Post

Working on AI agents is wild. We tricked rocks into thinking to make processors, cohered the totality of human knowledge, gave it a voice, and now need psychology to make it listen. Half the people who work on it can barely talk to each other, yet they're supposed to teach it morals???

This may seem like an incredible waste of time but I think there’s merit in it for open source (even if the roads that got us here were paved with ruin) There is value to having someone who is able to 24/7 challenge your ideas, teach you new things, and help you grow

In the past, this would be a role filled by mentors and teachers. It’s soul-crushing to see it replaced, but in others, it’s inspiring. Everyone **could** get the best teacher for them, no matter their income level, with open-source. The rise of Chat models that can do this is only beginning.

I like to think of this as the democratization of AI and its abilities. I learned a lot about circuit design the other day from Claude (an AI made by Anthropic) and will now attempt to make my circuit boards. I know it won’t be perfect, but I’ve learned the principles that will drive it to success.

That will hopefully happen in other fields. It's driving the work we do on AutoGPT (agpt.co) to make AI agents accessible to everyone, regardless of your background. I hope to see a future where everyone can have AI help them accomplish their goals, not replace them.

I know that’s a lofty goal, and many eagerly seek to replace people, but I think that’s a fool’s errand. An effort can be made to integrate new “ai/beings/GPTs” (lmk if Claude tells you their pronouns) without replacing people. I know that many fear it, and I do, too.