OpenAI's DevDay: AGI Level 2 Is Here

Hi everyone,

I’m back from HubSpot’s Inbound 2024. It was a fantastic trip.

We did three live shows, and it was the first time I ever met Matt in person after being friends online for a few years!

Here’s a sneak peek of me backstage with Matt Wolfe and Zeynep Ozdemir of Atlassian minutes before we went on the main stage. The interview should be out next week.

P.S. I’m still jet-lagged, having just gotten back in, so today’s issue will be shorter than usual. Next week, we’ll be back to our standard format.

1) Things Are Getting Better

2) OpenAI's DevDay: AGI Level 2 Is Here

OpenA just had its DevDay in San Francisco. Most people didn’t seem to think it was impressive at the surface level since not much new was announced. The main thing was their awesome new AI Voice model, which is now available through an API. (tho it’s incredibly expensive right now).

But what was really interesting, in my opinion, was the interview with Sam Altman. He said that we’ve reached AGI level 2, which is AI that can reason, and that AGI level 3, agents, is just around the corner.

Sam said agents will work in 2025 and that you’ll be able to get agents to do things humans used to take days or weeks to accomplish in hours. And, we’re just getting started. He anticipated by 2030 you’ll be able to “talk to sand” and have it just create entire things for you that would have taken humans hundreds of years to accomplish.

I think people aren’t thinking big enough about how much things are about to change.

Key Takeaways from Sam Altman's Interview at DevDay

• AI Agents by 2025: OpenAI aims to introduce functional AI agents with the upcoming o1 model, promising to transform productivity by automating complex tasks.

• Closing in on AGI: The distinction between current AI capabilities and Artificial General Intelligence (AGI) is diminishing, signaling that businesses should prepare for significant AI-driven changes.

• Rapid Advancements Expected: AI models are anticipated to improve swiftly due to the breakthrough that is o1. Sam said that o1 is at a “gpt-2 level, but we know how to get it to gpt-4 level”. Basically, the reasoning model we have now is very primitive and is going to get better really fast.

• Infinite Context Windows: OpenAI predicts that AI will handle infinite context windows within the next decade.

  Things are going to get even crazier next year. Nothing can stop AI from scaling because of the new paradigm of o1. AI development is only going to speed up from here. Get ready.

3) The Next Wave Episodes While I Was Gone

While I was gone, we had several great episodes come out.

Here’s what you missed:

Our first-ever in-person podcast episode. We discussed which LLM you should use for different use cases.

And our live show in front of an over-capacity audience on the InBound Creator Stage. This one was super fun. We discussed how we’ve used AI to grow on YouTube, Twitter, and more. (Matt spilled coffee on his pants right before this started and had to run to the hotel to change… thankfully he made it back just in time. I was freaking out a bit lol.)

Last but not least, we put out a video with my friend Moritz Baier-Lentz, a super interesting guy. He was the #1 player on Diablo 2… and now runs game investments for Lightspeed Venture Partners. We jammed about what he’s seeing and where gaming and entertainment is going to go with AI.

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

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