Pushmeet Kohli
All peoplePushmeet Kohli, full AI read
Pushmeet Kohli, VP of Research (AI for Science & Safety), Google DeepMind (United Kingdom), ranks #387/500 on the Power & Influence Index (45.3). Greatest lever: AI power (64.0, Strong). Entrenchment 64/100; net worth $0bn.
Dimension read
| Dimension | Value | Standing | What a high vs low value means, and where Pushmeet Kohli sits |
|---|---|---|---|
| PII Power & Influence (PII) | 45.3 | Moderate · #385/635 | Mid-pack. High would mean near the apex of civilizational influence; low would mean lower relative influence within this elite set. ▲ high: near the apex of civilizational influence · ▼ low: lower relative influence within this elite set |
| Economic power Economic power | 48.0 | Developing · #470/635 | Low here, limited economic control. ▲ high: controls vast wealth and companies · ▼ low: limited economic control |
| Political power Political power | 30.0 | Developing · #534/635 | Low here, little formal political power. ▲ high: commands states, law or policy · ▼ low: little formal political power |
| AI power AI power | 64.0 | Strong · #174/635 | High here, shapes or controls the trajectory of AI. ▲ high: shapes or controls the trajectory of AI · ▼ low: limited sway over AI's direction |
| Platform reach Platform reach | 48.0 | Moderate · #367/635 | Mid-pack. High would mean commands a vast audience or network; low would mean limited direct reach. ▲ high: commands a vast audience or network · ▼ low: limited direct reach |
| Institutional control Institutional control | 62.0 | Moderate · #397/635 | Mid-pack. High would mean controls pivotal institutions and capital; low would mean limited institutional control. ▲ high: controls pivotal institutions and capital · ▼ low: limited institutional control |
| Entrenchment / tenure Entrenchment / tenure | 64.0 | Moderate · #296/635 | Mid-pack. High would mean power is locked in (indefinite or controlling stake); low would mean power is contingent, term-limited or contestable. ▲ high: power is locked in (indefinite or controlling stake) · ▼ low: power is contingent, term-limited or contestable |
Strengths
- AI power (64.0, Strong), shapes or controls the trajectory of AI.
Risk factors
- Significant but balanced influence with no single dominant lever.
AI worldview
Ideas & positions
Pushmeet Kohli emphasizes the importance of AI safety and robustness, particularly in the context of scientific applications such as protein folding with AlphaFold. He advocates for rigorous testing and validation of AI systems to ensure they are reliable and safe. Kohli has also been involved in research on adversarial attacks and defenses, aiming to make AI systems more resilient. While he does not explicitly take a strong stance on existential risk, his work suggests a focus on mitigating specific risks associated with AI deployment. He has not publicly endorsed a particular position on open vs closed models but has emphasized the need for collaboration and transparency in AI research.
What shapes the view
Kohli's background in computer science and machine learning, combined with his leadership roles at DeepMind, has shaped his focus on technical robustness and safety. His work on AlphaFold highlights his interest in applying AI to solve complex scientific problems. The emphasis on safety and reliability likely stems from the practical challenges and potential risks associated with deploying AI in critical domains such as healthcare and scientific research.
The AI-powered future they see
Kohli envisions a future where AI significantly advances scientific discovery and improves human health through applications like AlphaFold. He promotes the idea that AI can accelerate progress in fields such as drug discovery and materials science. However, he also warns about the need to address technical challenges and ensure that AI systems are trustworthy and robust.