Summary
Deepmind co-founder Demis Hassabis believes that the capabilities of AI are currently overestimated. In the long term, however, the technology’s potential is still underestimated, especially with regard to artificial general intelligence.
Demis Hassabis, co-founder and CEO of Deepmind, sees the public perception of AI as a double-edged sword. On the one hand, the great interest shows that the technology is ready to have a positive impact on people’s everyday lives, says Hassabis in a Podcast.
On the other hand, the attention also brings a lot of noise into the field. For those like Hassabis who have been working on AI for many years, it is surreal to see the public slowly realizing how important AI will become.
Although he thinks current AI models are overrated, Hassabis is surprised by their ability to infer things about the real world, even though they only learn from language: “They live and learn in an abstract domain, so it’s quite surprising that they can infer some things about the real world from that,” says Hassabis.
Advertisement
THE DECODER Newsletter
The important AI news directly to your email inbox.
✓ 1x weekly
✓ free
✓ Can be cancelled at any time
Too much hype in the AI market in the short term, long-term effects are underestimated
Hassabis believes that the capabilities of AI are overestimated in the short term. Many people and startups claim that the technology can do things that it still cannot. There are crazy ideas and venture capitalists chasing these ideas even though the technology does not exist.
In the long term, however, the potential of AI is still underestimated, especially when it comes to artificial general intelligence (AGI) and the time after that. Hassabis believes it is possible that AGI will be achieved in the next decade. Deepmind was designed as a 20-year project, and this limit will be reached around 2030.
“I would say I would not be surprised if it [AGI] in the next decade,” says Hassabis. He sees enormous potential for humanity through AGI, for example in curing diseases, clean energy and solving the climate problem.
At the same time, however, the risks of human misuse and uncontrollable AI systems must also be contained. The Deepmind boss called for international cooperation in regulating AI, as the technology is difficult to contain within national borders.
In the final stages of AGI’s development, Hassabis hopes to create a scientific “Avengers” group of the world’s best AI researchers.
Recommendation
