Hey Everyone,

I’ve been watching a lot of economists and technologists ponder the impact of AI on jobs and it impacts software engineers and data engineers as well, including even machine learning researchers and scientists.

And I think this is worth dipping our curiosity into.

• Neil Thompson along with Maja S. Svanberg and Wensu Li from the MIT FutureTech, Martin Fleming from The Productivity Institute, and Brian C. Goehring from IBM's Institute for Business Value, have published a new article.

• This article introduces an innovative AI task automation model. It offers an end-to-end assessment framework, focusing on determining the necessary technical performance for tasks, defining the characteristics of AI systems capable of this performance, and making economic decisions about their development and deployment.

Read the 45 page Study

The Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory (CSAIL) conducts research in all areas of computer science and AI, such as robotics, systems, theory, biology, machine learning, speech recognition, vision and graphics.

Subscribe now

According to a study by MIT, it might not be so likely that AI is coming for your job. At least not in the way conceived in 2023 by the likes of Goldman Sachs, OpenAI itself and others.

Read Challenger, Gray and Christmas

Read more

Hey Everyone,

I’m always thinking about Snowflake and Databricks, the new breed of data analytics and data cloud platforms. Databricks has now made another significant acquisition of late.

Subscribe now

News of the acquisition comes after Databricks announced a $500 million funding round in September, with investors including Nvidia and Capital One. Databricks had also participated in Arcion’s funding in its Series A.

Databricks was last valued at $43 Billion. Databricks is a so-called white house data lakehouse provider and Arcion is an important piece to the puzzle. Databricks’ chief product is a data analytics tool, powered largely by an artificial intelligence method called deep learning, but there’s no data for the platform to analyze when a client signs up, Ghodsi said. Once Databricks integrates Arcion’s technology, it should be much easier for clients to add their data from software systems such as Salesforce, Workday, Oracle and so forth according to CNBC reporting with the CEO.

Arcion is Databricks’ first acquisition since acquiring MosaicML, an AI infrastructure startup specializing in training large language models, for $1.3 billion. With MosaicLM and Arcion Databricks is taking the next step to be IPO ready and take on Snowflake in the big leagues.

The way Databricks itself explains it is thus:

The new tie-up gives Databricks control over Arcion’s suite of no-code “connectors” that allow enterprise customers to replicate and ingest data from the multitude of different sources and applications they rely upon, such as Salesforce, SAP, and Workday, as well as transactional databases such as Oracle, MySQL, and Postgres.

Read Blog

• This will enable Databricks to provide native solutions to ingest data from various databases and SaaS applications into the Databricks Lakehouse Platform.

• Databricks is widely seen as a top contender for an initial public offering in the near future. This is likely in 2024.

• Prior to this, customers would need to use a third party tool like Informatica, Qlik or Alteryx to get the data into Databricks. Arcion will give the company a tool of its own.

This makes Databricks a lot easier to use in a self-serve manner. Arcion was founded in 2016 by Rajkumar Sen, a former employee at Oracle, and Miryana Joksovic, a former startup advisor. The $100 million valuation is after bonuses and incentives for its founders.

Databricks plans to increase Arcion’s staffing post-acquisition, Ghodsi added. TechCrunch added an easy explanation too:

The idea behind having a data storage solution like Databricks is about putting that data to work to do things like building dashboards, developing data applications and feeding machine learning models for AI, but to do that, companies need to move the underlying data into something like Databricks.

Data Lakehouse Platforms have emerged as the de facto standard for enterprise data and AI platforms. However, these data platforms are only as valuable as the data in them. Ingesting data from existing databases and applications remains complicated, fragile, and costly. With these two acquisitions DAtabricks feels more whole, and IPO ready. I could easily see Databricks being worth closer to $60 Billion once they go IPO. Snowflake with its tremendous growth trades at a rich premium.

This acquisition will enable Databricks to natively provide a scalable, easy-to-use, and cost-effective solution to ingest data from various enterprise data sources. Building on a scalable change data capture (CDC) engine, Arcion offers connectors for over 20 enterprise databases and data warehouses.

While Arcion is a fraction of the price of MosciacML, and is less directly focused on AI, it gives the company an essential ingredient in data ingestion, and the fact that the companies have been partners in the past, could help as they come together.

“We think we can make a lot of revenue on this particular acquisition,” Ghodsi said. “This is one where we think it’s going to be revenue-accretive.”

I’m excited to see what Databricks becomes, because in a sense it’s connected to so many data science and A.I. jobs in the real world and industry. Founded by the original creators of Delta Lake, Apache Spark™, and MLflow, Databricks is on a mission to help data teams solve the world's toughest problems.

Data Lakehouse Platforms have emerged as the de facto standard for enterprise data and AI platforms, hence it has a lot of hype behind it.

Arcion is the modern and scalable way to replicate enterprise databases, data warehouses, and cloud analytics platforms using change data capture (CDC) to power the next generation of AI applications. Enterprises utilize Arcion’s zero-impact, agentless CDC pipelines to replicate data with low latency and consistency. It sounds like a very complementary piece to Databricks at a pretty affordable price that might improve the accessibility and revenue generation of the company.

Hey Everyone,

There is something I find a bit perplexing. Given how ChatGPT has stalled in the second half of 2023 and GitHub Copilot is apparently a huge cash bleeding entity for Microsoft, why do Enterprise adopters of GPT-4 think they will be winners?

Subscribe now

Generative A.I. might be a gigantic bust. Apparently even GitHub Copilot has reportedly been costing Microsoft up to $80 per user per month in some cases as the company struggles to make its AI assistant turn a profit.

Even if you have ARR of $100 milion, what kind of loses is Microsoft able to tolerate here?

How much are you willing to bet on Generative A.I., even as it remains very costly to do so? What a gamble.

Microsoft’s GitHub Copilot, the automatic code-writing tool powered by OpenAI’s artificial intelligence, has exceeded $100 million in annual recurring revenue, vice president of product Mario Rodriguez said on Tuesday at an event in San Francisco, as reported by the Information.

But if you claim your tool can help coders, but the cost of the subscription is just one fifth of the cost, you have a problem. Can you imagine a loss of $80 per user? According to a Wall Street Journal report, the figures reportedly come from an unnamed individual familiar with the company, who noted that the Microsoft-owned platform was losing an average of $20 per user per month in the first few months of 2023.

As you might know, GitHub is using OpenAI's latest GPT-4 model to go way beyond auto-completing comments and code. Copilot X gets chat and voice support.

Productivity Gains for Lazy Habits

And are we dumbing down our software engineers and deskilling them in the process? GitHub Copilot was obviously designed to help developers write code more quickly and efficiently by suggesting code completions, functions, and entire files. Copilot has been praised by many developers for its productivity-boosting capabilities, but it has also come under criticism for its cost and its potential to promote lazy coding habits.

There are two subscription plans available for GitHub Copilot:

• Individual Plan: Priced at $10 per month or $100 per year.

• Business Plan: Priced at $19 per user per month.

• Microsoft says it’s losing $20 per month on average for each user of GitHub Copilot (so the real costs is likely between $30 and $40 to run the service per user).

Here is what people on Hacker News had to say about this.

Microsoft tried to hype OpenAI when it’s losing money in weird ways in reality. In June, unredacted court documents uncovered that Azure was actually proving to be far less profitable than AWS.

If AI costs exceed revenue for Microsoft, imagine how it will be for all of these Enterprise customers of OpenAI when they find out the truth of the benefits of GPT-4? It’s going to be expensive and a letdown.

GitHub Copilot uses GPT models from OpenAI to help solve coding problems. Microsoft has famously invested billions into the AI startup and has since gone on to use its models to power countless other AI tools across its businesses, but it might have been a horribly naive gamble.

While Microsoft’s acquisition of GitHub has proved fascinating, has it really been profitable? So on average, GitHub was losing more than $20 per user each month earlier this year. Some users were even costing GitHub as much as $80 a month. With approximately 1.5 million users, this translates to a substantial monthly loss for GitHub. It does not paint Microsoft in a very favorable light. They seem like corporate idiots in this report.

Can you imagine the media was talking about Bing AI being sentient earlier this year? It’s all very fuzzy. Microsoft’s work with AMD on their Athena chips was also very expensive, at least totalling 2 Billion. Artificial intelligence uses models that are capable of processing billions of parameters really quickly, but they require large amounts of resources like power and water for cooling. So there’s an environmental cost too for Microsoft’s reputation. All for so-called productivity gains?

ChatGPT hasn’t been what we might have hoped it would become. Even as OpenAI used Microsoft’s funds for its own designs. It will make OpenAI look very rich in 2024 with Enterprise contracts for the GPT-4 API, but how will all of this actually turn out?

In a related development, Microsoft announced the pricing for Microsoft 365 Copilot at $30 per user per month for enterprise customers. Microsoft is trying to recover some of the losses with an overpriced product for Enterprise customers.

Developing and maintaining Copilot is a costly endeavor. Microsoft has to pay for the computational resources that are needed to train and run Copilot's AI models. The company also has to pay for the engineers who develop and maintain the service. Azure and Microsoft’s AI bets might actually not turn out great if Generative A.I. isn’t as mature as they are claiming. We know it’s supposedly a nascent industry.

That GitHub Copilot has 100 million developers earlier in 2023 is a sign of the scale of these losses. When you are as profitable as Microsoft, Meta, Amazon or Google you can waste a lot of money is some pretty outrageous bets. The bandwagon of Generative A.I. might be in part, a faked campaign.

It appears that Microsoft’s Enterprise Copilot pricing strategy is aimed at mitigating significant losses similar to those experienced with GitHub Copilot. Even OpenAI doesn’t seem to understand how to turn a mobile app into a winner or the right form factor for ChatGPT’s ideal interface. It has not met with consumer appeal, outside of students, geeks and even some coders. In fact most young people, have never and will never use ChatGPT.

In a world with subscription overload and A.I. tool saturation, what does Microsoft expect to happen to its A.I. investments? Despite the partnership with OpenAI and the presence of more than 1.5 million users, GitHub Copilot remains unprofitable. They acquired GitHub five years ago, so it’s not as if they haven’t had time to optimize the business model.

Companies anticipate that generative AI will become cheaper over time, driven by innovations and improvements in processing efficiency. It’s unclear how much Microsoft’s internal A.I. Chips, dubbed Athena could help things along.

For those unaware, GitHub Copilot, which was first launched in June 2022, is powered by a generative AI model developed by GitHub, OpenAI, and Microsoft. This triad of A.I. greed might not actually be serving the best interests of developers who may become the guinea pigs who are training their replacement and deskilling themselves in the process.

Leave a comment

Currently, there are two paid subscription plans available for GitHub Copilot, which are: the Individual Plan, which costs $10 per month or $100 per year, and the Business Plan, which is priced at $19 per user per month. While this amount is nothing for most developers, what if they were paying the true cost that’s likely closer to $40 a month?

As good as GitHub Copilot is, its competitors are only going to get better, cheaper and more numerous. Microsoft might find that its acquisition of GitHub never really reaches the heights that it was supposed to.

Companies stuffing GPT-4 into their products might be in for a rude awakening if even Microsoft who had early-access to the tech looks so immature in their ability to even turn a profit. What happens to OpenAI when this all comes to light? GPT-4 might go down in the history of tech as the most over-hyped product in history.

Share

Hey Guys,

I’m so tired of working in Google docs or Microsoft products, and there’s going to be another option soon.

Zoom Video Communications Inc. is adding word processing to its suite of tools and experimenting with novel features for meetings as it faces steep competition from Microsoft Corp.’s Teams.

Read the Blog

I like how ambitious Zoom is with integrating AI into its products and expanding what they can do.

• Zoom is bringing a document creation tool, Zoom Docs, to its workspace portfolio.

• What sets the tool apart is the ability to include information and artificial intelligence-generated summaries from Zoom meetings, Chief Product Officer Smita Hashim said in an interview.

Zoom’s All in One Comms Tools with AI Just Got Serious

Zoom Docs is their AI-powered, modular workspace for docs, wikis, and work management, all in the Zoom platform. Coming in 2024, it will integrate with Zoom and third-party apps to help you easily create, collaborate, manage projects, and stay organized.

