@Zicklag I'm in the same boat as you 🙂 I'm a general software engineer, but I recently started working for an AI startup, which has given me a chance to work with a lot of brilliant ML engineers and neuroscientists implementing their work, and have had to accelerate my understanding of AI programming which I always had an interest in but never had the time / justification / confidence until recently. I'm also on the long journey of re-learning math so that I can eventually get to the more advanced ML related math.
Recently I've started a group of neuroscientists, engineers and game devs trying to utilize deep learning AI for games, currently with LLMs but eventually with active inference (Bayesian neural network) and take the same ideas from active inference to try to apply them to our current technology in the meanwhile, but I'm not trying to solicit anything, and I know you're very busy, so I'll keep you in the loop over time and then you can tell me if you're ever interested in joining / collaborating! Those of us who are less experienced in ML / DL are also on the journey of learning more about AI through developing this project.
With active inference (which uses Bayesian methods), there is a very real possibility that we can reproduce sentient / human intelligence, or at least something indistinguishable from it. This would also be happening actively, like a game loop. I highly recommend watching these relatively short videos by my Chief Scientist and Director of Research at the startup I work for to explain why, and I think when you watch these videos it will be intuitively obvious why when this is implemented it will be significantly more advanced than reinforcement learning, and it's potential application in games: