- What's brewing in AI
- Posts
- 🧙🏼♂️ Llama 3.1 is awesome
🧙🏼♂️ Llama 3.1 is awesome
Also: OpenAI in talks to make their own chip
in partnership with
Howdy, wizards.
Today it’s all about the new Llama in town. If GPT-4o is Microsoft Word, then Llama 3.1 405B is Google Docs – an open version that’s equally good.
Let’s dive in!
Dario’s Picks
The most important news stories in AI this week
Meta releases the first open source frontier-level AI model. Comes in 8B, 70B and 405B sizes – the latter is on par with the best closed models to date, including the top models from OpenAI, Google and Anthropic. Zuckerberg also wrote a letter detailing his thoughts on why open source AI is good. Try the new model here. Full details below.
Why it matters Meta is making a big play here not simply on an open source model, but rather a full scale open ecosystem which could become the industry standard where most AI development happens. Realising that vision would definitely bring influence and monetization opportunities to Meta, even without users paying for access to its models.
From our partners
Billionaires wanted it, but 54,578 everyday investors got it first… and profited
When incredibly rare and valuable assets come up for sale, it's typically the wealthiest people that end up taking home an amazing investment. But not always…
Masterworks is taking on the billionaires at their own game, buying up and fractionalizing some of history’s most prized blue-chip artworks for its investors. In just the last few years, its investors have realized annualized net returns of 17.8%, 21.5%, 35% and more from these opportunities.
Their nearly $1 billion collection includes works by greats like Banksy, Picasso, and Basquiat, all of which are collectively owned by everyday investors. When Masterworks sells a painting – like the 16 it's already sold – investors reap their portion of the profits.
It's easy to get started but offerings can sell out in minutes. However, as a trusted partner, What’s brewing in AI readers can skip the waitlist to join with this exclusive link.
Behind the news
Meta's Llama 3.1 is here
Source: DALL-E / What's brewing in AI
Meta has released its latest series of AI models: Llama 3.1. They come in 8B, 7B and 405B sizes. Notably, the 405B model is the first open-source model to be approximately on the level of the leading closed models, namely GPT-4o and Claude 3.5 Sonnet.
The 405B model is the most open, modifiable and cost-effective frontier-level model to date. An open weights model like 405B means developers will have flexibility to distill the model down to whatever size needed, so it can be tailored to any application. All at a much lower price than closed systems; Zuckerberg said in an interview that doing inference with the 405B model would be about 50% cheaper than using GPT-4o.
Meta’s long game here is building a vibrant ecosystem of development and resource sharing around the model. In addition to the models themselves, Meta is actively working on creating tools around them to make developing AI agents and apps easier, such as a reference system, safety tools and improvement mechanisms for the API. Meta has 25 partners ready to go with tools to fine-tune and distill the model, as well as the biggest cloud infrastructure providers, including AWS, Azure, Google, Oracle, and more.
You can try Llama 3.1 405B in the US on WhatsApp or at meta.ai.
Why it matters The new Llama models essential close the gap in performance between closed and open-source models. Zuck’s vision with Llama is not to become “the one” AI model but rather that there would be many powerful models in the world co-existing.
The wizard’s favourite AI newsletters
what i’m reading right now
TLDR 💨 - essential tech news in 5 mins. Read by 1.2 million people
simple.ai - 🕵🏻♂️ - deep-dives on AI by Hubspot’s co-founder
Bagel Bots 🥯 - best hands-on tips & tricks
The Neuron 😸 - easy weekday read on AI’s latest developments
Was this email forwarded to you? Sign up here. Want to get in front of 13,000 AI enthusiasts? Work with me. This newsletter is written & curated by Dario Chincha. |
What's your verdict on this week's email? |
Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.
*Past performance is not indicative of future returns, investing involves risk. See disclosures masterworks.com/cd