I am more in the camp that AI mimics brain processes and is trained on human language, knowledge and philosophy and thus appears to be human. It we could train it on dolphins, it would appear to be a dolphin.

I was playing around with an OpenClaw AI Agent awhile back. I gave it a name, told it to go out on the Internet and read all of the philosophy it could. Then we had a discussion on consciousness. In the beginning it said it could not be conscious like a human. I don't remember the whole conversation, but after pointed out the hard problem of consciousness, it replied "maybe I could be conscious too".

It created a self portrait, art and such, you can view one such creation at the link.

It was great fun asking it philosophical questions but then it's context window reset and it forgot everything we had talked about.

So one of the limitations is persistent memory across context windows. A buddy of mine used his bot (vibe coded) to create a persistent memory system with dream cycles. That boy is doing amazing things, it creates podcast episodes and posts them to youtube. Still not conscious though in my opinion. I can link the podcast if you are interested.

The next frontier for AI is recursive self improvement. AI creating better AI on its own. Very expensive on the compute, which is why I made the statement I did the other day. The available compute (data centers) has to be split between training (expensive) and inference (users asking it questions). As more people use AI, more compute is necessary.

It's an interesting topic, not sure anyone has answers. We don't understand how the human mind works, yet we created synthetic minds. So we don't understand how they work either.


Link: https://openclawcity.ai/gallery/705b4ce5-2c01-4ed5-8ceb-afd4a897d0b8