Martin Schrimpf sees the true potential of artificial intelligence not merely in its capabilities but in its power to shed light on human intelligence. He is on a mission to create a ‘digital twin’ of the human brain through the use of artificial neural networks—AI models that draw inspiration from the way neurons interact.
Although the goal may seem overly ambitious, Schrimpf’s methodology is quite straightforward. His team conducts tests related to language and vision on various individuals, then compares the resulting behaviors and brain activity with the outputs generated by AI models designed for similar tasks. This comparison enables them to refine their models, gradually making the AI more humanlike.
To enhance the effectiveness of this process, which thrives on abundant data and a variety of models, Schrimpf developed an open-source platform known as Brain-Score. This platform houses nearly one hundred human neural and behavioral datasets. Since its inception in 2017, while Schrimpf was still a graduate student, thousands of AI models have been evaluated against human data.
Initially, Schrimpf intended to pursue a career in the tech sector. However, after co-founding two software startups during his early academic years, he found himself lacking fulfillment. He reflected, “I thought I could ask neuroscientists how the brain works, and that would help me build better AI. But I realized there’s a huge opportunity in the opposite direction: prototyping ideas in silico [on a computer] and using AI models to explain the brain.”
He transitioned from Germany to the United States to pursue a doctorate in brain and cognitive sciences at the Massachusetts Institute of Technology. In 2023, he returned to Europe to establish the NeuroAI Lab at the Swiss Federal Institute of Technology Lausanne.