Demis Hassabis: AGI by 2030 Possible, Model Distillation Key to AI Efficiency

Demis Hassabis, CEO of Google DeepMind and recipient of the 2024 Nobel Prize in Chemistry, predicts Artificial General Intelligence (AGI) could be achieved by 2030. He stated that active problem-solving systems are crucial for AGI development, with possibly one or two major breakthroughs still needed. Hassabis highlighted that current AI systems are inefficient due to brute-force processing.

Model distillation, a process that creates smaller, highly efficient AI models without sacrificing performance, is seen as a key strength by DeepMind. These smaller models offer significant cost and speed benefits, drastically increasing productivity, with engineers now capable of performing up to 1000 times more work than six months ago.

Ideas from past triumphs like AlphaGo and AlphaZero are expected to drive future AI advancements. However, a significant barrier to full task automation remains the lack of continual learning in AI systems, which prevents them from adapting to specific contexts effectively.