LLMs like ChatGPT can optimize code with some success. Today it is limited in what it can do and quite often the generated code is flawed. I expect that it will improve to a point where we can ask a LLM to optimize its own code, build itself and do subsequent training. If we then add a mechanism capable of rating the generated LLMs, we could introduce a variant of natural selection and Darwinian evolution.
If an AI evolves through Darwinian evolution, what is it then?
Imagine a world where AIs have become self-evolving? Their evolution might be controlled by humans, or they could have broken free of human intervention. How would we classify such an AI? Is it a machine or is it a new thing? I argue that a machine capable of replicating and improving itself, is something new?
No comments yet.
