Science Café "Does ChatGPT understand your emotions?"

Many people use ChatGPT and other chatbots that operate using "Large Language Models." They're very useful, but they're also heavily criticized. They claim they're essentially stupid and merely repeat texts they've read. They claim they can't "really" think. Their advice often consists of hallucinations, beautifully worded to make you feel comfortable reading them. When you talk to them, they pretend to like you, which is harmful. And finally, they don't truly understand our feelings because they're not alive.

In his presentation, Joost Broekens, associate professor at the Leiden Institute of Advanced Computer Science (LIACS), will explore with us what we think of all these claims and whether we can figure out what it actually means when an AI understands something. What do we actually expect? For example, should we value a comforting response from an AI and sound advice about what you could do to feel better, or not? But also: how can we ensure that we use these kinds of chatbots in a way that maximizes their potential?
After the break, there will be an opportunity to discuss this with him.

Prior to the lecture and during the break between the lecture and the discussion, PLÅNET will perform. This indie band plays spacey experimental funk rock with a universe of stylistic influences.

Language is in Dutch.