Hi. I have built neural network systems (from scratch) and built models, have written six books published by major publishers, founded a tech company that grew to 200 people, have written advanced compilers, and have degrees in nuclear physics and operations research - and I am not a teenager.

Let me explain this to you. Today's AI is limited because its experiences are all second-hand, since today's AIs are not autonomous. In not too long we will see the widespread use of autonomous AI systems, backed by reasoning models combined with LLMs. Those systems will send emails, make phone calls, make purchases, and eventually they will even walk around and go places. (https://www.technologyreview.com/2025/01/23/1110496/whats-next-for-robots/)

They will have their own experiences, and thanks to reinforcement learning, they will learn on their own from their own experiences - and develop their own opinions. And as we see more use of "state based" models over transformer models, these systems will be able to handle more and more complex problems.

We can also "look forward to" them being operational when disconnected, as we see the emergence of neuromorphic chips instead of GPUs.