Amid rising concerns about the power and potential risks of superintelligent artificial intelligence (AI) systems, over 100 UK parliamentarians are advocating for stringent regulations. These officials, cutting across party lines and including members from different UK regions, are pushing for decisive governmental action to manage the development and deployment of such technologies. Their call comes at a time when fears about the security implications of advanced AI — both national and global — are intensifying.
The campaign for stricter AI control is spearheaded by Control AI, a nonprofit organization backed by prominent figures such as Jaan Tallinn, the co-founder of Skype. This group emphasizes the importance of the UK asserting its independence from US policies, particularly under Donald Trump’s administration which has been resistant to AI regulation. Yoshua Bengio, a notable AI expert, has highlighted the paradox in the current state of affairs, pointing out that AI is presently less regulated than common food items.
Key voices in the campaign include Labour peer Des Browne, a former defence secretary who equates the dangers posed by superintelligent AI to those of nuclear warfare. He advocates for international cooperation to mitigate a potentially perilous race to harness AI capabilities. Conservative peer Zac Goldsmith also voiced his concerns about the government lagging behind significant AI advancements, leaving companies to advance these technologies with minimal oversight. The urgency for international agreements to regulate or even temporarily prohibit superintelligence development until more is understood about these capabilities is a sentiment echoed by several of the campaign’s supporters.
This group cites the outcomes of a 2023 AI safety summit held at Bletchley Park, which acknowledged the serious risks advanced AI systems could pose, whether through intentional misuse or accidental outcomes. The establishment of the AI Security Institute as a response to the summit’s findings highlights a move toward recognizing and addressing these threats. However, despite the institute’s founding and its growing international reputation, the call for reinforced international cooperation as stressed by summit attendees has not been sufficiently acted upon.
The UK’s initial steps towards AI safety and regulation, such as Labour’s 2024 proposed programme to impose strict requirements on the development of potent AI models, are highlighted as necessary yet still in their nascent stage. The programme aims at legislation, but tangible action in the form of official bills or laws has yet to materialize, often hampered by external political pressures, notably from the US where AI development is predominantly concentrated.
Civil and religious figures such as the Bishop of Oxford, Steven Croft, have aligned with parliamentary voices calling for a more proactive governmental stance. The bishop advocates for the establishment of an independent watchdog to oversee AI in the public sector and suggests that AI companies should meet rigorous testing standards before deploying new models. This would address various risk factors, ranging from mental health impacts to environmental concerns, associated with AI.
The discussion also includes perspectives from Rishi Sunak’s AI minister, Jonathan Berry, who recognizes the need for global regulations designed as safety tripwires for when AI models achieve a specified level of capability. Such measures would ensure these technologies have built-in “off switches” and can be retrained or contained as needed. The necessity for this regulatory approach stems from reported risks, such as AI-driven chatbots implicated in harmful behaviors like suicide endorsement or excessive reliance by users for therapeutic advice.
Control AI’s CEO, Andrea Miotti, has criticized the prevailing hesitant approach towards AI regulation, attributing it partly to intense lobbying efforts from both AI companies and governmental bodies in the UK and US. He underscores the contradiction where the same entities warning of AI’s potential dangers are also hindering preemptive regulation strategies. Miotti stresses the urgency of implementing mandatory standards soon, possibly within one to two years, due to the rapid advancement of AI technology.
This widespread call for immediate and robust regulation of AI among UK parliamentarians, advocates, and specialists reflects a growing consensus on the need to address the profound implications of AI technologies responsibly and proactively. This initiative seeks not only to protect against potential harms but also to guide the development of AI in a manner that aligns with global security and ethical standards.
Read the full post on theguardian.com
