Pushmeet Kohli, Vice President of Research, Google DeepMind
Pushmeet Kohli is the Head of AI for Science at DeepMind, leading efforts including AlphaFold, a state-of-the-art AI system for predicting the 3D structure of proteins. Prior to this, he was the director of research at the Cognition group at Microsoft Research. Pushmeet's research centres on using AI to solve impactful real world science-related problems. He is particularly interested in the use of machine learning techniques to enable and accelerate life sciences research, and the use of this knowledge for understanding and intervening in disease. Pushmeet also actively leads research on new techniques to ensure that AI systems are safe, reliable and trustworthy.
At DeepMind, Pushmeet Kohli leads the research team as a principal scientist. Pushmeet was the director of research in Microsoft’s Cognition group before to joining DeepMind. at his ten years with Microsoft, Pushmeet held a variety of positions and responsibilities, including technical adviser to Microsoft Chief Research Officer Rick Rashid, working at Microsoft labs in Seattle, Cambridge, and Bangalore. Pushmeet writes in the domains of machine learning, computer vision, information retrieval, and game theory. His research focuses on computational sciences and intelligent systems. His current areas of interest in study are interpretable and verifiable knowledge representations from deep models, probabilistic programming, and 3D reconstruction and rendering.
Pushmeet’s study has earned him several honors and trophies. His doctoral thesis, “Minimizing Dynamic and Higher Order Energy Functions using Graph Cuts,” was a runner-up for the British Computer Society’s “Distinguished Dissertation Award” and the recipient of the British Machine Vision Association’s “Sullivan Doctoral Thesis Award.” The conferences for Computer Vision, Machine Learning, Robotics and AI, Computer Graphics, and HCI have published papers by Pushmeet. Numerous stories on his study have also appeared in widely read publications including Forbes, Wired, BBC, New Scientist, and MIT Technology Review. Pushmeet is a participant in the Distinguished Speaker Program of the Association for Computing Machinery (ACM).
2023
An AI research team at Google’s DeepMind project has created a particular kind of AI system that may exhibit social learning skills. The team details how they created an artificial intelligence program that demonstrated its capacity to pick up new abilities in a simulated environment by imitating the behaviors of an implanted “expert” in their publication published in the journal Nature Communications.
The team’s initial task was to create GoalCycle3D, a virtual environment with uneven terrain, different barriers, and multicolored spheres. Next, artificial intelligence (AI) agents were included. Their goal was to navigate the virtual environment by dodging barriers and entering circles. All that was provided to the agents in terms of knowledge about the world they would live in were learning modules. They used reinforcement learning to learn how to go forward.
Demis Hassabis, CEO of Google DeepMind, discussed the future of AI and the progress made by the company. He highlighted the merger of Google Brain and DeepMind, the success of ChatGPT, the potential of non-generative AI, the future of AI chatbots, and the path towards Autonomous General Intelligence (AGI). Hassabis also highlighted the need for collaboration between multiple models, such as a language and learning model for communication and another for specific tasks, to drive advancements towards AGI.
Latest Social posts of Pushmeet Kohli
Update on the next gen AlphaFold model we have been developing. This model takes structure prediction beyond proteins to general biomolecules like nucleic acids (RNA/DNA), small molecule ligands, PTMs etc. opening us entirely new possibilities in accelerating biological research. https://t.co/UbGMl9VWz5
We at @GoogleDeepMind are excited to announce #GNoME – an AI tool that has discovered 2.2 million new materials, and helps to predict material stability. We're releasing 381K stable materials to help scientists pursue materials discovery breakthroughs. https://t.co/nMeo3qGOTN
