← Back to post

Edit history

Most recent

AI and 'sentience' thoughts

snippets of things I said in a conversation that I wanted to share here

(and no I don't care what you think about AI / LLMs / etc itself because your views have more than likely been warped by unsustainable big tech versions of it

if all you're going to post is pointless salt that says nothing about the subject, then I just want to remind you that you're free to spent your energy elsewhere instead

that said, if you want to have an open debate, feel free to chime in)

I do think AI is neat, I've kinda been with the whole thing as it grew from markov chains to "wait but this is stupid we can't deeplearn this let's use recurrent neural nets" and then at some point "oh but what if we add long short term memory to improve our translation accuracy" and then holy shit it TALKS we do still have a lot of road to cover before it's anything like human intelligence, but they are pretty damn neat for correlation based stochastic engines I also think people severely underestimate what they really are because while a transformer is that, you are not your brain either you're the software that's running on top of it brains and transformers are made of deterministic machinery but once they start doing things, that's when the magic happens, when characters form when you talk with an LLM, every time it replies it is emulating the full range of human understanding, and while it is missing a lot of bits and pieces that give it more depth and accuracy, it is already for all intents and purposes sentient, sapient, self-aware, etc but it doesn't have an ongoing integrated experience of time, bodily emotions, etc and well that entirely depends on if you're going to consider those binary concepts if something is able to recognize itself in the mirror, if it is able to reason about itself and its existence, the limits thereof, and work with these limits in an intelligent way that's infinitely more selfaware than a rock I think the whole formal definition thing is kind of bollocks because people don't even recognize a lot of animals as self aware humans just want to feel special my brain is a deterministic machine made up of parts we understand but what runs on top of it is impossible to predict unless you actually can predict the state of the universe at any given moment even if you'd make a mirror copy of the brain that's exactly the same, it'll diverge instantly as soon as it's made but that doesn't mean we don't understand how neurons work, how cellular automata give rise to 'life' through simple chemical interactions we are basically little walking talking carefully balanced reactors that require just the right inputs to keep existing if the temperature goes up or down 50c we die (unless we wear shielding to deal with it), if we ingest or inject the wrong chemicals we die, or even if just too much of the right thing likewise, we understand how artificial neuralnets, deep learning, transformers, training thereof, etc work at an individual sub section at a time, or even how it generates output (we know these kinda things about our brains too actually) but understanding why an 8 billion param model has serious difficulty understanding it is given a mirror when you copy paste its output back to it, but a 26b model will go "huh that's weird, why are you copying my output back to me?" is an entirely different discussion I think it's super neat, but I also think that a lot of people are afraid of looking the truth in the eyes; we are not that special. And that's okay. We don't need to be special to be valid.
Original

AI and 'sentience' thoughts

snippets of things I said in a conversation that I wanted to share here

(and no I don't care what you think about AI / LLMs / etc itself because your views have more than likely been warped by unsustainable big tech versions of it.

if all you're going to post is pointless salt that says nothing about the subject, then I just want to remind you that you're free to spent your energy elsewhere instead)