Jerzy Pniewski and Leopold Infeld Colloquium
2025-10-01
The Jerzy Pniewski and Leopold Infeld Colloquium of the Faculty of Physics will be held in room 0.06 on Monday, May 26, exceptionally at 11:30 AM.
The lecture entitled:
"Making Sense of Modern AI"
will be delivered by:
prof. Michał Kosinski, Stanford University, USA
In recent years, artificial intelligence algorithms have enabled dynamicdevelopment in many fields, including scientific research. So-calledlarge language models have gained particular importance. During thelecture, we will look at these models from a different perspective,i.e., one in which they themselves are the subject of scientificinquiry.
The abstract of the Colloquium can be found below this news item. The lecture will be held in English.
Before the Colloquium, from 11 AM, please join us for informal discussions over coffee and cakes in the lobby outside room 0.06.
We invite you to these events.
With best regards,
Barbara Badełek
Jan Chwedeńczuk
Jan Suffczyński
The dates for the Colloquia in the current semester are:
November 24, 2025
December 1, 2025
January 19, 2026
Abstract:
"Making Sense of Modern AI"
prof. Michał Kosinski, Stanford University, USA
Large Language Models (LLMs) trained to predict the next word in asentence have surprised their creators by displaying emergentproperties, ranging from a proclivity for biases to an ability to writecomputer code and solve mathematical tasks. This talk discusses theresults of several studies evaluating LLMs' performance on taskstypically used to study psychological processes in humans. Findingsindicate that as LLMs increase in size and linguistic ability, they cannavigate false-belief scenarios, sidestep semantic illusions, and tacklecognitive reflection tasks. This talk will argue the intriguingpossibility that LLMs do not merely model language but alsopsychological processes echoed in human language..
