Oddbean new post about | logout
 My answer might be tricky. I don't want the current AI architecture (Transformers) to define the future of LLMs or ML, that would be disastrous. We need a breakthrough!

It's not that I can't live without it, but AI has countless potential use cases across fields like computer science, medicine, and space exploration. It's already penetrating every industry, these new generative models being just a part of it.

That's why I believe it's crucial to shape AI for a better, less centralized future.

What amazes me is how the open-source community has kept pace with frontier models. If OpenAI disappeared tomorrow, we could still rebuild up to GPT-4o using just open-source models.

I hope it makes sense 😅, I'd love to hear your thoughts. 
 I think a lot of AI news is hype. Companies are claiming AI is the reason for layoffs because it sounds better than, "We hired a bunch of people we didn't really need, screwed up our finances like a moron, and now have to get wise."

Meanwhile, I do see some benefits of LLMs, but the benefits are small for me. I'm willing to recognize that I may not know how to leverage them best, but I always have to check their work, and almost always find bad info. With the same amount of time/energy, I could have gathered the info myself, and have a better understanding.

Since I've only used the free AIs I've seen mentioned on Nostr, maybe I'm using sub-par tools, but if open source really is keeping up, then the delta isn't that large.

I think we will see more decentralized efforts popping up, and that collaboration could easily surpass the corporate projects.  

One of the side benefits of Nostr is how people that never considered software development are asking how they can get involved. People that haven't written code in years are firing up IDEs and compilers/interpreter. The tide is rising, and not just in our digital equivalent of a college town.

Persistence pays! Patience is required though.