Companies publishing AI-generated content may need to label their content with a “watermark” in the future to reduce the risk of deceiving the public, Rishi Sunak said.
Governments are considering a number of regulations that may apply to artificial intelligence in the next few years.
The prime minister told the Commons Liaison Committee that he believed AI could boost GDP by 10% over a decade and dramatically increase the productivity of the economy.
But he also warned of serious risks, including social upheaval, the spread of misinformation and even an existential threat to humanity.
Sunak plans to host an international leaders’ summit in the UK later this year, where he hopes to coordinate plans for a multilateral regulatory regime, which he believes is being used to monitor the world’s nuclear agenda. He also mentioned the creation of a similar institution. Industry was not imminent.
Asked by lawmakers what kind of regulation he was considering and whether new legislation needed to be passed, the prime minister stressed the importance of transparency. For example, to ensure that high-tech companies register their research and explain the company’s training methods. computer model.
“What we have to do, and I think we can probably do a lot without the law, is sit down and think about what safety features and guardrails we want to put in place.
“I think I understand what some of them are: watermarking, pre-release information sharing, reporting and licensing for large training runs, pre-training or pre-deployment. disclosure and transparency, and I think that’s the category of things we need to spend our time on.”
The idea of a forced watermark (a sort of indicator that some piece of content was generated by AI) has already been championed by Joe Biden.
Sunak said AI could soon make it possible to treat “incurable diseases like cancer and dementia,” and that the economy could grow much faster than in the past, resulting in “increase in productivity.” predicted the possibility of a J-curve.
But he warned that the dangers included job losses as a result of the “massive societal risks posed by technology.” Use of AI programs as “tools to generate false information”, including deepfakes. Fraudsters deploying AI in the physical world. And because “at the moment there is a lack of understanding of what the potential of these models is,” it could be a matter of human survival.
In recent months, the UK government has sought to take the lead in global AI regulation, positioning itself as a mediator between the EU’s restrictive approach and the US’s more laissez-faire stance.