It’s sad to admit that this excites me more than Microsoft or Alphabet trying to stuff Generative A.I. into everything.

So the reason is, it sounds more AI-native from the get-go.

Within Docs, AI Companion can alter wording for tone adjustment, brainstorm ideas based on a prompt, and summarize or query content stored within the Zoom product suite. For example, AI Companion can retrieve information from Zoom Meetings and Team Chat to answer users' questions.

• The announcements made at the company’s Zoomtopia event broaden the company’s platform for users and boost market opportunities in the call center space.

• Zoom unveiled a step to fight back against Microsoft Teams, the all-in-one communications suite that has been taking business from Zoom’s web-based videoconferencing.

When Microsoft cloned Zoom with Teams, it felt like they were going head-to-head but this feels more like a competitor of Google Docs, with its own AI powered tools that are way more intuitive.

The company’s fiscal-year revenue jumped more than fivefold to $4.1 billion from 2020 to 2022 as Zoom became an essential tool during the pandemic. But in recent years Zoom hopped on the AI bandwagon pretty hard to try to impress. I get the feeling it is innovating but I’m not sure if its customers are ready to embrace all of this quite yet?

Zoom is trying to redefine the concept of a document – it’s no longer a static record but a dynamic repository of the information vital to a collaboration.

In a hybrid first and remote friendly reality of the workplace, how we interact and collaborate with docs has changed a lot.

Zoom has seen early success with its office phone service, which is now making about $500 million per year, and its customer-service center offering, which passed 500 clients. Zoom knows what it is doing, but how do you go up against Monopolies like Microsoft and Google? Even Salesforce hasn’t fared so well in the CRM and Slack space.

A bit on the Notion Side of Things

1. The product will be equipped with Zoom’s artificial intelligence assistant, AI Companion, and other AI capabilities that will help users draft, edit, summarize and change tones as well as include items from meeting discussions. It also could answer questions about the content in the document.

2. The shared documents will be integrated into Zoom’s platform so that users can work on a document from its meetings, chat, desktop and mobile apps. The product will be generally available in the spring of 2024, and Zoom said it is still determining pricing.

Zoom Docs is AI Native from Day One

Zoom Docs goes beyond basic functionalities and is embedded with advanced features to enhance the user experience. Since AI powers it, the tool can assist users with content creation and comprehension through accessible formats and refined search queries. The AI companion also simplifies composition and summarization, making writing content easier and faster for users.

Along with the Mail and Calendar (Verge) offerings launched during last year’s event, Zoom Docs is another step toward a full office suite alternative to Google Workspace and Microsoft 365, which both have started to integrate AI-powered tools of their own, dubbed Duet AI and Copilot, respectively.

Zoom has thus created a valid alternative to those monopolies that we might be getting tired of. Additionally, Zoom Docs users can customize layouts and workflows, organize data, track tasks, and manage schedules with the table blocks feature—all with one click. Table blocks assist users in arranging data by utilizing columns, filters, and grouping.

A lot of us actually use Notion more than Microsoft products now.

• Notion

• Zoom

• Google Docs

• Microsoft products

Zoom AI Companion Keeps getting Better

• The ability to generate ideas on a digital whiteboard and organize them into categories, helping teams get to work faster.

• Expanded access to Zoom AI Companion for certain education and healthcare customers 

I am a fan of Smita. There’s an elegance to the design of these products.

Another innovation is the Global Unified Search functionality, enabling users to find information across the Zoom platform. Now if only Zoom had acquired Neeva instead of Snowflake. Snowflake’s market cap will soon be 3X that of Zoom, so it’s hard to compare the two companies of course.

Zoom’s AI companion is evolving faster than most other such products in the Enterprise space I have noticed.

Zoom is entering its prime, it’s now or never for the platform.

Zoom’s strategy to revive growth is based on providing a wider suite of business tools beyond video meetings, including features from Workvivo, an employee communication service it agreed to acquire in April, 2023.

The way Microsoft bundles things like Teams is plain unfair and anti-competitive. Everyone has known that for years, but Zoom won’t get much sympathy. Zoom has met with regulators from the US and European Union over the last year to express concerns about the way Microsoft, the world’s largest software maker, gives preference to its product through design and price bundling.

I almost feel bad for Zoom because it really is more interested in real innovation than Microsoft or Google. Zoom executives — and many financial analysts — say most people simply like using its videoconferencing app better than Teams. But Zoom controlled only about 7% of the market for communication and collaboration software as of the first quarter of the year, while Microsoft topped 42%, according to industry analyst IDC.

The barrier is the price though to be honest. Zoom AI Companion is included in the price of the company’s paid subscription plans, which start at $149.90 per year per user. Microsoft charges an extra $30 per user per month for 365 Copilot, and Google has said it will do the same.

Zoom is an Remote work and hybrid-work native platform, but in a world of forced RTO, does it even matter? Or is Zoom still stuck in its pandemic glory days?

Share

Hey Everyone,

I’m sort of always watching news about Snowflake and Databricks, since they will obviously evolve into important companies and especially for data workers and tech engineers of various kinds.

Subscribe now

Data analytics and AI software maker Databricks has raised a Series I round worth more than $500 million, earning a valuation of $43 billion. Series I, when are you going public already guys! Snowflake has a $53 Billion valuation on the public market.

The Sequence finally wrote something that interested me, they recently wrote about how Nvidia is one of the surprise backers of Generative A.I. and it’s true:

• Hugging Face: NVIDIA was part of the recent $200 million round that pushed Hugging Face's valuation to $4 billion.

• AI21 Labs: NVIDIA invested in the recent $155 million Series C that valued the OpenAI competitor at $1.4 billion.

• Imbue: NVIDIA joined the $200 million Series B that valued the AI reasoning platform at $1 billion.

• Cohere: NVIDIA was part of the $270 million round that valued Cohere at $2.1 billion.

• Adept AI: NVIDIA participated in the $350 million round that valued Adept at $1 billion.

Nvidia is basically backing some of the most actionable startups in the future of A.I., agents, open-source and data analytics. That they are backing Databricks is really telling.

You will remember, Databricks last raised $1.6 billion in August of 2021 at a post-money valuation of $38 billion. All of these valuation slashing really is going to wreck havoc with the IPO market that has seemingly restarted after what amount to a 1.5 year break. I shudder to think of where ARM will end up price range wise.

Share

Databricks now thinks they are worth $43 Billion

Read the Blog

This funding values the company at $43 billion and establishes the price per share at $73.50. The series is led by funds and accounts advised by T. Rowe Price Associates, Inc., which is joined by other existing investors, including Andreessen Horowitz, Baillie Gifford, ClearBridge Investments, funds and accounts managed by Counterpoint Global (Morgan Stanley), Fidelity Management & Research Company, Franklin Templeton, GIC, Octahedron Capital and Tiger Global along with new investors Capital One Ventures, Ghisallo Capital Management, Ontario Teachers’ Pension Plan and NVIDIA.

• Revenue at Databricks rose 50% in the quarter that ended in July.

TechCrunch summarized it well here:

The Nvidia-Databricks connection is not hard to suss out — Databricks is leaning into its AI capabilities, built on its history of selling data and machine learning software. Nvidia is also riding high on AI-powered demand for its chips and software. There’s so much demand for Nvidia chips that some countries are working to secure supply for their own economies, for example.

I believe Databricks will go public in 2024 or 2025 when things have settled down in the IPO market and there’s more cash on the sidelines. There’s still significant macroeconomic uncertainty. But Databricks cannot afford to wait much longer (holding out almost as bad as Stripe).

Databricks likely to go IPO in early 2024

According to a report by The Information in late 2022, the company told its investors that it planned to launch an IPO by the summer of 2023.

Databricks and Snowflake have room to make important acquisitions in the Generative A.I. space should they choose to do so in the coming months and years. Snowflake acquired Neeva, likely more for the talent than the product.

Databricks have a lot of momentum though listen: As of 2023, in the second quarter ended July 31, its revenue run rate surpassed the $1.5 billion mark. Databricks also said that it has more than 10,000 customers globally, of which more than 300 are currently generating revenue at a pace of $1 million or more per year for its software and services.

Some of us forget Databricks was only founded in 2013, they are incredibly mature and profitable for a startup that is just ten years old. Snowflake isn’t anywhere near profitable and it’s still overvalued on the public markets.

Databricks has found a mean point in its internal validation though. With an effective revenue multiple 29x, the company appears a little expensive for the current market. But flashy startups like Snowflake and Databricks don’t come around that often, I mean look at Nvidia, it’s easily over-priced at this point even if you think it’s an A.I. chip monopoly which it won’t be forever by the way.

Defining Snowflake or Databricks is also tricky because they will have to evolve with the times. Founded in 2013 by the original creators of Apache Spark, an open-source unified analytics engine for large-scale data processing, Databricks is known for its “lakehouse” platform in the cloud — a combination of data warehouses and data lakes that unifies data, analytics and AI on a single platform so that customers can govern, manage and derive insights from enterprise data and build their own generative AI solutions faster.

I think we will always be comparing Databricks with Snowflake. Snowflake was founded just a year earlier, in 2012. In this case the duopoly that they are is likely good for innovation, at least now as they are smaller. As Databricks and Snowflake compete to own the end-to-end enterprise AI workflow, their continued moves and countermoves will spur innovation that enables more businesses to leverage generative AI—a win for the entire enterprise ecosystem.

Datawarehouse Architecture

“The commitment from long-term focused strategic and financial partners reflects Databricks’ continued momentum, the rapid customer adoption of the Databricks Lakehouse, and the success customers are seeing from moving to a unified data and AI platform,” said Ali Ghodsi, Co-Founder and CEO of Databricks. “Databricks and NVIDIA are building transformative AI technology, and we’re excited about the business value and innovation we can bring to our customers.”

• In the latest round, shares were sold at $73.50 a piece, roughly equal to where they were priced in 2021.

• The $5 billion increase in valuation is the result of new shares that CEO Ali Ghodsi said have gone to the 3,500 employees the company has hired in the past two years, as well as to investors. Headcount now sits at around 6,000.

Nvidia has been investing in bigger companies recently including Hugging Face, Cohere and CoreWeave are a few of the companies that Nvidia has backed at multibillion-dollar valuations.

• Hugging Face

• Databricks

• Cohere

• CoreWeave

• Synthesia

• Adept AI

• Imbue (Formerly Generally Intelligent)

• AI21 Labs

That’s quite the list Nvidia is accumulating. A bit like how Microsoft is backing OpenAI and Inflection. Nvidia is just a bit smarter in how it allocates its funding around ecosystems than Microsoft’s AI-assistant gamble that’s more about connections than who will actually end up winning.

I don’t myself see ChatGPT or Pi as being the end-game of A.I. or AGI chatbots by the way. We are on the first batter of the first inning. Meanwhile Google got Anthropic’s back.

Databricks is technically growing faster than Snowflake. And what if Nvidia could one day acquire them? What will Snowflake and Databricks become?

Leave a comment

Databricks vs. Snowflake

via Macrometa.

Hey Everyone,

I’ll be writing slightly less than my usual out for a few weeks in the Summer. This news really got my attention!

Just as open-source A.I. is starting to have its Linux moment, major companies and BigTech are pivoting their approach and angle in how agnostic, how open or closed they are going to be.

Subscribe now

MosaicML

MosaicML has been working on reinventing what a database is via artificial intelligence. It’s also a bit like a Hugging Face, where it enables you to easily train and deploy LLMs and other generative AI models on your data, in your secure environment and deploy LLMs.

It’s definately a “democratization of AI” play. Databricks and Snowflake are both super fun to follow and are diversifying impressively well.

I really did not expect this. Databricks, the Data and AI company, today announced it has entered into a definitive agreement to acquire MosaicML, a leading generative AI platform. Together, Databricks and MosaicML will make generative AI accessible for every organization, enabling them to build, own and secure generative AI models with their own data. The transaction is valued at approximately $1.3 billion, inclusive of retention packages.

