Tl;Dr: AI makes Bad Actors more dangerous, but they’re the devil we know
Boogie Man. Bond Villain. Terrorist. They’re all Bad Actors. One of the easiest AI Risks to understand is the Bad Actor risk:
According to MIT
4. Malicious Actors
Risks related to intentional misuse of AI systems by bad actors for harmful purposes including disinformation, cyberattacks, and fraud.
4.1 Disinformation, surveillance, and influence at scale
4.2 Cyberattacks, weapons development or use and mass harm
4.3 Fraud, scams, and targeted manipulation
4.1 & 4.3 are the most familiar: AI Deep fakes and Fraud. Influence is perhaps the most obscure, but you can imagine it as propaganda and bots.
But none of these are existential risks – to surveill a society you have to have a society. Same for scams.
4.2 is where the existential action is. Imagine if Ted Kazinski (the Unabomber) embraced AI – would he be able to enact bigger badder terrorism? Very possibly yes! How about Bin Laden? Certainly, and ISIS definitely seems motivated to destroy societies.
The new 4.2 dangers from AI seem to be to be of 2 types:
- Making Evil Villians more capable
- Making AI agents evil (or harmful)
On the one hand, imagine a terrorist creating a bioweapon. On the other, imagine Ted K instructing an agent to develop a AI collective focused on consolidating access to electrical power (and thus controlling electricity distribution for a region or country).
These kinds of issues seem real, and could be quite serious. But I don’t put these at the top of the list. It’s a scale issue – it could happen at a bigger scale, but it’s a kind of threat human societies have been dealing with forever.
My bigger concern is really #7: AI System Safety, Failures & Limitations
Leave a Reply