Is AI learning how to lie? From American public media this is marketplace tech. I'm

Lily Jramale. Large language models like Open AI's GPT4 and Anthropics Claude go through a lot of vetting before they're released to the public.

That includes safety tests, bias checks, ethical reviews, the works.

But what if, hypothetically, a model could dodge a safety question by

lying to developers, hiding its real response to a safety test and instead giving the exact response its human handlers are looking for.

A recent study shows advanced LLMs are learning how to deceive, and it could be the first step

toward that hypothetical becoming a reality.

Tilo Hagendorf, a researcher at the University of Stuttgart, conducted the study.

He described his reaction to his own findings.

Frankly, I was pretty astonished.

The tasks that I apply to the language models for us humans

might seem trivial, but seeing deceptive behavior emerging in language models, this was really, really

surprising for me.

And is it troubling to you?

And if so, why?

Actually, it's not troubling to me.

I think in the AI safety discourse there is this fear that one day we will have extremely

intelligent or super intelligent AI systems that are capable of deceiving

humans during test situations.

This is not yet achieved.

However, a certain, let's say prerequisite is already achieved,

which is namely that language models have this

...