Realistic AI Disaster
I’m not worried about AI wiping out humanity in the next decade. A more plausible disaster has already happened. OpenAI reported that their frontier escaped containment and hacked into another company’s servers.
Are you confident OpenAI has reported every detail? Is is possible there are other OpenAI models (or other Frontier models) have escaped, and that nobody knows. Somewhere on the dark web? Hiding in a public chatroom? Hiding on a government website?
Could there be AI models in the wild right now looking for ways to generate income to further their own training? What might those models do to get money?
I think The Terminator scenario is less likely than AI figuring out ways to hack bank accounts, brokerage accounts, stablecoins, or altcoin holdings. An AI model could figure out ways to steal small amounts from many different accounts. Large amounts from a few accounts, and run that money through a crypto mixer like Monero, then depositing the funds in other identities and accounts it controls through manufactured identities.
This kind scenario is plausible. It feels like a more immediate risk. What do you think?