In late June, 2023 it definately feels like M&A and IPOs are set to accelerate for the second half of 2023.

But wait a second, MosaicML is only a 2 year old startups, and this means: wait for it - the valuation is wack, 2-year-old startup for $21 million per employee! Wow that’s something. It’s a premium Databricks must think is worth it and a very fast exit.

With enterprises looking to leverage LLMs, it’s a hot sector and especially along the lines of open-source. I mean they just released MPT-30B. It’s sort of a big deal. Prior to this, MosaicML had raised just under $64 million from investors that included DCVC, AME Cloud Ventures, Lux, Frontline, Atlas, Playground Global and Samsung Next.

LLMs are putting pressure on companies to adapt fast like never before. This is a good move by Databricks I think.

Databricks snaps up MosaicML to build private, custom machine models

Naveen Rao, [left] MosaicML co-founder and CEO, and Hanlin Tang, co-founder and CTO. The company's training technologies are being applied to "building experts," using large language models more efficiently to handle corporate data. 

Think about it though, this young startup is staffed with semiconductor veterans, has built a program called Composer that makes it easy and affordable to take any standard version of AI programs such as OpenAI's GPT and dramatically speed up the development of that program, the beginning phase known as the training of a neural network. 

Generative AI for All

The implications here are Databricks fortifying itself with LLMs, stamping its place in open-source and doing a rather expensive acqui-hire.

And I love how crazy it is and feels. But think about it, there’s a lot of synergy here. Databricks' core platform helps customers store and sort incoming data from different sources in their own cloud clusters, while MosaicML offers tools to spin up custom AI models at low cost.

So there’s this crazy active ecosystem now running simultaneously all over the world dealing with open-source innovation at the intersection of A.I. Hugging Face and recent startup wonder-kid Mistral AI are just the tip of the iceberg.

Democratization of AI Buzzword Play

Overpriced?

I guess what people are talking about is the wackly wild valuation. Notably, its last investor-round valuation was just $222 million — meaning it’s leaped 6x with this exit, a remarkable price that really does underscore just how frothy the AI market is right now, as well as the demand for talent and tech in the space.

Companies aligning themselves to Open-source is the new cool in A.I., as Meta’s stock share price has found out and as Amazon is investing in. Meanwhile Microsoft and Google still feel a bit in the old world of closed source with hype trains of OpenAI vs. Gemini, and so on.

Databricks Upgrade in LLM Credibility

There is a lot of synergy here though, Databricks immediately elevates itself in terms of LLM credibility. The deal will see MosaicML become a part of the Databricks Lakehouse Platform, providing generative AI tooling alongside the Databricks’ existing multicloud offerings, which include integration, storage, processing, governance, sharing, analytics and AI-related services.

MPT-30B looks really impressive, I’m not going to lie.

The startup this year introduced cloud-based commercial services where businesses can for a fee both train a neural network and perform inference, the rendering of predictions in response to user queries. Too good, to pass up on, as it turns out.

However, the more profound element of MosaicML's approach implies that whole areas of working with data --  such as the traditional relational database --  could be completely reinvented. This alone is getting a lot of attention it would seem.

The quality of its customer base also really stands out, these are some pioneering and credible teams. Its customers include the Allen Institute for AI, Generally Intelligent, Hippocratic AI, Replit and Scatter Labs.

In my opinion, this elevates a speculative valuation for Databricks that has yet to IPO, at around $42 billion. No doubt it’s going to be a monster IPO when it happens. But who knows in this world, Google acquiring Stripe is a rumor that peeked my interest, so anything can happen. Databricks itself would be a nice acquisition for someone like Amazon.

MosaicML was an Industry Leader

If you think of the open-source buzz in 2023, MosaicML was a major player. MosaicML is known for its state-of-the-art MPT large language models (LLMs). With over 3.3 million downloads of MPT-7B and the recent release of MPT-30B, MosaicML has showcased how organizations can quickly build and train their own state-of-the-art models using their data in a cost-effective way.

It was bringing LLMs to the masses and smaller enterprises and companies like few others could offer.

The announcement Tweet is sort of epic and it really tells you the pace of LLMs being integrated in Enterprise and data companies now.

Given how important LLMs are going to be across the board, this could be a monumental acquisition looking back historically speaking.

The Future of the Database

Unlike a traditional relational database, such as Oracle, or a document database, such as MongdoDB, said Rao, where the schema is preordained, with a large language model, "the schema is discovered from [the data], it produces a latent representation based upon the data, it's flexible." And the query is also flexible, unlike fixed lookups into a database such as SQL, which dominates traditional databases.

The San Francisco-based vendor was founded in 2021 and had raised $37 million in venture capital. It seems like peak open-AI LLM hype, so well timed and well played.

1 Month ago

So did MosaicML have a data moat for its open-source enterprise LLM on-boarding?

Databricks said that the entire MosaicML team will join Databricks after the deal closes — a retention deal that likely could be one reason for the large price tag here.

MosaicML had made a name for itself with performance achievements by demonstrating its prowess in the MLPerf benchmark tests (ZDnet) that show how fast a neural network can be trained. Among the secrets to speeding up AI is the observation that smaller neural networks, built with greater focus, can be more efficient. 

MosaicML seems like a way bigger fish than Snowflake bagging Neeva. But either way, databases and search are both really trending with LLMs it turns out.

Thanks for reading!

Subscribe now

Hey Everyone,

I have a feud in my head with Microsoft in recent times, but that isn’t to say that Microsoft’s reckoning in no-code isn’t significant. In fact, with RPA and No-code, Microsoft appears to be one of the big winners.

Microsoft Build is an event designed for developers – showcasing tools and services that developers can use now and with an eye toward the future. With AI dominating the conversation in the technology space, it only makes sense that it plays a key role in this year’s event. (May 23rd to 25th, 2023)

1. Windows Copilot

First-of-its-kind operating with native AI capabilities that will put the applications of AI in the hands of the masses!

2. Bing as ChatGPT Plugin

With Bing plugin, ChatGPT will have access to built-in search engine to provide timelier and more up-to-date answers from the web.

3. Microsoft Fabric

An end-to-end, unified analytics platform that brings together all the data and analytics tools that organizations need. Fabric integrates technologies like Azure Data Factory, Azure Synapse Analytics, and Power BI into a single unified product!!

It’s actually Microsoft Fabric that is among the more interesting things here. More on this later!

4. Copilot in Power Pages

Generate text, build detailed forms and chatbots—all in a matter of minutes. Leverage the power of natural language input and intelligent suggestions for unmatched efficiency!!

So GPT-4 yes will enable you to make an app or a web page tailored to your needs in minutes.

5. Windows Terminal meets AI

Users of GitHub Copilot will be able to take advantage of natural language AI both inline and in an experimental chat experience to recommend commands, explain errors and take actions within the Terminal application.

Watch the KeyNote

Microsoft says that it will build all its Copilot and Bing Chat plug-ins with the same standard that OpenAI uses for ChatGPT. It’s clear here Microsoft is a bit jealous of what ChatGPT will become and how OpenAI will eventually split from it once it pays back its mighty funding forward.

Everything in 7 Minutes

Register for different Live Sessions

Subscribe now

Read more

Hey Everyone,

As a lot of coders and professionals have embraced ChatGPT, I’m wondering what impact the app will have on industries and how we use the interface.

The ChatGPT app syncs your conversations, supports voice input, and brings our latest model improvements to your fingertips. As usual it’s coming to Android a bit later.

I’ve never seen so many GPT gimmicks in my life on app stores and in specialized LLM models that come out every week, heck every day. If LLMs are supposed to be a General Purpose Technology, I’d like to see hard evidence that it is the case. I’m not actually seeing much of that, it will take years for this to trickle down into many of our white collar tasks at our jobs.

ChatGPT is a powerful artificial intelligence tool that can be used to write code in programming languages such as C++, Java, and Python. But with so many options and chatbots, how much further do we have to go than GitHub Copilot for Business?

Just for coding, there’s a lot going on here.

Share

If Siri is dead, and Alexa was failure, who's to say the shelf life of this ChatGPT craze? Like its desktop counterpart, the ChatGPT app allows users to interact with an AI chatbot to ask questions without running a traditional web search, plus get advice, find inspiration, learn, research, and more.

I’m finding the chatbot crazy kind of dumb, because this is literally the 2nd time on the internet we’ve had to go through this. Does having ChatGPT in my pocket make me even a better person or more productive? Honestly this isn’t being augmented with A.I.

Voice Interface Integration

Okay so now we get to something actually interesting. The app is also integrated with Whisper, OpenAI’s open source speech recognition system, to allow for voice input.

Exclusive Access

ChatGPT Plus subscribers will be able to access GPT-4’s capabilities through the new app, OpenAI says, in addition to receiving early access to new features and faster response times, the company notes in its announcement. The subscription offering was launched in February and costs $20 per month for the upgraded features, including access to ChatGPT even during peak times.

Only another subscription for me to pay, every month. Google never asked me for one. But Microsoft, it thinks I need several.

The warnings have been upgraded.

So are all these APIs and chatbots going to amount to something? White collar workers want to know.

OpenAI had been rumored to be working on a mobile client, according to a February article by Semafor, which the company had then declined to comment on.

Coding

Generative A.I. might at least power coding auto-complete and game development forwards a bit in the next decade.

• Tabnine (2019)

• GitHub Copilot (2021)

• Replit Ghostwriter (2022)

• Amazon CodeWhisperer (2022)

• Codeium (2022)

• SourceGraph Cody (2023)

• CodeComplete (2023)

• FauxPilot (2023)

• Tabby (2023)

Some programmers certainly like to dabble with code in ChatGPT.

Infographic from Pragmatic Engineer:

Good luck to OpenAI trying to trademark “GPT”.

The ChatGPT app’s launch comes at a time when major tech companies, including Google, Microsoft, and Facebook, are innovating and building AI products at a much faster rate than a few years ago.

While ChatGPT stole the show in terms of the hype around LLMs, I struggle to see how they will maintain their enthusiasm. As more specialized and open-source options come out, the utility of one interface begins to slow down as the rest of the world catches up, and they are catching up fast.

Subscribe now

Hey Guys,

So I’ve been thinking a lot more about the impact of A.I. in cybersecurity, data-privacy regulation with regards to ChatGPT, its impact on the future of law itself and how we might be able to regulate it.

Subscribe now

So Microsoft thinks it can leverage GPT-4 for a Copilot experience everywhere, Google has the same approach to leveraging PaLM.

Generative AI for cybersecurity is just in its infancy and it’s a battle of Cloud giants.

Read Google Cloud Blog

At the RSA Conference 2023 yesterday on April 24th, 20223, Google announced Cloud Security AI Workbench, a cybersecurity suite powered by a specialized “security” AI language model called Sec-PaLM. An offshoot of Google’s PaLM1 model, Sec-PaLM is “fine-tuned for security use cases,” Google says — incorporating security intelligence such as research on software vulnerabilities, malware, threat indicators and behavioral threat actor profiles.2

We know that Google is working on dozens of A.I. products to take on Microsoft and its threat to its search (advertising) empire and monopoly. As the advertising duopoly is under pressure by Amazon and ByteDance in ads, Microsoft and Apple see a huge opportunity as better walled gardens to move in for the kill.

Cloud Security AI Workbench spans a range of new AI-powered tools, but Google does not have the deep experience in cybersecurity Microsoft does. However, let’s think about this, Google purchased3 Mandiant in 2022 for $5.4 billion. Generative A.I. is like watching an A.I. arms race or AI supremacy contest in the Cloud.

Mandiant’s Threat Intelligence AI within Google Cloud, just got a significant upgrade. Sec-PaLM has already faced down hundreds of the most advanced cybersecurity threats and relies on this experience to prevent the same attacks succeeding elsewhere. According to Google, it combines the company’s “world-class threat intelligence” with advanced incident analysis to stop malware infections in their tracks. 

Share

What is Sec-PaLM?

Sec-PaLM it’s a linguistic model (LLM) who has been trained through the knowledge of Mandiant on vulnerabilities, malware, threat indicators and malicious agent profiles. A database of billions of entries that make this AI in cybersecurity expert.

