虎嗅

Dialogue with Demis Hassabis: AI will become the ultimate tool of science

原文:对话德米斯·哈萨比斯:AI将成为科学的终极工具

Core Summary

Demis Hassabis, co-founder of DeepMind (the company behind AlphaGo and AlphaFold), shares his journey from an AI engineer to a scientist. He believes that AI is humanity’s ultimate tool for solving complex scientific problems and outlines his vision for the responsible use of this technology. He discusses what General Artificial Intelligence (AGI) truly means, how it is transforming fields such as biology and physics, and his theory that information could be the fundamental building block of the universe. His goal is to use AI to cure diseases, combat climate change, and unravel the mysteries of reality, all while prioritizing safety and ethics.

1. From Game AI to Life-Saving Science: How AlphaGo & AlphaFold Changed the Game

Hassabis began by developing AI for games, such as AlphaGo, which defeated the world champion Lee Sedol in 2016. However, he quickly realized that AI could do more than just win games; it could also solve real-world problems. For example, AlphaFold revolutionized protein structure prediction. For decades, biologists struggled to determine how proteins fold into 3D shapes, but AlphaFold learned from thousands of known protein structures and can now predict these shapes with high accuracy in seconds. This breakthrough has significantly accelerated drug discovery, enabling the development of treatments for diseases like cancer and malaria.

2. What Is AGI, and Why Aren’t We There Yet?

AGI refers to a system with human-like cognitive abilities, capable of inventing new theories, writing poems, or solving complex problems. Currently, AI is “narrowly focused”—it excels in specific tasks but struggles with other areas. The main challenges to achieving AGI include the lack of creativity, consistency, and adaptability.

3. AI as the Ultimate Tool for Big Scientific Problems

Hassabis identifies three key criteria for selecting problems suitable for AI: a vast search space, sufficient data, and a clear goal for measuring success. Examples of AI applications in science include predicting climate patterns, optimizing energy systems, and studying quantum gravity. He believes that AI can become a powerful ally for scientists in solving major scientific challenges.

4. The Philosophical Twist: Is Information the Building Block of Reality?

Hassabis proposes that information might be the fundamental component of the universe, as everything in nature follows stable patterns. If this is true, understanding information could help us unravel the mysteries of reality. He uses AI to test this idea by simulating complex systems, suggesting that it supports the notion that information is the foundation of all existence.

5. Responsible AI: Using It for Good (and Safely)

Hassabis emphasizes the importance of using AI responsibly, including following the scientific method, being honest about its limitations, and considering long-term impacts. He advocates using AI to address global issues, such as curing diseases and combating climate change, to benefit all people.

In conclusion, Demis Hassabis believes that AI has the potential to transform science and solve some of humanity’s most pressing problems. However, we must use this technology wisely and responsibly to ensure its benefits are widespread and safe for everyone.