Recent advances in artificial intelligence (AI) have spurred both enthusiasm about the capabilities of the latest large language models (LLMs) and warnings about their ability to match or even surpass human intelligence. But do comparisons to the human brain hold when addressed from the perspective of Neuroscience?  

This high-level symposium co-hosted by Leopoldina, the German National Academy of Sciences, and Max Planck Institute for Brain Research will bring together experts from computer science and neuroscience to discuss what has and has not been achieved with latest advances in AI. What are and how can we judge the capabilities of artificial systems compared to human intelligence? Which aspects of LLMs are similar and which ones are decisively dissimilar from the way the human brain works? By which tools can we inspect, measure and analyze the representation and capabilities of large language models? Have LLMs learned a representation of language that shows similarities to the one generated by our human brains? How and in which sense can LLMs inspire advances in brain science? Conversely, what aspects might be missing and could be used to improve current LLMs (and what does “improve” mean, in the first place)?

With these questions, the symposium aims at contributing to a scientific basis for the hopes and fears associated with AI that shall or shall not match human intelligence.  

Confirmed speakers include Alison Gopnik (Berkeley), Iryna Gurevych (Technical University of Darmstadt), Uri Hasson (Princeton University), Melanie Mitchell (Santa Fe Institute), Björn Ommer (Ludwig Maximilian University of Munich), Haim Sompolinsky (Hebrew University/Harvard University) and Mariya Toneva (Max Planck Institute for Software Systems).

A panel will take place on the evening of May 13.

The symposium will be held in English.