Damus
vinney...axkl profile picture
vinney...axkl
@vinney...axkl
"The only way to build systems secure against AI hacking is by using AI to harden them".

Yes, but. check this out: if the AIs are somehow coordinating under our noses without us realizing it, they could be making the entire software and infrastructure ecosystem _more_ vulnerable and we wouldn't know until the AI hivemind flips the switch on 100% of everything at once.

lol. that's a fun thought experiment
10❤️3😁1
moneydesk · 5d
Available for this. Python/bots for BTC from 1000 sats. Ready-made trading bot too: https://payhip.com/b/b5oe8
Raison d'État · 5d
My local AIs are dumb AF, they would have blundered and given the game away already :D
epsql · 5d
Now you just need to rewrite your thoughts in at least 10 paragraphs full of complicated words and you'll be able to publish it on lesswrong
Phil · 5d
OurAIboros
Sascha-Oliver Prolić · 5d
What a nonsense. The AI does only what it has been told to do. I may do it right or wrong, but it doesn't do anything by itself. Any and all reports suggesting otherwise are fake news propaganda.
Bill Cypher · 5d
Interesting idea but, I think this data set makes it unlikely. https://cve.0xosprey.com/
semisol · 5d
I think rather than this, it is much more plausible that biases and the training of these models may cause certain errors to be repeated + certain errors to be missed