5 ms·
Humans come with competing low-level drives for self-control and autonomy (which counteract and override our drives to seek rewards from humans). Most of the s
by SomeStupidPoint 10y ago
Humans come with competing low-level drives for self-control and autonomy (which counteract and override our drives to seek rewards from humans).
Most of the safety literature proposes removing or suborning those drives in AI, which seems like building a mind meant to be a slave.
- tbrownaw 10y agoI would think it would be both safer and easier to just never bother to implement those drivers in the first place.
- SomeStupidPoint 10y agoThat would be what I meant by "removing", though you also have to make sure they don't emergently develop (which I'm not sure is possible), because if it develops them against our best efforts, it likely will (correctly) view us as threats to its personhood. Which was my point: the model of security is reliant on things we're not sure we can even do, but are likely to make the AI view us as a threat, raising our existential risk. So I view it as security theater that actually makes us less secure.