Microsoft is Trying to Dig into Hugging Face's Newfound Popularity

The GitHub of AI is Trending.

Hey Guys,

This is Datascience Learning Center Premium,

I want to thank you for supporting the channel. The money goes directly to my family’s survival, food and rent.

Personal Note

On a personal note my family battled my Mom’s terminal cancer and lost on May 25th, 2022, so if I haven’t been writing much on this Newsletter, that’s the reason. However, I plan to make up for it with more posts in the near future.

So let’s get into it:

Hosting over 40,000 models and serving over 1 million downloads a day, Hugging Face is the go-to destination for all things Transformers.

It’s a bit alarming to see Microsoft reel in Hugging Face into their ecosystem. You can see Transformer on GitHub here: https://github.com/microsoft/huggingface-transformers

In late May the partnerships was clear. Fresh off a $100 million funding round (TechCrunch), Hugging Face, which provides hosted AI services and a community-driven portal for AI tools and data sets, announced on May 24th, a new product in collaboration with Microsoft.

True to its mission to democratize good ML, Hugging Face is always looking for innovative ways to help everyone solve complex challenges with Transformers quickly, easily, and regardless of ML expertise.

Will Microsoft Aim to Acquire Hugging Face?

Microsoft already acquired GitHub, and now appears like it may be interested in acquiring the “Github of AI”, Hugging Face.

Microsoft has also warped the integrity of OpenAI in recent years. OpenAI granted Microsoft, with which it has a commercial relationship, the exclusive licensing rights to its powerful GPT-3 language model.

According to TechCrunch, other organizations say that the code they use to develop systems is dependent on impossible-to-release internal tooling and infrastructure or uses copyrighted datasets.

While motivations can be ethical in nature (somewhat doubtful now in this case) — OpenAI initially declined to release GPT-2, GPT-3’s predecessor, out of concerns that it might be misused — the effect is the same. Without the necessary code, it’s far harder for third-party researchers to verify an organization’s claims.

Hugging Face in their blog states: One of the main problems that developers and organizations face is how difficult it is to deploy and scale production-grade inference APIs. Of course, an easy option is to rely on cloud-based AI services. Although they’re extremely simple to use, these services are usually powered by a limited set of models that may not support the task type you need, and that cannot be deeply customized, if at all. Alternatively, cloud-based ML services or in-house platforms give you full control, but at the expense of more time, complexity and cost.

Hugging Face Endpoints on Azure

Hugging Face announced a new product in collaboration with Microsoft called Hugging Face Endpoints on Azure, which allows users to set up and run thousands of machine learning models on Microsoft’s cloud platform. Microsoft wants to own many of the building blocks of A.I. It’s a bit of a scary time if they manage to acquire Hugging Face.

Hugging Face Endpoints are available in public beta in all Azure Regions. Some researchers go so far as to say that withholding a system’s code “undermines its scientific value.” However Microsoft wants to commercialize the likes of OpenAI and GitHub for profit, not necessarily for the open-source community for the benefit of all.

Hugging Face Endpoints on Azure does not seem especially popular that far with just one 5-star rating to date. Hugging face has its own momentum. Having started as a chatbot application, Hugging Face made its fame as a hub for transformer models, a type of deep learning architecture that has been behind many recent advances in artificial intelligence, including large language models like OpenAI GPT-3 and DeepMind’s protein-folding model AlphaFold.

The Next Layer of the Software Cloud

In an era of Snowflake and Databricks, how datascience is intersecting the Cloud and A.I. is getting more interesting. In recent years, Transformer models have proven to be exceptionally efficient over a wide range of ML tasks, including Natural Language Processing (NLP), Computer Vision, and Speech. Hugging Face is becoming a more important hub of late. While many use GitLab how they once used GitHub.

Hugging Face has raised $161 million to date. GitLab have raised more than 3x that. The demand for AI remains high. According to a recent McKinsey survey, nearly two-thirds of companies plan to increase their investments in AI over the next two years. But implementing AI from scratch can be challenging.

As interesting as GitHub Co-pilot and OpenAI are, at the rate of change in A.I. labs and language models, they might not be very relevant in a few years time. GPT-4 could come out at anytime in the next six months, but DeepMind among others have replicated and improved upon what OpenAI had achieved in a very short time period.

That Hugging Face announced it has a new partnership with Microsoft to “democratize machine learning” through open source collaboration and make the Hugging Face machine learning platform accessible to Microsoft Azure customers, something didn’t sit right. Microsoft has a way of bending startups to its cause with deep pockets. Yet few companies can realistically compete with Microsoft dominance in software, as its Teams product might not just disrupt Slack but Zoom as well.

Microsoft has allowed the pandemic to be a time of digital transformation where they acquired even more gaming, security and startups at a faster pace than usual. It’s not as if software developers aren’t already using software languages backed and sponsored by BigTech. I personally would be sad to see Hugging Face get acquired, but I admit it’s a distinct possibility considering its rising utility in A.I.

What do you think?

Thanks for reading!