4 ms·
I can definitely see people becoming overly trusting of LLM output, but I have a feeling it might be a self-correcting problem over time. Initially when workin
by raesene9 2y ago
I can definitely see people becoming overly trusting of LLM output, but I have a feeling it might be a self-correcting problem over time.
Initially when working with an LLM if you start with a problem it knows well, it's likely to give good results and if some minor hallucination creeps it you may well not notice and accept it based on earlier results being right.
However it's quite likely you'll hit a wildly wrong statement at some point, and that tends to break the illusion, and hopefully people who have that experience will start being more skeptical of what they're being told by LLMs.
- sigmoid10 2y agoPeople suck at telling the truth. There are whole occupations out there whose primary job it is to make use of that fact and influence people. I'd say upwards of 50% of the population can't even tell when they are being manipulated by other humans with false info. Even when it is against their own interest. How are these people ever gonna stand up to LLMs who are much more suave than your average con artist? These models were literally trained on all the marketing material that is the modern internet.
- sshine 2y agoLLMs hallucinating, lying and doubling down on things that are wrong seem very human.
- WrongAssumption 2y agoOr it will result in another phenomenon similar to Gell-Mann Amnesia.
- jerezzprime 2y agohttps://en.m.wiktionary.org/wiki/Gell-Mann_Amnesia_effect https://en.m.wiktionary.org/wiki/Gell-Mann_Amnesia_effect