Llama-v2 is open source, with a license that authorizes commercial use!
Big Moment for Open-Source AI arrives.
Hey Everyone,
So today July 18th, 2023 something impactful was announced.
Yann LeCun on LinkedIn posted:
This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers
Pretrained and fine-tuned models are available with 7B, 13B and 70B parameters.
Llama-2 website: https://ai.meta.com/llama/
Llama-2 paper: https://lnkd.in/djeTCvfn
A number of personalities from industry and academia have endorsed our open source approach: https://lnkd.in/dSge3PGV
Hugging Face aficionados reported on it:
What does it all mean?
Llama 2 is free for research and commercial use.
Microsoft and Meta are expanding their long-standing partnership, with Microsoft as the preferred partner for Llama 2.
They’re opening access to Llama 2 with the support of a broad set of companies and people across tech, academia, and policy who also believe in an open innovation approach to today’s AI technologies.
Meta AI is also committed to building responsibly and are providing resources to help those who use Llama 2 do so too.
Recently the AI arm of Mark Zuckerberg owned Meta, has introduced a multimodal model “CM3leon” (pronounced like chameleon), that does both text-to-image and image-to-text generation.
Meta AI has been more active in the A.I. space in recent weeks than many presumed they would be.
Microsoft was boasting with PR about the partnership.
Today, at Microsoft Inspire, Meta and Microsoft announced support for the Llama 2 family of large language models (LLMs) on Azure and Windows. Llama 2 is designed to enable developers and organizations to build generative AI-powered tools and experiences. Meta and Microsoft share a commitment to democratizing AI and its benefits and we are excited that Meta is taking an open approach with Llama 2. We offer developers choice in the types of models they build on, supporting open and frontier models and are thrilled to be Meta’s preferred partner as they release their new version of Llama 2 to commercial customers for the first time.
Microsoft’s $13 Billion in OpenAI, did anything but help democratize A.I. to be honest. That BigTech is pretending they are behind open-source innovation is getting a bit weird.
It’s as if Azure is trying to take credit? Very strange. The PR adds: “Expanding Azure AI model catalog and Windows availability. Llama 2 is the latest addition to our growing Azure AI model catalog. The model catalog, currently in public preview, serves as a hub of foundation models and empowers developers and machine learning (ML) professionals to easily discover, evaluate, customize and deploy pre-built large AI models at scale.”
Meta is boasting about CM3leon and now this. BigTech a tyranny of PR, now somehow admitting open-source is important.
Listen that the announcement, is coupled with details of Microsoft’s pricing plans for artificial intelligence in some software, underscores rapid AI development at big tech companies and the sad monopolization of the state of Generative A.I.
Meta AI the leader in open-source LLMs? Microsoft being agonistic? Hardly.
Llama 2 is available for free for research and commercial use.
Over 100,000 requested access to Llama 1, Meta Says.
The paper is 76 slides long.
Llama 2 was trained on 40% more data.
The benefits of open-source have BigTech worried, so they are pretending they are a big part of the wave.
Open-source models certainly have a mix of benefits and negatives. Providing a wide group of people with an AI's differing capabilities can help the model learn faster, given the increased data it's receiving. Additional eyes can also spot bugs and security issues, giving developers a chance to fix these problems sooner.
Each model trying to out-do the others. So this is what it comes to.
According to Meta, LLaMa 2 was trained on 40 percent more data when compared to LLaMa 1, which includes information from “publicly available online data sources.” It also says it “outperforms” other LLMs like Falcon and MPT when it comes to reasoning, coding, proficiency, and knowledge tests.
Benchmarks
Llama-2 looks good on paper.
BigTech keeps trying to hype up what LLMs can do. Large Language Models (LLMs) have shown great promise as highly capable AI assistants that excel in complex reasoning tasks requiring expert knowledge across a wide range of fields, including in specialized domains such as programming and creative writing.
GitHub
https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md
These guys seem to take A.I. models very seriously.
In opening up LLaMA, Meta said it wants to improve safety and transparency. Meta AI ain’t no Anthropic, I can tell you that. Google has a 10% stake in Anthropic A.I.
The company said the LLaMA 2 model has been “red-teamed,” or tested for safety by “generating adversarial prompts to facilitate model fine-tuning,” both internally and externally.
Safter is Better
Meta AI branding themselves as the good guys in A.I. innovation.