Simon Kornblith
About
Originally trained as an experimental neuroscientist, Simon Kornblith now sees understanding artificial neural networks as a necessary foundation for understanding the brain. Artificial neural networks are free of the challenges of biological experimentation, yet explaining how these systems can give rise to intelligent behavior has remained challenging. Simon’s research has focused on developing and validating analytical tools for understanding how artificial neural networks represent information and applying these tools to build better neural networks. Given the rapidly improving coding and research abilities of neural network-based large language models, Simon now sees these models as analytical tools in themselves. He currently leads a research team at Anthropic that develops novel architectures and algorithms to improve Anthropic’s large language model Claude.
Relevant Publications
- Muttenthaler, L., Linhardt, L., Dippel, J., Vandermeulen, R. A., Hermann, K., Lampinen, A., & Kornblith, S. (2023). Improving neural network representations using human similarity judgments. Neural Information Processing Systems.
- Chen, T., Kornblith, S., Norouzi, M., and Hinton, G. (2020). A simple framework for contrastive learning of visual representations. International Conference on Machine Learning.
- Kornblith, S., Norouzi, M., Lee, H., and Hinton, G. (2019). Similarity of neural network representations revisited. International Conference on Machine Learning.