Main content

AI Without Limits?

AI is beginning to do something remarkable: teach itself. Are we creating systems that could eventually improve themselves faster than humans can understand or control?

Artificial intelligence is beginning to do something remarkable: teach itself. What once sounded like science fiction is now being taken seriously by some of the world's leading AI researchers. Could self-improving AI spark a new golden age of scientific discovery, helping humanity cure disease, grow replacement organs, tackle pollution and even uncover laws of nature that have so far remained beyond our reach? Or are we creating systems that could eventually improve themselves faster than humans can understand or control?

Joining AI Decoded is Richard Socher, co-founder and CEO of Recursive and author of The Eureka Machine. He explains his vision for a new generation of AI systems capable of accelerating scientific breakthroughs by learning from vast amounts of data, designing experiments and continually improving their own capabilities. Could these 'Eureka Machines' become the greatest tool humanity has ever created?

Putting that vision to the test are AI Decoded co-hosts Priya Lakhani, founder of Century Tech, and Stephanie Hare, technology researcher and author of Technology Is Not Neutral. Together they challenge some of the biggest claims in AI, from whether machines can generate genuinely original ideas to whether self-improving systems can ever be made safe.

The discussion also explores growing concerns raised by historian and Reith lecturer Rutger Bregman, who argues that advanced AI may be progressing much faster than many realise. Should governments ban AI systems that build better AI? Can humans remain in control if machines start improving themselves? And are fears of runaway superintelligence justified, or overblown?

11 months left to watch

26 minutes

Broadcasts