Meta Releases First Open Source Model That Rivals ChatGPT

Hi everyone

This is Lore, the techno-optimist newsletter. 

Every week, I share 5 interesting things to help you thrive in the age of AI.

Thank you for being a part of 20k+ techno-optimists and growing. Feel free to forward this to your friends if you feel the urge. 🙏

P.S. Watch the latest episode of The Next Wave with Greg Isenberg for his SEO 2.0 Playbook!

NEW: Lore Newsletter App

We’re excited to announce that we now have a Lore Newsletter App! You can download the Lore newsletter app and be notified when new issues come out!

Just visit our newsletter on your mobile phone, and make sure you’re logged in and follow the instructions.

1) Things Are Getting Better

Elon Musk estimates that Humanoid Robots will be available for internal use next year and widely released by 2026. The world is about to get way cooler.

We will have a sci-fi world where robots are walking around, and combined with AI, we can build entirely new things in the physical world that have never been possible before.

Keep accelerating.

2) Meta Releases First Open Source Model That Rivals ChatGPT

Today, Meta released Llama 3.1, the best open-source AI model ever.

Key Features of Llama 3.1:

• Improved performance across various benchmarks, rivaling closed-source models like GPT-4 in some areas.

• It comes in 8B, 70B, and 405B parameter versions.

• Enhanced multilingual capabilities, now supporting over 100 languages

• Reduced hallucinations and improved factual accuracy

• Fine-tuned versions for specific tasks such as coding and mathematical reasoning

Try out the 8B and 70B version here with Groq. It’s nuts how fast it is.

Meta's commitment to open-sourcing Llama 3.1 is inspiring. I strongly believe that we can’t have one AI model to rule them all. Otherwise, we’re going to end up living in 1984.

People have already found a way to run Llama at home, using 2 Macbook Pros.

Imagine what it means now that open-source models as powerful as GPT-4o can be spun up by any company or government in the world. Things are about to get wild.

My friend Rowan Cheung conducted an exclusive interview with Mark Zuckerberg about this release. Watch the interview below for more insights directly from Meta's CEO.

This release represents a significant step forward for open-source AI.

While I believe we'll soon see even more advanced models from companies like OpenAI, Llama 3.1's performance demonstrates that open-source alternatives can keep pace with proprietary models. This competition is crucial for preventing a monopoly in AI technology.

We can’t allow one AI company to dictate the future of intelligence and truth.

3) SEO 2.0 Playbook with Greg Isenberg

In this week’s episode of The Next Wave, Matt Wolfe and I spoke with Greg Isenberg, our first repeat guest. This time, we talked about SEO 2.0 and how you should be thinking about SEO in the Age of AI. Check it out!

It would also help if you subscribe to The Next Wave on Apple, Spotify, and YouTube. That’d make HubSpot and me very happy.

4) Elon Announces World’s Most Powerful AI Training Cluster

This week, Elon Musk announced on X the launch of what he claims is the world's most powerful AI training cluster in Memphis, Tennessee. The Supercluster, a collaboration between xAI, X, and Nvidia began training at 4:20 am local time on July 22nd.

Key points:

• The cluster boasts 100,000 liquid-cooled H100s (NVIDIA's most advanced AI chip).

• In a recent interview with Jordan Peterson, Elon stated he expects Grok (xAI's language model) to be the most powerful LLM globally by December.

By the time they release the new LLM in December, GPT-5 will probably be out already, and OpenAI will still be ahead. Still, this shows that Elon is taking catching up in AI seriously.

Combining Elon’s speed of execution with their real-world data from cars(and soon robots) and real-time data with X, and Elon’s AI could end up being #1.

5) Things I’m Learning From

• ARTICLE: Confronting Impossible Futures by Ethan Mollick —This is a great article from Professor Mollick about how, despite there being so much hype about AI and many tech leaders involved in AI publicly stating that AGI is coming in 2-5 years, companies still aren’t changing their strategies enough to account for the changes that are coming.

• VIDEO: AI News: We're One Step Closer To AGI This Week!— Video from Matt talking about the updates from OpenAI this week.

• PAPER: THE EMPLOYMENT EFFECTS OF A GUARANTEED INCOME: EXPERIMENTAL EVIDENCE FROM TWO U.S. STATES - Paper showing the results of a UBI experiment supported by OpenAI. Interestingly, some people read this as a very positive outcome, and others see it as a negative outcome. The TLDR is it seems people work less if you give them money. But maybe that’s not always a bad thing.

That’s all for today. Please consider sharing the newsletter with your friends if you think they’d enjoy it. 🙏

Please give me feedback on Twitter. I am still figuring out the best format for this newsletter. Which part did you like? Dislike? Do you have other suggestions? Please let me know—just tweet @NathanLands.

-Nathan