Meta is Surprisingly Relevant in Generative AI

🦙 Llama-3 is going to push open-weight LLMs into a new world. Enterprise AI and Agentic AI apps.

Image Credit: Ivy Liu

Hello Everyone,

I’m not a big fan of Facebook historically, and I had noticed Meta was a huge buyer of Nvidia H100 AI chips. However with Llama-2 and in particular Llama-3 will do to the open source (open-weight) community has been and will be tremendous.

Help support my work for as little as $2 a week. It makes a big difference for my coverage of emerging tech.

I stand by the “Metaverse” as a nefarious potential ecosystem, however Meta’s investment in Generative AI is clearly going to lead to some good in the world. It’s actually getting impossible to ignore Facebook’s investment in AI and Meta’s contribution to the open-weight space.

Nvidia robotics mentioned Meta Llama 3, the latest open-weight LLM is also optimized to run on our NVIDIA GPUs across cloud, data center, edge, and desktop. For instance you could run it on Jetson Orin following the guides on Jetson AI Lab. I’ve been covering a thing or two about robotics lately.

Analyzing and Curating News 🔔

So why am I talking about this? I’m a news fanatic always analyzing the latest in exponential and emerging tech like robotics, semiconductors & AI chips, quantum, synthetic biology and other fields - as well as BigTech in AI. Never would I have guessed the impact Llama-2 would have on the ecosystem where now dozens of decent open-weight models are released each week on HuggingFace who have over 1 million models.

Llama-3 Signals the Open-Weight Era of Generative AI

I’m calling it, Llama-3 signals a milestone of sorts; April 18th, 2024 marks the start of the open-weight era of Generative AI.

---

Details in the full announcement ➡️ https://ai.meta.com/blog/meta-llama-3/

Download the models ➡️ https://llama.meta.com/llama3/

Experience Llama 3 with Meta AI ➡️ https://www.meta.ai/

Models on Hugging Face ~ https://huggingface.co/models?other=llama-2

---

So amid these ecosystems like HuggingFace, GitHub, now Mistral, even Meta has a special contribution to make with its announcement of Llama-3 yesterday.

It makes me actually stop to listen to what Mark Zuckerberg’s says he’s doing with Generative AI. Meta is adding real-time AI image generation to WhatsApp and an AI Assistant across its family of apps.

Mark Zuckerberg on Meta’s new AI Scene

A new version of Meta AI goes live.

• Meta’s free artificial intelligence assistant, Meta AI, is rolling out across its social media platforms WhatsApp, Instagram, Facebook and Messenger.

• Meta is rolling out real-time AI image generation in beta for WhatsApp users in the US.

• Check out

  Dwarkesh Patel
  podcast with Mark recently here if interested.

  Mark Zuckerberg - Llama 3, Open Sourcing $10b Models, & Caesar Augustus

According to Dwarkesh the podcast covers Mark Zuckerberg on:

• Llama 3

• open sourcing towards AGI

• custom silicon, synthetic data, & energy constraints on scaling

• Caesar Augustus, intelligence explosion, bioweapons, $10b models, & much more

So jump to the section that interest you the most! That time Mark sounded more like Sam Altman? Is this even the real Zuck? Ha, nevermind.

Timestamps

(00:00:00) - Llama 3

(00:08:32) - Coding on path to AGI

(00:25:24) - Energy bottlenecks

(00:33:20) - Is AI the most important technology ever?

(00:37:21) - Dangers of open source

(00:53:57) - Caesar Augustus and metaverse

(01:04:53) - Open sourcing the $10b model & custom silicon

(01:15:19) - Zuck as CEO of Google+

Why is Llama-3 a Big Deal?

According to Jim Fan of Nvidia, “the upcoming 400B+ tier will mark the watershed moment that the community gains open-weight access to a GPT-4-class model. It will change the calculus for many research efforts and grassroot startups. I pulled the numbers on Claude 3 Opus, GPT-4-2024-04-09, and Gemini.

Llama-3-400B is still training and will hopefully get even better in the next few months. There is so much research potential that can be unlocked with such a powerful backbone. Expecting a surge in builder energy across the ecosystem!”

I think in many ways Llama-3 could be more significant than GPT-5 for 2024. This is because it actually helps the world build better LLM and SLM models and potentially better Agentic AI, rather than just making OpenAI and Microsoft more profitable.

Consider supporting what I do for less than $2 a week. More insights on Llama-3 and Meta’s conception of AGI below.