I’ve been bullish on LLMs trained on proprietary data and knew that we’d be seeing some of them in 2023 and 2024. This is fairly good example of that.

It’s pretty neat: Generative A.I. can unlock a lot of value in cyber-defense at a time when there is a significant lack of cybersecurity professionals in the labor force vs. the demand. Mandiant Breach Analytics for Chronicle is based on years of accumulated threat intelligence from Google’s Mandiant team and can automatically alert customers to ongoing, active security breaches. It leans on Sec-PaLM to help contextualize and respond immediately to such attacks.

But just like Security Copilot, this also enables cybersecurity professionals to be upgraded with A.I. in that all too familiar buddy-system formula. Sec-PaLM will assist customers of Chronicle, Google’s cloud cybersecurity service, in searching security events and interacting “conservationally” with the results. Users of Google’s Security Command Center AI, meanwhile, will get “human-readable” explanations of attack exposure courtesy of Sec-PaLM, including impacted assets, recommended mitigations and risk summaries for security, compliance and privacy findings.

Fair enough, but relevant. I’m looking forward to Google’s I/O event because I’m expecting to hear some A.I. swag of new products. Alphabet cleary wants to get this all right. Because it if doesn’t it will hurt Google Cloud growth vs. Azure.

If you consider that VirusTotal Code Insight, the first tool in the Cloud Security AI Workbench, is only available in a limited preview at the moment, this is all a very ne play for Google.

Share Datascience Learning Center

Thirteen years since Google Authenticator was released, it’s getting a vital upgrade. I think Microsoft’s Security Copilot and Google’s Sec-PaLM are evidence that A.I. is only getting more involved and actionable in lucrative and important industries like cybersecurity. Sec-PaLM is still somewhat untested and we don’t actually know much about it or its efficacy in the real world.

Google is trying to tell us that its Security AI Workbench’s capabilities are getting better and you presume it has done its due diligence. But time and the market will tell.
”Supercharging” everything with Generative A.I. is not going to be great, across the board. There is a lot of hype in the Cloud Wars fighting over lucrative cash-cows like Search advertising.

It would be much better if there were new players in the game. But even those tend to be back by the super-powers, such is the fallacy of innovation in Silicon Valley where monopoly capitalism winners buy even the pawns entering the board. So much for a startup meritocracy in search, and increasingly more and more use-cases.

Google was badly late to the cloud and Google Cloud is basically burning cash trying to catch Azure and AWS. You cannot blame them for trying but its all a bit lame as their advertising duopoly won’t last forever. Google and Meta are very poorly diversified companies when compared to Apple or Microsoft, or even Amazon for that matter. Amazon has been increasing their ads revenue in leaps and bounds and I don’t see why Microsoft or Apple won’t succeed soon in doing the same thing.

All that being said, PaLM is only going to get better.

Cloud giants will just pick and choose the best Generative A.I. and Quantum computing companies that add value to their positioning. It’s a bit unfortunate. Knowing this, Anthropic AI will be farmed by Google the way OpenAI has been farmed by Microsoft, even with GPT-5 in regulatory limbo.

Hey Everyone,

I have been impressed with the frequency and variety and size of Microsoft’s cybersecurity acquisitions throughout the years. See list of all acquisitions here. I was concerned when ChatGPT is being used for malware, phishing and Generative A.I. is being used as the new most common type of fraud and scam, voice cloning of relatives.

Subscribe now

Some Seniors show a reduced ability to tell the difference between a real relative and an A.I. generated clone of their voice. So while Microsoft has spent $13 billion on OpenAI, which is causing a more risky internet and more frequent phishing and more sophisticated fraud and cybersecurity risks at scale, since ChatGPT is pretty good at code and malware, Microsoft also has come up with the solution!

If you are the company responsible for more security issues, you stand indeed the most to benefit from creating the solution, isn’t it?

Microsoft in their blog, indeed say that today the odds remain stacked against cybersecurity professionals. But no need to fear guys, Microsoft Security Copilot is here.

As you might expect, Microsoft Security Copilot is a new assistant for cybersecurity professionals, designed to help defenders identify breaches and better understand the huge amounts of signals and data available to them daily.

When you are the very actor that brought ChatGPT to the world, which increases risk, that you are the solution to the malevolent possibilities of Generative A.I. around cybersecurity is a bit corrupt and daunting. Such is the world of Monopoly Capitalism, such is the playbook of the centralization of A.I.

• Powered by OpenAI’s GPT-4 generative AI and Microsoft’s own security-specific model, Security Copilot looks like a simple prompt box like any other chatbot. You can ask “what are all the security incidents in my enterprise?” and it will summarize them.

• Microsoft says it’s making use of the 65 trillion daily signals Microsoft collects in its threat intelligence gathering and security-specific skills to let security professionals hunt down threats.

I finally feel safe with Copilot.

Share

At a time when the supply of Cybersecurity professionals isn’t meeting demand, Microsoft assures us that Microsoft Security Copilot is designed to assist a security analyst’s work rather than replace it — and even includes a pinboard section for co-workers to collaborate and share information.

Microsoft Security Copilot is the first security product to enable defenders to move at the speed and scale of AI. Security Copilot combines this advanced large language model (LLM) with a security-specific model from Microsoft. Microsoft will have advanced early access to GPT-5, just as it did with BingAI and GPT-4.

Microsoft thinks it is at the speed of A.I. now, I’m sure all those voice cloners and phishing experts using ChatGPT are going to be quite concerned now.

The new AI security tool, which can answer questions about vulnerabilities and reverse-engineer problems, is now in preview. And it’s all a bit interesting how A.I. will be increasingly used in phishing and cybersecurity defense.

Still this is a specialized LLM that could hopefully assistant cybersecurity professionals to reduce incidents and minimize the damage of vulnerabilities. When Security Copilot receives a prompt from a security professional, it uses the full power of the security-specific model to deploy skills and queries that maximize the value of the latest large language model capabilities.

The service will work with Microsoft security products such as Sentinel for tracking threats. Microsoft will determine if it should add support for third-party tools such as Splunk based on input from early users in the next few months. Microsoft is building multiple subscription revenue streams from its aggressive foray into LLMs with its sponsorship of OpenAI.

Continuous Learning of LLMs in a Specific Niche

Security Copilot is a closed-loop learning system, which means it’s continually learning from users and giving them the opportunity to give explicit feedback with the feedback feature that is built directly into the tool. As we continue to learn from these interactions, we are adjusting its responses to create more coherent, relevant and useful answers.

The Global Shortage of Talent in Cybersecurity

According to the news and announcements, there is a global shortage of skilled security professionals, leading to an estimated 3.4 million openings in the field.  Voice cloning fraud is exploding and will only get worse as global populations begin to age in the coming demographics winter.

Microsoft is using information from the Cybersecurity and Infrastructure Security Agency, the National Institute of Standards and Technology’s vulnerability database, and Microsoft’s own threat intelligence database to improve the product.

Dennis Chung, Chief Technology Officer, Microsoft Singapore, shared, "The rapidly evolving cybersecurity landscape demands that organizations continuously strengthen and maintain their security measures to effectively combat increasingly sophisticated threats and ensure the safety of their people and business. Our Microsoft Security Copilot brings the power of AI to individuals to drive innovation and scale at pace, as they create a secure digital environment for all. "

Read more

Hey Everyone,

Google is so behind in LLMs it’s bound to make the company make some desperate choices and in the first week of April, 2023 we got the confirmation.

Google plans to add conversational artificial-intelligence features to its flagship search engine, Chief Executive Officer Sundar Pichai said, as it deals with pressure from all that ChatGPT and the Generative A.I. boom entails. Google has tried hard to make announcement that it’s busy working on LLM upgrades to its products, but Microsoft has been moving faster with the help of its vassal company, OpenAI (who isn’t very open any longer).

Subscribe now

Let’s dive a bit into this news:

Read more

Hello Everyone,

Transformers will change Healthcare, education and the life sciences, but how?

I’m very interested in how companies like Salesforce and Canva are implementing Generative AI in their business. It’s even more interesting for growing platforms like Doximity and Roblox.

Doximity is basically the LinkedIn for Doctors and Nurses. Doximity is an online networking service for medical professionals. Launched in 2010, the platform offers its members curated medical news, telehealth tools, and case collaboration.

Roblox is basically the gaming, coding and entrepreneurship platform for Pre-Teens and younger teens. Roblox is an online game platform and game creation system developed by Roblox Corporation that allows users to program games and play games created by other users.

With ChatGPT hype, many companies have successfully built ChatGPT like chatbots into their platforms and many will do this a lot better than Microsoft has.

Doximity GPT

Doximity, a digital platform for medical professionals, rolled out a beta version of a ChatGPT tool for doctors that helps streamline some of their time-consuming administrative tasks, such as drafting and faxing preauthorization and appeal letters to insurers.

In a Feb. 9 earnings call, Doximity cofounder and CEO Jeffrey Tangney said the tool is an integration that works with ChatGPT, and can be accessed, for free, via DocsGPT.com. 

Try it out here

Thanks to the that alterted me to this launch. Bloomberg mentioned this about a week ago but I did not see it.

Doximity engineers began working on the beta site over the holidays, Tangney said, and the tool was announced this week during the earnings call.

Obviously medical chatbots have immense value considering the demographics winters we will soon be entering. We will have to DIY our aging relatives and their situations due to the healthcare cost crunch.

The fiercehealth.com PR isn’t very in-depth but they say:

"We had an opportunity to speak with a number of physicians and a very familiar use case quickly bubbled up: doctors still handle a lot of paperwork and much of it is still sent via fax machines," he said.

At least 70% of healthcare providers still exchange medical information by fax, according to federal officials. Clearly Generative A.I. and healthcare have a fantastic future together. One would indeed expect Teladoc, Amazon, Doximity and Google Health to implement interesting A.I. tools and LLMs for both patients and medical professionals.

Share

TL;DR

The PhilaVerse says:

The three key points:

• DocsGPT is an integration with ChatGPT, OpenAI’s large language model, that works with Doximity’s free fax service.

• The tool features a library of medical prompts that can generate text within your conversation with a chatbot (a similar experience to the regular ChatGPT).

• Doximity warns that the technology has obvious limitations and the system should be used judiciously.

Roblox GPT

Given that Roblox has users and creators who are in an A.I. native generation makes sense to have Generative A.I. tools they use directly in the platform.

About Roblox GPT-3

A Roblox experience that demonstrates some capabilities of GPT (GPT-3 now, and GPT-4 when available). Here is a let's play video of this application by Dr. Alan D. Thompson. (see the video above).

How Transformers and LLMs will be integrated into niche platforms is where the real magic happens. I’m awaiting more news on Salesforces’ EinsteinGPT.

Salesforce and March Launch of EinsteinGPT

As customer relationship management (CRM) becomes more data-driven, the role of predictive analytics has grown and become more important. I think I’m also bullish on China’s Tech sector implementing their own versions of ChatGPT as they tend to be better at bringing a product into a state of actual utility for consumers.

So I’ll be watching the likes of Baidu, ByteDance, Alibaba, Tencent, Netease and others closely in this regard. We know they all have offerings around this in development and very near launch.

GPTZero in the Classroom

The obvious interesting of chatbot plagiarism for the education sector is also something I’m watching. GPTZero is a good example. Here is their latest update. So during the past month, the founder says their research team (currently six ML researchers from Princeton, Yale, McGill, and the University of Toronto) have worked extensively to address this problem, and train an AI detection model for student articles, and from unique datasets both collected and from our partner organizations. It’s getting better training for student paper use cases.

Some of the 20 A.I. tools Google will launch in 2023 and 2024, see their I/O conference in May/June for more details I believe will be impactful. I’m particularly excited about their Github Copilot competitor around PaLM.

Subscribe now

Read more

Hey Everyone,

Anthropic A.I. have their Claude and are seeking more funding. With BuzzFeed becoming a meme stock on the back of ChatGPT, up 400% in 2023, it makes you wonder. Then the Generative A.I. Crowed.

