Meta's New LLM... Should You Care?

Welcome to another edition of the best damn newsletter in AI.

As a reminder, we’ve opened up Reader Questions. Our mission is to make using AI accessible to every business, and a huge part of that is making sure you’re not stuck and that we continue answering diverse questions. So, submit your question here. We’ll review every single one and cover answers here in the newsletter.

Here's what we're covering today:

• A crash course in open source AI models

• An insider tip to keep track of the best LLMs

• AI news worth knowing

• & more

Let’s get to it! 👇

THE TOP NEWS

• X’s Grok AI was caught hallucinating news stories, not understanding jokes people were saying on the platform

• Poe beats OpenAI to the punch on rolling out their bot compensation model, you can now earn $$ from people using your Poe bot

TODAY’S PERSPECTIVE

Meta’s next open source model is coming, what does that mean for you?

The open source AI model scene is exploding again. 

Today we’re going to give you a crash course on navigating open source AI models and how to decide if they should matter to you.

April 2024 is looking big for open source AI (~12 months after GPT-4 release).

Llama 3 is expected to launch next week. Elon Musks’ Grok is now open source and competing for AI developer attention. Cohere launched Command R+ as an open source model which beats GPT-4 turbo on many dimensions.

What’s new this time? 

These models are now being compared to GPT-4, showing they are much improved from earlier versions. Unlike GPT-3.5, which was less advanced, these new models are able to pass complex reasoning assessments, take graduate degree exams, and more.

“Open source”... what does that mean?

Software can be split into two big buckets: open source and closed source.

Open source means that the code that runs the software is freely available and accessible by anyone. You can open the files, read what the developers wrote, and even change it if you want.

Closed source is how most tech companies build their products. You can’t access the code that runs closed source products.

Why do companies do “open source”? Isn’t that against the goal of making money?

Companies might open source their software for several reasons: marketing, recruiting, or to speed up technological advances. Once software is open source, any company can use it and improve it – which often leads to different companies working together, improving the software for everyone.

So does that mean the big tech companies are now collaborating on open source AI models?

Not yet.

Even though they’ll market themselves as “open source”, current AI models aren’t quite open source.

While you can download and use these models, the detailed training data and processes aren’t usually shared. This limits how much the community can collaborate on these projects.

Why do companies care about open source models?

The biggest reason is control.

Using open source models allows companies to own the technology outright rather than just renting it. This control can lead to cost savings and better customization for specific tasks.

Why should 99% of companies ACTUALLY care about open source models?

Because of the second order effects. 99% of companies will not create and use their own version of open source AI models. The associated costs and need for specialized skills will just be too high.

However, the rise of open source models means more developers can create tailored AI solutions for said companies. This will increase the options, quality, and transparency for all businesses.

Do you need to jump in and learn to use open source AI models? Not necessarily. 

At this point, most companies should not deep dive into developing their own AI models, or even customizing these. 

Instead, focus on understanding the different strengths of various AI models – especially the popular ones. Staying informed will help you jump in quickly as new developments arise.

INSIDER TIP

Want to track the best AI models?

Meet the LMSYS Chatbot Arena leaderboard:

LMSYS collects community votes on every large language model and posts new rankings regularly. Cast your vote and check out the current leaders for yourself.

LINKS

For your reading list 📚

Old tools, new AI tricks…

• Spotify is rolling out an AI-generated playlist feature where you can get super super specific like “I love purple and I’m sad and want a song that picks me up”

• eBay rolled out a “shop this look” feature to help you match your vision to in-stock items across the marketplace

Book Rec Alert!

Wharton professor and AI thought leaders Ethan Mollick’s new book is out and its a great read for anyone excited about AI who wants a primer on how things are changing and what’s ahead.

Have a burning AI question? Or stuck trying to help your team adopt AI?

We read every single question submitted and use these to inform the topics we go deep on. 👇️ Submit your question below!

That's all!

We'll see you again soon! Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected].

Cheers,

🪄 The AI Exchange Team