#rp24 speaker Eva Wolfangel: Are chatbots on the road to super intelligence?

Chatbots suddenly seem to be able to do everything: they argue, they plan, they think, they show intelligence and seem to have a precise idea of the world. Perhaps it won't be long before they get equipped with robotic bodies and change history?

At re:publica 24, Eva Wolfangel will help us unravel the black box of AI large language models – which, on closer inspection, is not so opaque after all. Is this black box perhaps just a pretence, so that tech companies can perpetuate the myth of impending superintelligence?

Eva will tell us about some pretty crazy experiences with chatbots: They suddenly fall in love, want to break out of the computer or outdo the other person in logic puzzles. They reveal other people's secrets and know how to repair water damage – and if only they had a robot body, they would do it, too…

In research, emergent capabilities are tasks that, for example,  large language models (LLMs) can execute without having been explicitly programmed to do so. 
Have LLMs developed a model of the world solely on the basis of text data? Time and again, individual phenomena - such as superhuman intuition - are discussed as evidence of these impressive capabilities.

But what if we take a closer look? Do language models truly understand, and how can this be explained? What does "understanding" even mean - and how do we humans do this? Are these astonishing abilities real or hype? And is generative AI leading us straight to super intelligence?

Eva is a journalist, speaker and presenter. She received the Reporterpreis (Reporter Award) 2020 for her report "Reality Sucks" and the Georg von Holtzbrinck Prize for Science Journalism 2020. In 2019/20 she was a Knight Science Journalism Fellow at MIT. 

At #rp24, we look forward to Eva's fascinating introduction to the world of chatbots, large language models and generative AI.