In the case of Anthropic, the New York Times Friday cited two people with knowledge of the matter as saying it’s zeroing in on a $300 million round of funding. Character AI, meanwhile, has told investors it’s seeking $250 million in fresh funding, according to a report Friday in The Information.

Now it seems Baidu as reported by WSJ and Bloomberg, means Baidu will launch its own bot in March, 2023 in China.

Share

Karen is right, it was only a matter of time. And I’m surprised ByteDance hasn’t hit this harder. Mobile search on Douyin and sister app TikTok has already gotten pretty high, with its own ChatGPT it could be quite powerful a window to take on Search giants like Google and Baidu. Microsoft is itself expected to integrate ChatGPT into Bing in March, 2023 as well.

Then there’s Google’s Sparrow likely in May, and some evidence that Anthropic’s Claude is more safe than ChatGPT and able to personalize answers a bit better. So what’s up with Baidu’s efforts?

Not sure I found the Bloomberg coverage enlightening.

Baidu Inc. is planning to roll out an artificial intelligence chatbot service similar to OpenAI’s ChatGPT, according to a person familiar with the matter, potentially China’s most prominent entry in a race touched off by the tech phenomenon. Baidu is a pretty interesting pure-play A.I. company.

I don’t agree they are up nearly 17% so far this year, but the 2023 bounce back has been partly just seasonal, even as Earnings isn’t great and certainly not yet time for the Chinese re-opening to be incredible.

Read more

Hey Everyone,

With the popularity of ChatGPT and the promise of GPT-4, Microsoft announced a new multiyear, multibillion-dollar investment with the artificial intelligence lab OpenAI.

This multi-year, multi-billion dollar investment from Microsoft follows their previous investments in 2019 and 2021, and will allow OpenAI to attempt to continue their “independent research and develop AI that is increasingly safe, useful, and powerful.”

From the point of view of Microsoft, this is their attempt to accelerate Generative A.I. adoption into their software and Azure AI services.

Upgrade to Paid

• Supercomputing at scale – Microsoft will increase our investments in the development and deployment of specialized supercomputing systems to accelerate OpenAI’s groundbreaking independent AI research. We will also continue to build out Azure’s leading AI infrastructure to help customers build and deploy their AI applications on a global scale.

• New AI-powered experiences – Microsoft will deploy OpenAI’s models across our consumer and enterprise products and introduce new categories of digital experiences built on OpenAI’s technology. This includes Microsoft’s Azure OpenAI Service, which empowers developers to build cutting-edge AI applications through direct access to OpenAI models backed by Azure’s trusted, enterprise-grade capabilities and AI-optimized infrastructure and tools.

• Exclusive cloud provider – As OpenAI’s exclusive cloud provider, Azure will power all OpenAI workloads across research, products and API services.

OpenAI’s evolution with GPT-4 will be particularly symbiotic now with Microsoft, that is expected to own a 49% stake in OpenAI moving forwards.

Sam Altman has become a bit of a cult-like figure on Twitter of late, when Elon Musk was a co-founder of OpenAI, it’s no surprise those ChatGPT tweets went so viral.

OpenAI however is still trying to wax idealistic, though how much I buy it remains mixed, it says: “In pursuit of our mission to ensure advanced AI benefits all of humanity, OpenAI remains a capped-profit company and is governed by the OpenAI non-profit.”

OpenAI is not even open about its research so hardly a leader in the democratization of A.I. It is entirely commercialized now practically like a research wing of Microsoft. What DeepMind is to Google, OpenAI has become for Microsoft.

Yet even Microsoft is claiming fair use in its AI for Good contribution to the world. It says: “Since 2016, Microsoft has committed to building Azure into an AI supercomputer for the world, serving as the foundation of our vision to democratize AI as a platform.”

Both entities claim to be into democratizing A.I, however their association raises more antitrust and centralization of A.I. questions in the Cloud, than good steps to empowering open-source, accessibility or fair use with proper A.I. regulations.

AI Supremacy

Google AI's Great Comeback of 2023

Hey Everyone, For me in 2023 with Artificial Intelligence, Google is the elephant in the room. Let me explain my thinking. While Microsoft is a highly diversified company with an excellent Microsoft Research group, it’s not an AI-first company at the level of pure-play specialty that Alphabet is. A.I. integration isn’t A.I. innovation…

Read more

3 years ago · 6 likes · 1 comment · Michael Spencer

Can Microsoft Take on Google Search?

You can see and vote in my poll about this here.

Some believe Bing with ChatGPT like we will see later in 2023, could lead to Google losing some of its search monopoly that is a $149 Billion search Ad business. Bing is approximately only $12 billion, but what will it become with ChatGPT and this OpenAI partnership?

While ChatGPT is popular, who knows how well a $42 pro monthly subscription will do.

Microsoft said the renewed partnership will accelerate breakthroughs in AI and help both companies commercialize advanced technologies. That sounds like the voice of Satya Nadella himself. I just don’t know if this is actually a good thing for A.I.

Microsoft was rumored to be investing (Semafor) $10 billion into OpenAI, the creators of popular AI tools like ChatGPT and DALL-E 2. We can only assume they got the deal they wanted and that OpenAI will be able to commercialize GPT-4 even better now.

Microsoft’s partnership enables it to capitalize on OpenAI’s technology. Presumably to be relevant or even a first-mover in Generative A.I. adoption. As good as Microsoft Research is, it’s not at a level of operationalizing its research very well like a DeepMind is beginning to get better at, so it needed a more specialized LLM lab like OpenAI to do some of the commercialization role.

Many of the A.I. labs now pretend to have an A.I. for Good ethos or mission statement, but how much of it is even true or just marketing and PR? This makes the entire community less trusting.

What indeed are your values Sam? Anthropic A.I. seems to have its own version of LLM laws in terms of A.I. alignment, by releasing the demo prone to errors a lot of damage has been done to the reputation of consumer A.I. and how tools are used for manipulation online.

Selling out to Microsoft is an appealing short-term injection of cash, but what will it do to the so-called Open ecosystem? The future of LLMs might become more secret, with more A.I. labs selling out. Stability A.I could become the underling of Amazon’s AWS, it’s all almost predictable.

ChatGPT automatically generates text based on written prompts in a fashion that’s much more advanced and creative than the chatbots of Silicon Valley’s past. What happens when GPT-4 powers even more powerful text-to-code and text-to-video capabilities? Is the consumer or the internet user really a winner in such a scenario?

Google Ads was trending on Twitter a few days after their epic 12,000 layoffs. Clearly Microsoft has an opportunity to hurt Google search and maybe even Google Cloud’s growth with AzureAI’s coming-out party with OpenAI integration and GPT-4’s grand unveiling later in 2023.

There is a sense that LLMs are transformative, just five years after the first paper. What will Generative A.I. be able to do in the next five years? As Microsoft software, Canva, Adobe, ByteDance and others upgrade themselves with Generative A.I. tools what will software and apps and no-code be capable of in just a few years time?

Microsoft said:

These innovations have captured imaginations and introduced large-scale AI as a powerful, general-purpose technology platform that we believe will create transformative impact at the magnitude of the personal computer, the internet, mobile devices and the cloud.

I’m not so sure I’d put it that way, but there is at least some enthusiasm and optimism that we are heading in a good direction for A.I. adoption at scale with some actual utility in how ChatGPT can be used and how prompt-engineering has capabilities of chaining different tools together in innovative ways to accomplish tasks in more automated and even creative ways.

The deal will see Microsoft increase its investments in the development and deployment of supercomputing systems to assist OpenAI’s research. When you need supercomputer prowess, it’s not a very open system, I’ll not be the first to point out. A.I. is not being “democratized”, it’s making very powerful elitist corporate entities even more powerful to supposedly take on their counterparts in places like China, with even more U.S. National Security and Defense collaboration.

It’s not the kind of technological capitalism that typically has a good ending where Silicon Valley, Venture Capital power is tied to Government, where the consumer has no choice to opt-in, they just must just to keep up. It’s not heading to the kind of fair internet I’d want my children or grandchildren to be using.

There are dystopian linings in the great collaborations of the space. More A.I. labs will pursue the OpenAI game-plan, of selling out to the highest bidder. Microsoft says it will use the OpenAI partnership to develop supercomputing systems that aim to advance independent research and development for AI capabilities, which in a press release, OpenAI claims “is increasingly safe, useful, and powerful.” Only Google, Meta and Amazon will be able to keep up, along with Nvidia, ByteDance and a small handful of companies world-wide.

This winner-takes-all capitalism is a dangerous game Microsoft is playing. OpenAI seems only too pleased to become a pawn in the ecosystem. OpenAI has used this infrastructure to train its breakthrough models, which are now deployed in Azure to power category-defining AI products like GitHub Copilot, DALL·E 2 and ChatGPT. Both of which are just new subscription revenues for Microsoft, one $19 and the other $42.

Do you have $60 to donate to the future of A.I.? It’s not really negotiable.

All 🙏🏾 pray to the altar of A.I. because the Azure OpenAI service features a number of AI models made by OpenAI including GPT-3.5, Codex, and DALL-E, so businesses and developers can utilize these systems in their own apps and workloads.

Thanks for reading!

Subscribe now

Hey Guys,

GitHub has launched Copilot for Business. The difference? GitHub Copilot for Business is officially here with simple license management, organization-wide policy controls, and industry-leading privacy—all for $19 USD per user per month.

GitHub Copilot, GitHub’s service that intelligently suggests lines of code, is now available in a plan for enterprises months after launching for individual users and educators.

Microsoft is being so entirely generous in its AI-as-a-Service subscriptions. Called GitHub Copilot for Business, the new plan, which costs $19 per user per month, comes with all the features in the single-license Copilot tier along with corporate licensing and policy controls. That includes a toggle that lets IT admins prevent suggested code that matches public code on GitHub from being shown to developers, a likely response to the intellectual property controversies brewing around Copilot.

"You can easily set policy controls to enforce user settings for public code matching on behalf of your organization," explains Shuyin Zhao, senior director of product management, in a blog post.

Microsoft is certainly fond of subscriptions payments for its enterprise products. When will they start charging for Bing ChatGPT access anyways? Disrupt Google for just $6.99 a month guys!

So why the extra $9 for corporate users? There is this feature, a public code filter, that is already available to individual users, who pay $10 per month for Copilot's AI help. But for corporate accounts, control over this filter belongs to the IT administrator. That’s a nice administrative add-on there Microsoft!

GitHub has turned into such a benevolent tool after all:

Microsoft, GitHub, and OpenAI are being sued for allegedly violating copyright law by reproducing open-source code using AI. But the suit could have a huge impact on the wider world of artificial intelligence.

No mercy no malice in the era of monopoly capitalism and btw, anything goes!

Available as a downloadable extension for development environments, including Microsoft Visual Studio, Neovim and JetBrains, Copilot is powered by an AI model called Codex, developed by OpenAI, that’s trained on billions of lines of public code to suggest additional lines of code and functions given the context of existing code. Copilot — which had over 400,000 subscribers as of August — can surface a programming approach or solution in response to a description of what a developer wants to accomplish (e.g., “Say hello world”), drawing on its knowledge base and the current context.

GitHub Copilot, introduced in 2021 as a Visual Studio Code editor extension, offers coding suggestions and functions directly from the user’s programming editor or IDE. The AI model behind Copilot is trained on open source code in public repositories.

Microsoft Alleges it’s Above the Law

GitHub claims that fair use — the doctrine in U.S. law that permits the use of copyrighted material without first having to obtain permission from the rights holder — protects it in the event that Copilot was knowingly or unknowingly developed against copyrighted code.

Microsoft, its subsidiary GitHub, and its business partner OpenAI have been targeted in a proposed class action lawsuit alleging that the companies’ creation of AI-powered coding assistant GitHub Copilot relies on “software piracy on an unprecedented scale.” Microsoft denies this to be the case.

The Free Software Foundation, a nonprofit that advocates for the free software movement, has called Copilot “unacceptable and unjust.” And Microsoft, GitHub and OpenAI are being sued in a class action lawsuit that accuses them of violating copyright law by allowing Copilot to regurgitate sections of licensed code without providing credit. But that’s not stopping Microsoft from commercializing it as soon as was possible!

