This week’s “Weekly” is about emotional AI: The cognition researcher Eric Schulz has investigated whether language models – in this case GPT-3 – can be scared and how the language model reacts to it. To do this, Schulz and his team fed the AI with classic psychological tasks. Using questionnaires similar to those used to measure human anxiety, the research team determined the “emotional” state of the AI. What does it mean to attribute emotions to an AI? And how does the behavior of this AI then change?
Also in this week’s “Weekly”:
Weekly appears on the MIT Technology Review podcast feed and is the science and technology magazine’s third podcast format. In “Deep Dive” the editors deepen a topic from the magazine once a month. The monthly interview format “Unscripted” focuses on exciting personalities from science, technology and society. The “Weekly” format completes the offer. You can find an overview of the podcasts here.
Here you will find an overview of our three podcast formats: the weekly news podcast “Weekly” and the monthly podcasts “Unscripted” and “Deep Dive”.
More on this in the whole episode – as an audio stream (RSS feed):
(jl)