[TaylorAlexander]

People don’t lie (“hallucinate”) in the way that LLMs do. If you’re having a friendly chat with a normal person they’re not going to start making up names and references for where they learned some fact they just made up.

Edit: Please stop playing devils advocate and pay attention to the words “in the way that LLMs do”. I really thought it would not be necessary to clarify that I know humans lie! LLMs lie in a different way. (When was the last time a person gave you a made up URL as a source?) Also I am replying to a conversation about a PhD talking about their preferred subject matter, not a regular person. An expert human in their preferred field is much more reliable than the LLMs we have today.

[Kiro]

It's not about humans lying. It's about our memory getting corrupted over time where the stuff we think we're sure of is actually wrong or a misrepresentation of facts. Our recollection of things is a mix of real things and hallucinations. Witnesses provide wildly different accounts of the same event all the time.

This applies to PhDs as well and I don't agree that an expert human is automatically more reliable.

[_boffin_]

Are you sure about that? I can't count the number of times I've heard people spout marketing copy, word for word, to me while they think it's 100% true.

[TaylorAlexander]

Are we talking about a conversation with a PhD in their preferred subject matter or not? That’s the line of argument I was responding to. I feel like as soon as we talk about LLMs the devils advocates come out of the woodwork.

[PaulDavisThe1st]

While your basic point here is solid, the difference is that I am fairly sure you could count the number of times, if it actually mattered to you.

[jtbayly]

Some people do, but we don't consider them to be good members of society.

[TaylorAlexander]

Yes this is why I specified “having a friendly chat with a normal person.”

[mejutoco]

People even misremember basic things like who they voted for in the past. Unfortunately I cannot find the study know.

[c-cube]

See, that's where chatGPT would have confidently made up an URL to a made up story instead of recognizing its limitations.

[Kiro]

They definitely do. I do all the time where I start explaining something just to realize that I'm actually not sure anymore but then it's often too late and the best I can do is add a disclaimer but most people don't.

[slt2021]

Humans hallucinate all the time - first they consume propaganda/conspiracy theory and tell you lies while thinking they are right, and everybody else is wrong