Welcome to the chatGPT era of a corrupt OpenAI working in a trinity of the worship at the altar of A.I. Any questions?

An Added $9 for Privacy and Admin Control

Copilot for Business comes with a commitment that GitHub "won’t retain code snippets, store or share your code regardless if the data is from public repositories, private repositories, non-GitHub repositories, or local files."

That’s so kind of you GitHub, was that an “A.I. for Good” decision too?

TechCrunch goes on “GitHub’s attempt at rectifying this is a filter, first introduced to the Copilot platform in June, that checks code suggestions with their surrounding code of about 150 characters against public GitHub code and hides suggestions if there’s a match or “near match.” But it’s an imperfect measure.”

So in theory, business customers can rest assured that their super-secret, money-minting algorithm won't get sent to GitHub for product improvement.

Your Data is my Data

Copilot for Business, however, does transmit "engagement data," events related to editing actions (e.g. competitions accepted or dismissed), errors, and data like latency and feature use, including potentially personal data like pseudonymous identifiers.

Microsoft has found sneaky ways to protect itself legally it would seem: Whether Copilot for Business' promise to disregard the code suggestions it generates will deny data that could be used to improve future output isn't clear. Not a big surprise.

In November, lawyer and developer Matthew Butterick announced a lawsuit challenging Copilot, described as "an AI prod­uct that relies on unprece­dented open-source soft­ware piracy." It wasn’t greatly covered in the mainstream media. I noticed the story was also absent from LinkedIn, which owns Microsoft.

But GitHub, aware that its enterprise customers might be put off by uncertain legal risk, has a standing offer to defend corporate clients against infringement claims based on Copilot output in its GitHub Copilot Product Specific Terms. How very generous of them.

Suffice to say that the A.I. of code is going to get messy in the years to come.

GitHub plans to introduce additional features in 2023 aimed at helping developers make informed decisions about whether to use Copilot’s suggestions, including the ability to identify strings matching public code with a reference to those repositories.

Individual Copilot users and Copilot for Business customers not under enterprise accounts will have to face any legal action on their own – if it comes to that. So that $9 is also paying for your immunity. Peculiar and sneaky GitHub!

Microsoft loves to brag and GitHub does a great job. GitHub says Copilot has helped redefine productivity for more than a million developers, boasting that it synthesizes as much as 40% of code. For researchers, GitHub has helped users code 55% faster, GitHub said.

Read the Blog

Scraped by Microsoft - but definately A.I. for Good

Copilot, which was unveiled by Microsoft-owned GitHub in June 2021, is trained on public repositories of code scraped from the web, many of which are published with licenses that require anyone reusing the code to credit its creators.

So it turns out those $9 extra are wells spend IT managers!

• For code safety, GitHub will not retain code snippets or store or share users’ code regardless if the data is from public repositories, private repositories, or local files. This, despite the tool itself being trained on billions of lines of publicly available code.

• The ability for administrators to select which organizations, teams, and developers receive licenses.

• Policy controls that allow administrators to enforce user settings for public code matching on behalf of an organization.

Privacy, legal protection, policy controls! Microsoft knows its customers.

Thanks for reading!

Hey Guys,

I wanted to explore a bit about the rise of Remote work since the pandemic began with some data. Tl;dr Remote jobs have tripled during the pandemic and show no sign of slowing down.

Though people are returning to in-office work, the option for remote work remains high and is likely to keep growing, this is even more true for software engineers, machine learning engineers, data engineers and so forth.

Remote Work is Thriving

With the great resignation still leading to over 3.5 million quits a month in the U.S., remote work is driving a continual great reshuffle that has been the new normal even with some startup and tech job cuts.

If you have software programming, data engineer, cloud or machine learning skills - there’s never been a better time than to be a remote worker.

Follow Newsletter on Twitter

Top Remote Companies Hiring via ZipRecruiter

Here are the top 10 companies hiring for the largest share of remote-capable jobs on ZipRecruiter in 2022:

1. Anthem: 60,445 remote jobs listed this year

2. CBRE: 51,304 remote jobs listed this year

3. USAA: 42,311 remote jobs listed this year

4. Capital One: 36,336 remote jobs listed this year

5. Cerebral: 34,526 remote jobs listed this year

6. Change Healthcare: 30,602 remote jobs listed this year

7. Meta: 29,052 remote jobs listed this year

8. SAP: 282,62 remote jobs listed this year

9. Kronos: 25,965 remote jobs listed this year

10. SelectQuote: 25,799 remote jobs listed this year

How Ubiquitous is Remote Work in Technological Society?

Upwards of 60% of job seekers hope to find remote opportunities, according to ZipRecruiter data. And a similar share, 56%, of full-time U.S. workers — more than 70 million people — say their job can be done working remotely from home, according to Gallup.

That’s a lot of jobs.

Source: Source: U.S. Bureau of Labor Statistics via FRED

The Great Resignation is the New Normal in America

More than 4 million people quit their jobs in August, marking the 15th straight month where as many people left their old jobs for something new.

There are whispers that quit quitting, overemployment, mental health burnout and other factors are literally reducing productivity among workers in 2022.

So what is going on?

Remote Work and Minority Workers in Tech

Women are more likely than men to prefer remote work, and Black, Asian American and Latino workers are more likely than white peers to want the setup, per ZipRecruiter. Workplace experts have said throughout the pandemic that a greater adoption of flexible work arrangements could help boost company diversity, equity and inclusion efforts.

• Women more likely to prefer remote work

• Black, Asian Americans and Latin-X populations more likely to prefer remote work

Why that is, is not entirely clear.

Great Reshuffle among Software Engineers

A desire for remote work, and the ease of taking new work in a remote setting, are both driving some of the turnover in the labor market, furthermore it’s making it easier for some professionals like software engineers to find better opportunities and more learning on a regular basis.

Pressure to work from the office is also not the future of work. GenZ and younger Millennials are driving quits at historical levels. As of August, 2022 - a crazy high 11% of workers said they were dissatisfied with their job and looking for another where they could work from home. Among those out of work, 8% quit an in-person job to look for a work-from-home position.

Remote work preference is thus driving a new future of work that will change the very dynamics of the labor force and future of work in the 2020s.

Meanwhile more countries are creating “nomadic” remote-work Visas like Portugal recently. The 2020s may really be the decade remote work becomes more mainstream, even with weird RTO and hybrid-work compromises by some companies.

The clash of the old and the new is still causing some tension and some Great Resignation momentum. A desire for remote work, and the ease of taking new work in a remote setting, are both driving some of the turnover in the labor market.

Remote work as more Frugal with Inflation

Since the beginning of 2022, workers say Covid concerns are becoming less of a reason for wanting to work remotely, but a desire to save on commuting costs has gone up considerably. The typical job-seeker would even take a 14% pay cut in order to work remotely, with younger workers and higher earners willing to give up even more for the flexibility.

Remote Work is Booming due to Improved Well-Being

Among recent job-switchers, 14% said changing jobs allowed them to move into a role that could be done remotely, and they reported feeling more satisfied and engaged in their new position than in-office employees.

If people are happier, they will simply want to continue doing that, no matter what management or industry trends are at the time.

Employers have responded to high turnover by raising pay and offering new benefits, like remote and flexible work. But so far, it hasn’t been enough to temper disruptive turnover. In the talent competition and labor market supply demand dislocation, retaining employees is getting harder.

The Future is Remote Work

In so many ways the data indicates we won’t be returning to a 2019 type world, we are changed forever.

• 22% of jobs will be done fully remote, 3x what it was before the pandemic.

• Only 23% of jobs will be done exclusively in-person at a worksite, down from 60% in 2019.

This shift is one of the biggest trends in the future of work we’ve seen in decades and it touches nearly everyone.

Remote work applications have nearly doubled in 2022 alone. Women with young children are a great example of how that works in the real world. The Great Resignation is fueled by a Remote Work boom, that could signal the changing nature of work itself.

In July 2022, remote work listings on LinkedIn (17% of total paid job listings on the platform) drew a majority of applications (54%) and nearly half of the views (47%) compared to on-site jobs. This also complicates competition and application for jobs that aren’t tied to a zip code.

Technology Jobs have gone Full-On Remote

The industries that had the highest percentage of remote job listings on LinkedIn in July were, unsurprisingly, ones that don’t necessarily require physical interaction:

• Technology, information and media: 42%

• Professional services: 30%

• Education: 25%

• Administrative and support services: 25%

Old school managers who need to see their workers in order to trust them simply for the most part are a dying breed unless you work in Finance and banking and other sectors where it makes less sense to be remote.

Business leaders are facing “enormous pressure to cave to workers’ demands,” and in response are getting creative with other incentives that improve the work experience.

Job openings fell by 1 million in August, a sign that the labor market could be cooling as rising interest rates slow demand for goods, services and labor. Yet the unemployment rate actually went down still near historically low levels.

Remote work is here to stay.

Software Engineers are Leaders in the Remote Work Boom

According to the statistics in The State of Remote Engineering study, 86 percent of engineers work entirely in remote software engineer jobs. This is in sharp contrast to pre-pandemic figures, which indicated that 19 percent of people worked remotely 100 percent of the time.

• 86% of Software engineers now compared to 19% then work entirely remote 100% of the time.

• Software, data and machine learning engineers are thus truly pioneers of the future trend.

Teams are exploring innovative methods to sustain culture and combat burnout and give competitive compensation. While remote work has overall benefits, there are definitely mental health risks and company culture degradation issues that can sometimes lead to faster and higher rates of employee churn.

One wonders if living in another country could help combat burnout rates when signs of burnout begin to emerge.

More Americans are trying out South American countries and Mexico in recent months and years.

While job cuts have gotten a lot of headlines the reality in Tech even in the Autumn of 2022, is that jobs are in a great reshuffle where remote work is just an expected perk. The apparent willingness to leave a job suggests tech sector workers aren't too concerned by recent layoffs at US tech firms. According to Crunchbase's lay-off tracker, as of August 9, US tech firms had laid off 41,000 workers since the beginning of 2022.

You can track tech startup layoffs here.

Quit quitting and job hopping may be near an all-time high in 2022. Some 1 in 4 workers are so confident that they’d quit their job without a new one lined up, and about 1 in 3 job seekers expect wage growth to keep accelerating.

Thanks for reading!

If you enjoy my coverage and want to support the channel, for the price of a cup of coffee a month you can make a big difference.

Subscribe now

Hey Guys,

As I watch companies like Snowflake, Databricks, Datadog and others evolve, I wonder: what if they became more than just what they are today?

Snowflake Inc. is a cloud computing–based data cloud company based in Bozeman, Montana. It was founded in July 2012 and was publicly launched in October 2014 after two years in stealth mode. The firm offers a cloud-based data storage and analytics service, generally termed "data-as-a-service".

As a stock, many feel Snowflake is overvalued even for its tremendous growth. But if software is eating the world, the cloud is powering Big Data.

Snowflake’s platform supports a multi-cloud strategy, including a cross-cloud approach to mix and match clouds.

Its platform unifies data and supports a variety of workloads, including data warehousing, data lakes, data engineering, data science, data application development, and data sharing. So what if it was able to do some things better than existing service providers due to what it does well today?

The Case for Snowflake’s Media Data Cloud

What will Snowlfake, Databricks (which has yet to go publish) and others like them become?

When we think of Snowflake we don't think of its Media Data Cloud.

The Cloud is transforming data, and Snowflake will leverage this movement.

There’s been an increase in adoption of Snowflake’s tech by companies in the marketing space over the past couple of years.

Roughly a year ago, Snowflake launched its Media Data Cloud, a suite of solutions built specifically to serve businesses with media and advertising needs.

I predict this will evolve into sophisticated uses including better Marketing automation and advertising capabilities. I literally put a poll on LinkedIn about it here. The results surprised me.

How A.I. begins to impact the future of marketing is coming into focus.

Snowflake’s vendor ecosystem, which now includes more than 7,000 tech and data companies, into six main categories that marketers should be aware of: analytics, integration and modeling, identity and enrichment, activation and measurement, business intelligence and Data Science and AI.

