Making an AI available online to play with is almost certain to invite trouble. People will naturally want to break and prove it is dangerous, unreliable and should be shuttered. I am no fan of Meta or anything they do, but in this instance the reaction of the crowds and the machine's response to it was unlikely a Meta-only problem. Humans do want AI to fail - it's the only way to prove humanity is uniquely valuable and cannot just be programmed. It's one thing to game the AI to make it fail and make its creators lock it way as happened in the Meta instance. But there is AI out there that will fail of its own volition leading to unknown consequences.
AGI (artificial general intelligence) can be seen as a superset of all NAIs and so will exhibit a superset of failures, as well as more complicated failures resulting from the combination of failures of individual NAIs and new super-failures, possibly resulting in an existential threat to humanity. In other words, AGIs can make mistakes impacting everything. Overall, we predict that AI failures and premeditated malevolent AI incidents will increase in frequency and severity proportionate to AIs’ capability.
Comments