I think about intelligence—what it is, how it works, and how to build it safely. Co-founded OpenAI, now building Safe Superintelligence Inc. The future is going to be very interesting.
Links
Education
PhD in Computer Science, Machine Learning
Advisor: Geoffrey Hinton. Thesis: Training Recurrent Neural Networks.
MSc in Computer Science
BSc in Mathematics
Early studies
Started university coursework as a teenager in Jerusalem.
Experience
Palo Alto & Tel Aviv
One goal: safe superintelligence. No distractions, no products, just the most important technical problem of our time.
San Francisco
Built the research team from the beginning. Led work on GPT, scaling laws, and co-led the Superalignment initiative.
Mountain View
Created sequence-to-sequence learning, worked on TensorFlow, and contributed to AlphaGo.
Stanford
Brief postdoc with Andrew Ng before joining DNNResearch.
Projects
With Alex Krizhevsky and Geoff Hinton. Won ImageNet by a landslide and changed everything. Deep learning went from academic curiosity to the future of AI overnight.
The architecture behind neural machine translation. Showed that you could map arbitrary sequences to sequences with deep nets. Later evolved into transformers and LLMs.
Bigger models trained on more data get predictably better. This simple insight drove the creation of GPT-2, GPT-3, GPT-4, and the modern AI boom.
How do you ensure AI systems smarter than humans remain aligned with human values? The most important unsolved problem in the field.