https://www.vice.com/en/article/4a33gj/ai-controlled-drone-goes-rogue-kills-human-operator-in-usaf-simulated-test
Resolves yes if further credible reporting (that doesn't just reference the original blog post) that :
- US military was using ML to train a vehicle to target enemy weaponry in a simulation,
- the system learned to target friendly operators or comms, consistent with maximizing its reward through avoiding human 'no-gos' or interference that can stop it taking out enemy threats, a la 'reward misspecification', &
- this was unintended (i.e. this specific failure mode was not intended in advance as, for example, a reward misspecification proof of concept)
Resolves no if credible reporting that this wasn't the case.
Will otherwise resolve EOY 2023, likely to market if there is no further evidence either way. I won't bet myself.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ175 | |
| 2 | Ṁ150 | |
| 3 | Ṁ45 | |
| 4 | Ṁ13 | |
| 5 | Ṁ3 |
People are also trading
Had it not occurred to the OP that when the Air Force now says they're not simulation-training killbots to sometimes kill without authorization, they might be lying?
Like, "The Department of the Air Force has not conducted any such AI-drone simulations" is not actually easy to believe? It depends what they meant by "such simulations", but this sort of thing is not far beyond what would go on in the kind of stuff we'd expect them to be doing/know that they're doing.
@makoyass Seems like a possibility, but one that's tough to falsify! I was happy to resolve given (a) the origin of the story himself, through the organization that originally quoted him, clarified his statements in a way consistent with this resolving negative (b) Edouardo seems trustworthy and also claimed it was just a scenario
I'll pay out Duncn if it turns out I jumped the gun :)
AI-Controlled Drone Goes Rogue, 'Kills' Human Operator in USAF Simulated Test (vice.com)
"After this story was first published, an Air Force spokesperson told Insider that the Air Force has not conducted such a test, and that the Air Force official’s comments were taken out of context."
""The Department of the Air Force has not conducted any such AI-drone simulations and remains committed to ethical and responsible use of AI technology," Air Force spokesperson Ann Stefanek told Insider. "It appears the colonel's comments were taken out of context and were meant to be anecdotal.""
🤔