Brain research and large language models - a quantum leap?

The symposium explored the potential of large language models (LLMs), such as GPT-3, which have recently generated both excitement and concerns about their ability to rival human intelligence. While there are many similarities in the capabilities of LLMs and human intelligence, it is important to consider these features from a neurological perspective.

During the event, participants discussed the achievements and challenges in the realm of AI, comparing the capabilities of artificial systems to the human brain. They also examined the tools used to study the representation and potential improvements of LLMs. One interesting question raised during the symposium was whether these models have learned a representation of language similar to that of the human brain.

Furthermore, the symposium addressed the potential for LLMs to inspire progress in brain research. In this context, researchers explored whether neurological findings could be used to enhance the performance of current language models.

The symposium featured notable experts in computer science and neuroscience, including Alison Gopnik from Berkeley, Iryna Gurevych from TU Darmstadt, Uri Hasson from Princeton University, Melanie Mitchell from Santa Fe Institute, Björn Ommer from LMU Munich, Haim Sompolinsky from Hebrew University/Harvard University, and Mariya Toneva from the Max Planck Institute for Software Systems.

Additionally, a panel discussion was held as part of the symposium. The entire event was conducted in English.