- AI leaders disagree on how long it will take to realize AGI, a machine that can reason like humans.
- Google DeepMind CEO Demis Hassabis says “AI agents” are the next phase of AI after chatbots.
- These systems will be able to plan, act, reason, and use tools, among other traits, he said.
Experts are divided on what it’ll take to achieve artificial general intelligence — a still hypothetical form of robot intelligence that mimics human abilities.
According to Demis Hassabis, the CEO of Google DeepMind and a recently minted Nobel laureate, there isn’t any secret formula to get there.
At The Times Tech Summit earlier this month, Hassabis said the next iteration of AI after chatbots like ChatGPT — which passively answers questions, summarizes text, and does research — will be agent-based systems with the following traits and skills:
- Planning. These systems can think ahead, plan a trip, or book tickets.
- Acting. They need to be able to take action in the real world.
- Reasoning. They should think through problems. Hassabis pointed to AlphaGo, DeepMind’s AI that first defeated humans in Go, demonstrating reasoning skills in the game’s defined domain. The next phase of agent-based systems will apply these skills to real-world contexts.
- Better Memory. They should remember the details they’ve been told.
- Better Personalization. They should understand a user’s preferences, likes, and dislikes.
- Using tools. They can use hardware like robots in the physical world, software like a calculator, or even other AI systems.
Hassabis said at the summit, however, that he thinks the full realization of an artificial general intelligence that can reason as well as humans is still a decade away.