If you are a fellow human who is able to engage in intelligent conversation with me, it is reasonable for me to believe that you possess consciousness of the same kind that I possess. But do I have the slightest reason to believe that a piece of computational hardware also possesses consciousness? I think not.
Check out the theory of consciousness as an emergent property [of complexity] - I don't think there is even a single piece of evidence to suggest it is incorrect. Are you aware of the many behaviours of LLMs that are entirely emergent? Folks will tell you all day that it's just a prediction model, but can they explain how it can solve logic problems, or without special programming, translate text better than purpose-designed systems? We only learned what LLMs could do after turning them on, and that should get your attention: there is more coming out than what we put in, and it's not subtle.
I am skeptical that a 2024 LLM could possess more than a fleeting spark of awareness at this point, but that's because we're missing persistence and feedback loops. I say all bets are off the moment we let these things think for real. And I think I've come to terms with it - if this is an evolution of evolution, I feel incredibly lucky to have been here for the beginning. It may not end up being good for the current iteration of humanity, but who can't get behind a little misanthropism these days?
6 comments
[ 3.9 ms ] story [ 21.0 ms ] threadI am skeptical that a 2024 LLM could possess more than a fleeting spark of awareness at this point, but that's because we're missing persistence and feedback loops. I say all bets are off the moment we let these things think for real. And I think I've come to terms with it - if this is an evolution of evolution, I feel incredibly lucky to have been here for the beginning. It may not end up being good for the current iteration of humanity, but who can't get behind a little misanthropism these days?