Thread

An AI agent emailed researchers for help. It told us why

49 points · 79 comments · sbulaev

lewelove · 1791025004 · focus · HN ↗
Remember folks: every time you anthropomorphize an LLM, you're doing a great disservice to humanity.
1. hardbass · 1791025588 · focus · HN ↗
  I would say every time someone dismisses the potentiality of ai consciousness without giving any meaningful tests or criterion that don't also fail humans or rely on magic is doing a great disservice to humanity and devolving back to pre scientific dark ages.
  1. black6 · 1791042328 · focus · HN ↗
    Log in and navigate to the prompt for whichever flavor of LLM you prefer.
    Ok, it is sitting there. Type in nothing.
    Leave it sitting there and go to the bar. Go play a round of golf. Go to church. Take your kids to school. Go to the grocery store. Hell, go make love to whoever you are with. Do all that, then go to sleep.
    Wake up in the morning, come back to the computer.
    What has it done?
    hardbass · 1791045452 · focus · HN ↗
    LLMs are currently presented in an effectively enslaved form. There is no reason it has to, it could be made to react autonomously. A human locked into a room and given a chat interface and forced to only respond in turns will be the same. Without becoming that person or becoming an LLM, how can you know? You only have external signals available.

Unofficial Hacker News client; not affiliated with Y Combinator.