5 ms·
> People keep bringing up the "fake intelligence" vs "real intelligence" argument but it actually doesn't matter. It matters because you have to keep it in min
by nightfly 2y ago
> People keep bringing up the "fake intelligence" vs "real intelligence" argument but it actually doesn't matter.
It matters because you have to keep it in mind when judging its output EVERY TIME. You can't trust it to not tell you to poison yourself, or invent people/things that don't exist, or make up stories about people that _do_ exist.
- Asraelite 2y agoYou've completely missed the point of the parent comment. It doesn't matter if it occasionally hallucinates, because there are many use cases where that's okay and you can generate enormous value anyway.
- nightfly 2y agoI got their point and disagree with it because it's a highly cherrypicked example used to broadly dismiss a real and valid concern
- em-bee 2y agoi disagree with that. the point is that there are good and harmless uses of AI. the possibility of dangerous uses does not necessarily warrant the complete dismissal of AI unless a complete dismissal is the only way to protect us from harm caused by use of AI.
- skeledrew 2y agoDespite it being cherrypicked, it is a valid case. Hence it would seem you're knowingly disagreeing with a fact because it doesn't fit your purpose, which is disingenuous.
- reaperducer 2y agoHallucinate === Lie
- ryandrake 2y agoLying implies intent to deceive. An LLM has no intent.
- reaperducer 2y agoA lie is a lie is a lie. We don't need to make excuses for computers.
- ghaff 2y agoRight. Because people never make statements that are mistaken or even just patently untrue.
- nightfly 2y agoAnd now we have a machine to channel all those mistakes/lies!
- benterix 2y agoExcept that telling someone a lie, especially in professional context, can be completely discrediting, especially whenever trust is crucial. I work with people able to say: "I don't know", "I'm not sure", "Hard to say", "I need to do some extra work to verify this". Working with such people is very efficient because you don't waste context on fact-checking their every utterance. Can they be wrong? Absolutely, but it's relatively rare. Will they decide to just lie to me? Extremely unlikely, given the stakes.
- rimunroe 2y agoFurthermore, it also matters if we’re replacing existing systems with new ones with lower reliability, e.g. replacing technical documentation written by humans with ones generated by AI.
- xigoi 2y agoPeople can be held accountable for making wrong statements.
- ben_w 2y ago> You can't trust it to not tell you to poison yourself, or invent people/things that don't exist, or make up stories about people that _do_ exist. My mum was a big beliver in homeopathy and Bach flower remedies. Kept sneakily dosing me and my dad with one "for memory". She ended up with Alzheimer's just shy of 20 years younger than her mum. I could name a lot of elected officials over the years that made up stories about people, the hard part is picking one sufficiently uncontrovertial that nobody will object to the example. What matters is the rate at which these things happen, given AI is now sufficiently competent at presenting as a human to be a problem for those who need to know they're discoursing with a human (job interviews, grading essays, is this video call really with your relative who really needs an emergency payment or is it all fake, political propaganda).
- em-bee 2y agobut that is not relevant to ad-detecting. the worst ad-detecting can do is false positives, eg detect an ad inside a scene where someone is watching said ad on tv as part of the story or something like that. so your basic point still stands, you have to check if what it detects is really an ad, but beyond that the use of AI for ad-detecting is benign and harmless.
- drivingmenuts 2y agoIf you're having to constantly check the output, seems like it's not much of a stretch just to do the detecting yourself.