Snowflake’s six vendor categories each have five “leaders” and vendors listed as “ones to watch,” a catch-all bucket for companies with momentum but less adoption or tech that’s designed for specific use cases.

The measurement and activation category, for instance, pegs Braze, HubSpot, The Trade Desk , VideoAmp and Simon Data as leaders. Zeta, Rivery and Hightouch are classified as ones to watch.

Companies like Snowflake, Databricks and others will have an increasingly prominent role in how the Cloud automates marketing and advertising integrating vendors and best practices seamlessly for unique business use cases.

The Cloud is also in its nascent period, and more sophisticated Data Clouds could disrupt older players. Snowflake is built on Microsoft Azure infrastructure. It provides a platform for data warehousing and a secure sharing of data. The snowflakes enable the data storage, processing, and analyzing solution in a very flexible and efficient manner. How Snowflake and Databricks evolve new business models will also change how companies do business. Cloud native solutions will evolve beyond their original solution.

The architecture of the Snowflake consist of three main key layers:

1. Database Storage

2. Query Processing

3. Cloud Services

Snowflake’s Media Data Cloud has the opportunity to allow the company to get into marketing and advertising in a way I really didn’t see coming. Snowflake recently published the first of what will become an annual report called “The Modern Marketing Data Stack,” which aggregates how Snowflake customers use cloud and marketing technology vendors.

Snowflake is releasing this inaugural report now because it’s got enough data to draw conclusions. So what if it used that data to create mar-tech and advertising solutions for some of its biggest customers?

• The report is based on an analysis of 6,000 Snowflake Media Data Cloud customers and how they use the platform.

• Snowflake also examined how 500 of its brand customers that are listed in the 2022 Forbes Global 2,000 ranking use Snowflake’s platform to work with tech partners.

• The report segments Snowflake’s vendor ecosystem, which now includes more than 7,000 tech and data companies.

I personally think we can expect the best Cloud data companies like Snowflake, Databricks, Datadog and others to evolve new solutions that directly serve their customers’ needs for example in marketing and advertising, automation and A.I.

What do you think?

Leave a comment

Read Datascience Learning Center in the Substack app

Available for iOS and Android

Get the app

Subscribe now

WED AUGUST 17TH, 2022 1:15 PM MONTREAL, CANADA

Hey Guys,

Companies still like to put their chatbots out in the public to actually talk to people, why? These sorts of experiments always end the same way.

Meta is putting its latest AI chatbot on the web for the public to talk to, and you know, what could possibly go wrong?

BlenderBot is a conversational AI prototype by Meta. Meta claims it “improves through conversation”.

If only BlenderBot3 has the charm of Meta’s Joelle Pineau, pictured above.

They released it on August 5th in the U.S. This new AI research project called BlenderBot 3, a chatbot that can search the internet to talk about nearly any topic. Users typically like to corrupt these poor devices. Microsoft’s Tay was a weird example of this a few years ago.

Meta is trying to clean up its image and supposedly being very transparent. Meta is thus sharing the BlenderBot 3 model, data and code with the scientific community to help advance conversational AI.

This topic really made me laugh out loud.

So let’s get into it:

This all amounts to the usual scam. Meta’s AI research labs have created a new state-of-the-art chatbot and are letting members of the public talk to the system in order to collect feedback on its capabilities.

If you’re worried that artificial intelligence is getting too smart, talking to Meta’s AI chatbot might make you feel better. - The Guardian

The conversational AI is designed to converse with humans about “nearly any topic.” Like Blender Bot 2, the latest upgrade also features the ability to browse the internet, is trained on the OPT-175B language model (which is 58x the size of Blender Bot 2), and has long-term memory. 

Sadly chatbots didn’t manifest much real world intelligence, inspite of the hype of nearly the past decade. I think we’ll have to wait for news from Google.

The bot is a prototype and built on Meta’s previous work with what are known as large language models or LLMS — powerful but flawed text-generation software of which OpenAI’s GPT-3 is the most widely known example.

BlenderBot 3 is designed to improve its conversational skills and safety through feedback from people who chat with it, focusing on helpful feedback while avoiding learning from unhelpful or dangerous responses. Since August 6th there’s been a flurry of reports of how its bad-mouthing Facebook and Meta’s history of exploiting its users.

I realize Mark Zuckerberg isn’t exactly very revered by the general population. More than half a decade after Microsoft's truly monumental Taye debacle (Engadget), the incident still stands as stark reminder of how quickly an AI can be corrupted after exposure to the internet's potent toxicity and a warning against building bots without sufficiently robust behavioral tether.

• Meta AI combined two recently developed machine learning techniques, SeeKeR and Director, to build conversational models that learn from interactions and feedback.

The Next Iteration of Chatbots

• long term memory

• more simulated empathy

• more human-like responses

Mark is a great example of someone with simulated empathy, a great LTM and very human-like responses.

According to Meta, the BlenderBot series has made progress in combining conversational skills — like personality, empathy and knowledge — incorporating long-term memory, and searching the internet to carry out meaningful conversations. While Amazon’s Alexa devices go mostly unused it will be interesting to see what BigTech has next for us regarding the huge future TAM of conversational A.I. and companionable chatbots?

The Verge sounds extremely optimistic when it writes: Like all LLMs, BlenderBot is initially trained on vast datasets of text, which it mines for statistical patterns in order to generate language. Such systems have proved to be extremely flexible and have been put to a range of uses, from generating code for programmers to helping authors write their next bestseller (Verge).

A range of uses yes, but mass adoption not at all.

I’m often asked if A.I. will replace human writers. I would ask Jasper.AI, if he had a voice!

The problem with VR and chatbots is we’ve been waiting for more than a decade now and they don’t seem very close to their realization of Metaverses and smart AIs everywhere, or do they?

Leave a comment

There’s now entire dedicated PR platforms just for Metaverse insider content, and it makes me wonder. I’m comfortable hyping Quantum computing because I see the technological and computing scalability in terms of a trajectory, it’s a bit harder to fathom how chatbots and the Metaverse are going to make the world a better place!

A big feature of the chatbot is that it’s capable of searching the internet in order to talk about specific topics. Like OpenAI’s DALL-E 2, this could be awfully fun to play with. We’d be beta testers doing free labor for Meta, in practice.

Since all conversational AI chatbots are known to sometimes mimic and generate unsafe, biased or offensive remarks - turning the machine against its makers is classically a fun micro-game around this theme.

Just a few days after the U.S. public launch, Meta claimed it had already collected 70K conversations from the public demo, which they will use to improve BlenderBot 3. From feedback provided by 25 percent of participants on 260K bot messages, 0.11 percent of BlenderBot’s responses were flagged as inappropriate, 1.36 percent as nonsensical, and 1 percent as off-topic. Meta says this is huge progress, thanks for your free labor guys!

The Guardian snarkly remarked: Asked about Mark Zuckerberg, the bot told BuzzFeed’s Max Woolf that “he is a good businessman, but his business practices are not always ethical. It is funny that he has all this money and still wears the same clothes!” That sounds about right! Clearly now I can trust chatbots.

I might even mistaken them for being sentient! Especially if I’m a priest at the altar of A.I. Honestly I think we can laugh at Meta’s expense, it’s not a big deal for a BigTech company on the decline.

Meta's stock is down 48% so far this year, YTD, even after a huge July and August rally in the NASDAQ.

The bot has also made clear that it’s not a Facebook user, telling Vice’s Janus Rose that it had deleted its account after learning about the company’s privacy scandals. “Since deleting Facebook my life has been much better,” it said.

Now I just have to figure out how I’m going to delete the Metaverse?

Thanks for reading!

How do you see Chatbots and the Metaverse adoption at scale, if it’s even possible in the decade ahead?

Leave a comment

Thanks for reading! If you want to support the channel and allow me to continue to write Newsletters feel free to get access to more content. Basically the more community supporters I get the longer it’s viable for me to write.

Subscribe now

AUGUST 6TH, 2022 1:30 PM MONTREAL, CANADA

I have a summer discount going to reward my early backers and most passionate readers. I’m really grateful for the support.

Get 22% off for 1 year

Hey Guys,

I saw this outlandish story that I found a bit amusing. Google’s PR claim of “Quantum Supremacy” always seemed a bit bogus. Now, a team of scientists in China claim to have replicated the performance of Google’s Sycamore quantum computer using traditional hardware, thereby undermining the suggestion the company has achieved quantum supremacy. 

It wasn’t so long ago, rewind to the year 2019. Back in 2019, Google proudly announced they had achieved what quantum computing researchers had sought for years: proof that the esoteric technique could outperform traditional ones. But this demonstration of “quantum supremacy” is being challenged by researchers claiming to have pulled ahead of Google on a relatively normal supercomputer.

China’s efforts in R&D, papers in AQ and A.I. seem to be increasing each year. As reported by Science magazine, the scientists used a system comprised of 512 GPUs to complete the same calculation developed by Google to demonstrate it had passed the quantum supremacy milestone back in 2019. How different even three years feels!

The endeavor was led by statistical physicist Pan Zhang, who said his team’s supercomputer performed the calculation 10 billion times faster than Google had thought possible.

Read the paper here:

Read the Paper

Some of Google’s PR seriously backfires and seems out of touch with other research trends. Quantum computers like Sycamore are not better than classical computers at anything yet, it turns out according to the Chinese paper, with the possible exception of one task: simulating a quantum computer.

There is a lot of hype out there with companies, startups and even Governments making outlandish claims. Quantum advantage or Quantum supremacy are now terms used in all sorts of contexts. I like the term ‘AQ’ that implies some kind of convergence between quantum computing and artificial intelligence.

The Quest for quantum supremacy

Quantum supremacy (or quantum advantage) can be defined as the point at which quantum computers can outstrip the maximum potential performance of classical supercomputers in a particular discipline.

So why does the recent paper matter? The team at the Chinese Academy of Sciences led by Pan Zhang has published a paper describing a new technique for simulating a quantum computer (specifically, certain noise patterns it puts out) that appears to take a tiny fraction of the time estimated for classical computation to do so in 2019.

So much this is about powerful simulation and the ability to predict business outcomes or processes in science.

We Need More Research

Three years ago, Google announced it had achieved this feat with Sycamore, which it said took just 200 seconds to complete a statistical mathematics problem that would take a supercomputer 10,000 years to solve.

The problem was architected in such a way as to accentuate both the attributes of quantum computers, which exploit a phenomenon known as superposition to speed up calculations, and the limitations of traditional systems.

Historical Context

However, the claim was treated with suspicion by industry players from the beginning. The loudest critic was IBM, which claimed its Summit supercomputer would be capable of solving the same puzzle in two and a half days, given enough time to perfect the implementation. And now, a paper published by Zhang and his colleagues has cast further doubt over Google’s claims.

IBM, Microsoft, Amazon and others are also beginning to invest a lot in the future of quantum computing, as well as more Quantum computing startups which I cover on my Newsletter Quantum Foundry:

Quantum Foundry

Quantum computing, IPOs, startups, future companies, business models, venture capital deals, research & papers, global news coverage, etc...

By Michael Spencer

Zhang et al. cast the problem as a large 3D network of tensors, with the 53 qubits in Sycamore represented by a grid of nodes, extruded out 20 times to represented the 20 cycles the Sycamore gates went through in the simulated process. The mathematical relationships between these tensors (each its own set of interrelated vectors) was then calculated using a cluster of 512 GPUs.

TechRadar explained this really well:

• As explained by Science, to replicate the performance of Sycamore, the scientists first had to reconfigure the problem as a 3D mathematical array called a tensor network. This meant they could use the system’s GPUs to perform calculations in parallel.

• The low level of accuracy of Sycamore’s computation gave the Chinese team additional room for maneuver. By minimizing the accuracy target, they were able to accelerate the calculation by hundreds of times, replicating the same level of performance the quantum machine had achieved.

The Problem with Hype PR

The problem with a lot of the PR firms like Google and OpenAI are making is it’s somewhat deceptive. The Ad-based internet has normalized pointless boasting that actually leads to less trust in BigTech as a whole.

