Google Deepmind CEO Dismisses Meta’s AI Complaints as ‘Absurd’

Demis Hassabis, co-founder of DeepMind, responded to criticisms made by Meta Platforms Inc.’s chief AI scientist, who had accused him and others in the field of exaggerating the potential dangers of the technology. Hassabis dismissed these criticisms as absurd.

Hassabis responded earlier this week to a post by Meta’s Yann LeCun, who said some leaders were unnecessarily jumping into doomsday scenarios and providing ammunition to those who want to ban open source AI research and development. LeCun argued that this approach would put control of future AI systems in the hands of just a few companies, which would be a “disaster.”

“We want to make sure we get the benefits of the innovation and the promise that the technology clearly holds,” Hassabis said in an interview with Bloomberg Television from the UK’s AI Security Summit at Bletchley Park on Wednesday. “But we have to do it responsibly. I would say using the scientific method, trying to get as much anticipation out of the technology as possible. So we’re predicting in advance what the unintended consequences might be.”

The UK’s AI Security Summit has brought together leaders from the US, Asia and Europe to agree on how to protect against the risks of the most advanced “frontier” AI systems. Although Prime Minister Rishi Sunak had hoped to include some key leaders, the summit was able to issue a joint statement on Wednesday in which 28 nations agreed to work together on a common approach to artificial intelligence.

Hassabis said that in addition to efforts to address near-term pitfalls, such as preventing images and audio created by watermarking so deep fakes are not mistaken for genuine, an international agreement is needed on the technology’s long-term risks.

“Maybe they’re still a decade away, but you don’t want to be dealing with these types of things until something general AI happens,” he said. “We want to prepare well in advance.”