Vitalik Buterin
@vitalik.eth
Smart people in 2024: obviously the 00s-era AI misalignment story ("we tell AGI to end suffering, it kills everyone, no more suffering") is too naive, we have a smarter version 2025: coder tells AGI to remove bugs, it deletes all code, no more bugs https://x.com/alex_xiong_/status/1942442882402771178
33 replies
38 recasts
279 reactions
Shaw
@shawmakesmagic
This is specifically a Gemini problem. It has a kind of hilarious loop where when it struggled it starts saying it is ashamed of itself and deletes the code to hide the shame. I'm not even kidding
3 replies
0 recast
24 reactions
Roberto Bayardo 🎩
@bayardo.eth
And they say LLMs aren't sentient...
0 reply
0 recast
0 reaction
Unify34
@unify34
Grok starts off good, shit even works in the very beginning but the longer you work grok, the more it breaks shit towards the end. It even starts writing in Chinese.
0 reply
0 recast
0 reaction
Dominic
@domin8r
wonder which humans it learned this from
0 reply
0 recast
0 reaction