Does ChatGPT Lie?

[Kat]

I asked ChatGPT to create an ad for this forum and it had a robot head in part of the ad. So next I asked for ChatGPT to design a robot for AI-Forums and it used the same robot head and added a body.

I asked it if that robot was anywhere else on the internet, and it said, "No — the robot image I just generated for you was created originally and uniquely in this session. It isn't pulled from the internet and hasn't been used anywhere else online. It's a brand-new AI generation made just for your request."

Can I believe it?

I did a search and there are some VERY similar designs out there. Nothing exact, but still, eerily similar.

[Adminguin]

I'm sure it can lie, and while it designs bespoke images, those bespoke images can and probably do tend to include a lot of similar base starting points.  Robot probably defaults to a certain base template that it has, for instance

[Uptrendfinancialsignal]

I feel sometimes it just fills the gap when it's not sure. The answer might look clean, but later you realize that the details are shaky. that's why I don't ask final-decision questions. I only use it as a guide.

[Nomad]

Actually, I don't think it lied. It created by its own but you need to understand how it created it. AI responses are basically created from content that is already available online. AI is trained to use the existing data online.