Marc Warner, a member of the government’s AI Council and CEO of Faculty AI, has raised concerns about the need to potentially ban certain powerful artificial general intelligence (AGI) systems. According to Warner, AGI systems should have strong transparency, audit requirements, and enhanced safety technology. He emphasized the importance of making sensible decisions regarding AGI in the next six months to a year.
Warner’s comments follow a joint statement by the European Union and the United States, stressing the need for a voluntary code of practice for AI. The AI Council, an independent expert committee providing advice on AI to the government, is where Warner serves. Faculty AI, which Warner heads, is the technical partner of OpenAI, assisting customers in safely implementing AI products like ChatGPT.
While “narrow AI” systems used for specific tasks can be regulated similarly to existing technologies, AGI systems, which aim to exceed human intelligence across various domains, require different rules and raise more concerns. Warner highlighted the risks associated with creating AI systems that surpass human intelligence, emphasizing the need for caution and strong limits on computing power.
Warner acknowledged that the decision to ban algorithms of a certain complexity or compute power should be made by governments rather than technology companies. He argued that safety measures could provide the UK with a competitive advantage and stated that safety is crucial for deriving value from AI technology.
Some argue that concerns about AGI distract from existing technology-related problems such as bias in AI tools. However, Warner compared this perspective to questioning the safety of both cars and airplanes, emphasizing the need for safety in all aspects. While excessive regulation may raise concerns about stifling innovation and deterring investors, Warner believes that prioritizing safety can be advantageous for the UK.
Although the UK’s recent White Paper on regulating AI faced criticism for lacking a dedicated watchdog, Prime Minister Rishi Sunak has emphasized the necessity of establishing “guardrails” and positioning the UK in a leadership role. The United States and the European Union have also expressed the need for voluntary rules to be established swiftly. The EU’s Artificial Intelligence Act, currently undergoing legislative processes, is expected to regulate AI, but implementation is projected to take a few years. However, a draft voluntary code of conduct will be open for industry and public contributions in the coming weeks.