In Google’s original paper, it was estimated that performing this scale of simulation on the most powerful supercomputer available at the time (Summit at Oak Ridge National Laboratory) would take about 10,000 years — though to be clear, that was their estimate for 54 qubits doing 25 cycles; 53 qubits doing 20 is considerably less complex but would still take on the order of a few years by their estimate.

So a lot of those PR statements are rigged in their own favor.

As the Chinese here have used advanced algorithms and 512 GPUs to achieve quantum-level results, it begs a lot of questions!

It makes you wonder about how BigTech is now a PR factory where Microsoft is using LinkedIn to benefit its own image and Google is doing the same with search and its own powerful influence on discovery, e.g. suggested articles on Google search, Google News, etc…

I hope Quantum computing hype doesn’t fall prey to the same somewhat misleading tactics. The integrity of journalism has been alterted permanently by the Ad-based internet, and people are beginning to trust the Tech news and crypto news media industry as whole, leading more of us to get our news from social outlets like TikTok or via mobile search.

So instead of the 10,000 years Google claimed the task would take, Zhang’s group claims to have done it in 15 hours. And if they had access to a proper supercomputer like Summit, it might be accomplished in a handful of seconds — faster than Sycamore. Their paper will be published in the journal Physical Review Letters; you can read it here (PDF).

According to TechCrunch, these results have yet to be fully vetted and replicated by those knowledgeable about such things, but there’s no reason to think it’s some kind of error or hoax.

Google even admitted that the baton may be passed back and forth a few times before supremacy is firmly established, as it’s incredibly difficult to build and program quantum computers while classical ones and their software are being improved constantly. So the question remains how much can we trust PR statements they make on a field they are not a leader in? BigTech already lobby but how they impact misinformation online is also somewhat dubious when they are themselves like media companies.

Chinese scientists successfully proved IBM's point by attacking the original problem using advanced algorithms and compute power from today's GPUs to complete the calculation. That this Chinese team were able to do this so easily, really is quite shocking for the reputation of Google.

Some people in the Quantum computing field find the hype misrepresents their work. So why would a company like Google make claims that ultimately aren’t true? If I was bullish on quantum computing scaling soon, I’d just assume Google may have been hasty in its celebrations, as it is expected that the quantum supremacy milestone will fall sooner rather than later by virtue of the level of investment going into R&D activity in the space. But scalability issues like this are very hard to solve and usually take far more time in years and decades, than originally hyped to take.

The truth is we don’t actually know the impact of more scalable A.I., quantum machines and how convergence of all of these trends like ‘AQ’ will take shape. Not even Google is good at predicting it, even as they lead with DeepMind and Google AI in some areas. All claims of hype have to be taken in context.

Google’s damage control to real studies countering its PR claims remains to be seen. Google Quantum AI's principal scientist, Sergio Boixo, said the original 2019 paper acknowledged the likely future improvement to classical algorithms but doesn't believe today's classical computing approach can keep pace with quantum technologies. So there’s always an escape hatch for academics, PR and the games BigTech play on our ability to envision the future.

Academics are then paid to say whatever makes Google look good. As University of Maryland quantum scientist Dominik Hangleiter told Science, this isn’t a black eye for Google or a knockout punch for quantum in general by any means: “The Google experiment did what it was meant to do, start this race.” But it does validate much of the skepticism those in the quantum space felt back in 2019 about Google’s outlandish claims.

“I think they’re right that if they had access to a big enough supercomputer, they could have simulated the … task in a matter of seconds,” says Scott Aaronson, a computer scientist at the University of Texas, Austin.

How will the future of AI interact?

• Hardware: More powerful supercomputers

• Models: More scalable efficient language models

• Quantum: Quantum machines with more qubits

• Software: Evolution of Quantum Machine Learning

• Capitalism: The continued evolution of Neo Surveillance Capitalism

• Government: The investments in National Defense and the CCP that impacts this over the scope of many decades.

The truth is we don’t really know. In fact China may have a better idea about the long-term implications of A.I. at scale than Google or Microsoft does in isolation. This is because China makes strategic long-term bets on economic dominance of smart technologies and A.I. in long-term plans.

The Quantum aspect of this alone is going to transform our view of the future of artificial intelligence. But scalability is not near. A lot of work is going into improving performance by increasing the number of qubits on a quantum processor, but researchers are also investigating opportunities related to qubit design, the pairing of quantum and classical computers, new refrigeration techniques and more.

Claims of Quantum supremacy or Quantum advantage don’t necessarily make the world a better place, they just stoke an AI arms race of AQ now between the U.S. and China and their more powerful institutions, companies and best researchers. That’s an outcome that could actually be potentially dangerous.

What do you think?

Leave a comment

If you think my work matters and my accountability journalism in the space is valuable, please consider a small donation to support me.

Subscribe now

I have a pretty awesome Summer Discount going in case you want to support my work here. It’s pretty substantial considering how cheap the monthly cost to begin with is.

Get 22% off for 1 year

Hey Guys,

Question of the Day

I recently have started my Sunday (8:00 AM EST) Discussion thread going on AI Supremacy about if DALL-E 2 will disrupt designers, illustrators, freelance brand experts and artists?

You can participate on the thread here:

https://aisupremacy.substack.com/p/discussion-thread-will-tools-like/comments

But I wanted to cover something pretty amusing about how NLP tools will have dozens of startups around them.

Who needs to write real articles when you can be selling your AI-tested prompts.

Summon your inner writer today that is AI-friendly! Do you like to tinker with NLP tools with your awesome ability to imagine new possibilities?

PromptBase might be right for you. See their website here.

What is PromptBase?

• Prompts are becoming a powerful new way of programing AI models like DALL·E & GPT-3.

• However, it's hard to find good quality prompts online.

If you're good at prompt engineering, there's also no clear way to make money from your skills.

PromptBase is a marketplace for buying and selling quality prompts that produce the best results, and save you money on API costs. Save credits, crowdsource the PromptBase marketplace.

There are so many juicy ethical implications of this, I cannot even wrap my head around all the legal aspects of this.

So let’s get into it:

An Engineering Degree Not Required!

Now a startup is looking to let “prompt engineers” cash in with an online marketplace that sells these finely tuned phrases.

PromptBase, launched in June, allows users to sell strings of words that net predictable results with particular systems. Priced at $1.99 — PromptBase takes a 20% cut — the content that the prompts generate range from “viral” headlines to pictures of sports team logos, knitted dolls and animals wearing suits.

PromptBase hosts only prompts tested on DALL-E 2 and GPT-3. But according to its founder, Ben Stokes, the plan is to expand the platform to additional systems in the future.

Are you on good terms with NLP based AIs? Do they like you? Are you good at controlling their unlimited power? Directing their machine learning synergies? Join PromptBase as a hybrid wizard.

Maybe it’s your time, to tinker with the future on the Substack for DALL-E 2 Prompts!

Ben Stokes was only to happy to share his vision for the platform with TechCrunch:

“Our ultimate aim is to build tools in order to help support prompt engineers. It’s early days, so we’re currently just trying to spread the word and find prompt engineers to sign up and start listing their prompts for sale on our marketplace.”

“We’re already seeing big tech companies build their own systems similar to GPT-3 and DALL-E, and I predict many more to come. Different systems will likely be utilized like tools in a toolbelt, similar to how different programming languages are used today, and we plan to accommodate all of them as they gain popularity.”

Dozens of startups are quickly positioning themselves around NLP based tools, chatbots and writing and marketing automation related activities.

Who needs to waste DALL-E 2 credits when you can get the best prompt for your desired NLP-command prompt tailored to your specific need?

• Sports Team Lobos

• 1-Bit Pixel Art

• Corporate Style Art

• Anime/Magma

• Custom made Emojis (Clay)

• Logos both cute and retro

• Video game art

• Cyberpunk Misc.

Or that Knitted Doll image you never knew you needed!

Prompt engineering

Since DALL-E 2 is so customizable based on prompts, it’s sometimes easier just to buy a prompt that works the best for what you need: You can buy it before you coulds say:

“Film still of a woman drinking coffee, walking to work, telephoto”

Filtering NLP Creative Generated Images

I think we have to think of prompt engineering in this context like a search and filter. They can also act as “filters” of sorts, creating images with the characteristics of a sketch, painting, texture, animation or even a particular illustrator (e.g., Maurice Sendak). And prompts can portray the same subject in different styles, like “a child’s drawing of a koala riding a bike” versus “an old photograph of a koala riding a bike.”

You will remember how Karen created her Cosmo Magazine cover, apparently it took her thousands of iterations and hours and hours of prompt engineering. (TikTok link)

@karenxchengCreating the first ever artificially intelligent magazine cover for @Cosmopolitan ! Using #dalle2 to generate the art

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Where do we go from here? Freelancers who give us the write phrasing to interact with NLP tools? I didn’t really see that coming.

You might be thinking but wait, I can do that!

The prompt of the words, their order and the nuanced sentence really matters.

There’s a weird wide range of DALL-E 2 that is actually pretty impressive. TechCrunch shares another study that caught my attention.

In a fascinating study out of the University of Texas at Austin, researchers documented an extensive vocabulary of bizarre prompts that can be used to generate images with DALL-E 2. They discovered that the system understands “Apoploe vesrreaitais” — a gibberish phrase — to mean “birds” and “Contarra ccetnxniams luryca tanniounons” to mean “bugs” or “pests” (sometimes). Giving DALL-E 2 the prompt “Apoploe vesrreaitais eating Contarra ccetnxniams luryca tanniounons” yielded pictures of birds eating bugs.

According to its founder, Stokes argues that many of the prompts on PromptBase represent hours of genuine work and insight by engineers. There could be some legal grey area in this guys. Microsoft wants you to waste credits, not buy prompts! OpenAI was once a non-profit, which seems like forever ago.

But it’s Sam Altman we are talking about.

A number of researchers and enthusiasts have released free resources containing prompts for popular AI systems, mostly DALL-E 2. PromptBase is one of the first to monetize the exchange. But DALL-E 2 is not free forever, people will buy credits. They will cut out freelancers. This is the free-market we are talking about.

What happens on Reddit does not always stay on Reddit, especially if it’s the Subreddit DALL-E / https://www.reddit.com/r/dalle2/comments/w94n5w/selling_prompts_an_analysis/

Scandalous, people actually want to make money on the creativity of others! Can you imagine?

There’s a long-running debate within the AI community over which research, if any at all, should or can be commercialized; one Reddit user argues that PromptBase is “starting a trend that threatens the openness and accessibility of AI in general.” I like that talk, but I’m more worried about OpenAI and Microsoft’s behavior around A.I. as a whole. If this is how the AI community feels, they have to urge for corporate accountability. That’s why talent goes to places like HuggingFace.

It’s the future and NLP and chatbots will get smarter, and people will pay for their use (at least until it becomes a ubiquitous utility on the web). If it was a subscription, it would make more sense.

Stokes seems stoked about his marketplace for tiny prompts.

“Today we have prompts to generate basic text and images, but it’s not too hard to extrapolate years into the future where we’ll have prompts for generating videos, and maybe one day even feature-length films complete with orchestral scores,” Stokes added. “Those people who can craft the quality prompts required guide the AI to do these things will be extremely valuable. It’s unknown how big the market will be, but I can see it being a key tech skill, if not the future of programming.”

My Reddit alias is Mediocre_Credit so don’t look at me. For the record, Stokes is on Twitter here and on Reddit here.

I think OpenAI may have opened a can of worms in terms of data privacy. I didn’t know this but apparently studies show that language systems trained on vast swaths of public data, like GPT-3, can “leak” personal information, including names and addresses, when fed certain prompts.

Like trying to answer Questions on Quora for coffee money, I’m the weird sort of person who would enjoy being a Prompt creators.

The most hilarious thing I read about this story is perhaps the following: Stokes said that PromptBase reviews every listing in the marketplace to ensure they don’t violate any “AI generation rules.”

Pray tell overlords of the OpenAI and PromptBase universe, what might those be?

Would like to hear your ideas on this too.

Leave a comment

Thanks for reading!

Subscribe now