Spoke with a dev today who told me he doesn't believe AI is such a big deal, perceives of it as merely a new tool, and that it won't change much about his plan for his life or what matters to him. Huh.

solidity & product. i interview crypto builders @ web3galaxybrain.com

I find those most knowledgeable about the tech tend to have mostly extreme positions. Either it's nbd and just a tool or it's full on existential risk our species is ending and getting replaced by AI in our lifetimes.

The limitations are clear and we're tapping out available training data, but also the rate of progress has been insane the past decade

We might be just one novel architectural insight from orders of magnitude better performance that gives real versatility 

Or the whole thing's cooked, and now incremental gains only

mm, architectural improvements should reduce data requirements. babies don’t need so much data to learn.

Could be but gpt4 is so smart i wouldn’t believe it’s a human communicating with me because it’s so rigorous and tireless and fast. Trajectory feels solid for solving human+ intelligence. But nuance will appear as I get better at distinguishing LLM output (maybe?). What humanity won’t it be able to mimic? idk

it's also possible the models are still many orders of magnitude too small compared to a human brain, because there could be subcellular computation going on in neurons we don't yet understand

very speculative and most scientists are dismissive of the idea, because it's usually attached to quantum woo