Justin's AI Newsletter

The most interesting content, news, and insights in AI each week.

--

Hey friends 👋,

My birthday is tomorrow! And the funny thing is I started this newsletter exactly a year ago on my birthday.

So, happy 1st birthday to this newsletter! 🎉

Here's to a fantastic year in AI and many more ahead...

And as a birthday gift to this newsletter, consider sharing it with your friends. Don't let your friends miss out on AI knowledge! 📬🎁

Now, let's dive into my AI takeaways from the week…

---

AI DRAMA

🤯 Google Gemini: Beating ChatGPT

"We always viewed our mission as a timeless mission," says Google CEO Sundar Pichai. "It's to organize the world's information and make it accessible and useful." Well, it looks like mission completed with Google's new AI model Gemini.

Google released a demo for Gemini last week that took over the internet. Gemini is multi-modal meaning it can understand text, code, audio, image, and video. The demo showed Gemini's ability to reason and interact with the real world from identifying a rubber duck to creating games from scratch.

There's been some debate over how much of the demo was real and how much was "faked", but either way, the model is top-notch. It's beating ChatGPT at 30 out of 32 categories. "Each of the 50 different subject areas that we tested on, it's as good as the best humans in those areas," says Google DeepMind CEO Demis Hassabis.

If 2023 was the year of GPT4 and ChatGPT, then I think 2024 is going to be the year of other companies like Google releasing models that are better and better. This is the beauty of competition, right? We're going to keep seeing better AI models released into the world. Even if you love ChatGPT, it's worth trying other models because each has its different strengths and weaknesses.

More on that soon...

---

AI NEWS

🕵️‍♀️ AI Agents: The name is Bond, JAImes Bond

I made a TikTok this week about what's coming for AI in 2024...

In a year from now, I think AI products are going to look very different from the current state of the market.

Next year, we are no longer going to be just chatting with AI. It's going to be all about AI running asynchronously and completing tasks on its own. Imagine the smartest employees who accomplish tasks without constantly monitoring their progress—that's the direction I believe AI is moving towards.

These are called AI Agents, and there's going to be agents running at all times for everything from outbound sales to making content based on events in the news.

That's why we're building AI Agents at Cassidy: to give you the upper hand. Remember, you're not going to get beat out by AI, you're going to get beat out by someone using AI better than you.

---

AI ADVANCEMENTS

😎 DeepMind introduces FunSearch

Earlier this week, DeepMind introduced FunSearch, a method to search for new solutions in mathematics and computer science.

So far, LLMs have been great for functions like reading, writing, and creating images, but what if they could help us discover entirely new knowledge?

Is it possible? It turns out yes.

FunSearch represents the first time a new discovery has been made for a challenging open problem in science or math using LLMs. FunSearch discovered new solutions for a decades-old unsolved problem in math called "the cap set problem", and was also used to discover more effective algorithms for the "bin-packing problem" which is used to make data centers more efficient.

"FunSearch works by pairing a pre-trained LLM, whose goal is to provide creative solutions in the form of computer code, with an automated 'evaluator,' which guards against hallucinations and incorrect ideas. By iterating back-and-forth between these two components, initial solutions 'evolve' into new knowledge. The system searches for 'functions' written in computer code; hence the name FunSearch," explains the team at DeepMind.

For a full technical explanation of FunSearch, check out DeepMind's announcement and their paper published in Nature.

---

⭐ AI Quick Hits

1. The treasure hunt for more GPUs is over; Sam Altman announces ChatGPT Plus is back for everyone. [Read here]​
2. Stability AI launches membership fees for access to their models for commercial purposes [Read here]​
3. A new OpenAI paper explores how humans can supervise AI models that are much smarter than them [Read here]​
4. Balaji Srinivasan talks AI Gods, human-AI symbiosis, and AI control 🤯 [Watch here]​
5. OpenAI opens up applications for their 6-week program Converge-2 for engineers, designers, researchers [Read here]​

Stay tuned for more next week...

---

Thanks for reading,

Justin

​

P.S. Want more quick, hot AI takes?

Be sure to follow me on Twitter & LinkedIn! 🔥🤖