Cyber Scene - AI Regulation From the Top: Back to Bletchley Basics
By krahal
Despite focusing on the unimaginable future of cyber, the world seems to have a whiff of retro-mode these days. In 1938-39, international concerns increased focus on the rise of Hitler. The impact of the coming World War II launched tech collaboration in London's shadow, with help from the Poles and Americans as things developed, to combat the seizure of Eastern Europe, and more, by Naziism. Simultaneously, across the pond, an American and a Canadian-born high school duo in Cleveland, Ohio, created Superman, who made his breakthrough to DC Comics in 1938. Since then, Superman has had uncountable spinoffs to address new threats. So too are 28 countries which gathered at Bletchley Park in early November. At the same time, US President Biden issued an Executive Order, "On Safe, Secure and Trustworthy Artificial Intelligence," both profiting from and controlling the two faces of AI.
The UK hosted the first-ever AI Safety Summit in November, drawing the EU, US (Vice President Harris attended), representatives from Africa, the Middle East, Asia, Australia, the UN Secretary- General Antonio Guterres, King Charles III (virtually), Elon Musk, the Chinese Vice Minister of Science and Technology, and scores of organizations. The "Godfather of AI," Yoshua Bengio, likely familiar with this readership, delivered the "State of the Science" report. South Korea and France offered sequentially to host a virtual mini-summit (ROK) and an in-person full summit (FR) over the next year. The objective is to "understand and collectively manage potential risks through a new joint global effort." Both nations and leading AI companies have agreed to support the safety testing of emerging types of AI.
Like Turing's view of the future, UK's Prime Minister Rishi Sunak believes that through the AI Safety Summit organization, "…we can secure the rapid international action we need for the safe and responsible development of AI around the world." Why is this of interest to Cyber Scene readership? The AI Safety Summit explains:
"All actors have a role to play in ensuring the safety of AI: nations, international fora and other initiatives, companies, civil society and academia will need to work together. Noting the importance of inclusive AI and bridging the digital divide, we reaffirm that international collaboration should endeavour (sic) to engage and involve a broad range of partners as appropriate, and welcome development-orientated approaches and policies that could help developing countries strengthen AI capacity building and leverage the enabling role of AI to support sustainable growth and address the development gap."
Back in the US, President Biden's Executive Order released on 30 October is principally focused on defense in the future, to "…ensure that America leads the way in seizing the promise and managing the risks of artificial intelligence." It was built upon the Biden White House's efforts to lead to voluntary commitments from 15 leading companies to "…drive safe, secure, and trustworthy development of AI." It requires the responsible governmental entities to implement the following:
- New standards for AI safety and security including 4 serious governmental implementations, largely defensive, for Americans
- Protection of Americans' privacy via bipartisan data privacy legislation and implementations
- Advancing Equity and Civil Rights through keeping AI algorithms not used for discrimination, but using best practices with help from DOJ (Department of Justice) and Federal civil rights offices
- Standing up for consumers, patients and students regarding healthcare and education
- Supporting workers to avoid surveillance, bias, and job displacement
- Promoting innovation and competition for AI startups, inter alia
- Advancing American leadership abroad
- Ensuring that governmental use of AI is responsible and effective as it accelerates the "rapid hiring of AI professionals" led by four organizations
In closing, it notes that it already is working with countries on all six continents and cites its $30 billion Bipartisan Infrastructure Law and Inflation Reduction Act which will bolster the US electrical grid underpinning AI.
Returning briefly to the Chinese presence at the AI Safety Summit, it is of serious import that the California Xi Jinping-Biden meeting—a landmark in itself—after the AI Summit and White House Executive Order, has taken place. Given the last few years of conflict (tech trade issues impacting cyber directly, intrusive attacks to the US, the famed balloon, Sino-Russian coziness, etc.), agreement for the two leaders to meet is extraordinary. The Economist pursues this in "Why Xi Jinping sounds friendlier to America." The author (nom de plume: "Chaguan," meaning "hot" Chinese teahouses) had compared Sino-US relations to the two remaining characters in Jean-Paul Sartre's “No Exit” where those who hate each other are condemned to be locked up together for eternity. However, surprise: the California meeting is now seen as a bilateral inflection point… in a good way.
Chaguan cites this hugely significant change: "In California he (Xi) came close to conceding, for the first time, that China is engaged in an economic, technological and geopolitical contest with America, and has an obligation to agree on a set of rules and guardrails that might prevent that competition from veering into disaster." This is not a misreading of the meeting, as the official Chinese readouts he cites discuss the Chinese-US relationship as "co-operating in areas of shared interest, and responsibly managing competitive aspects of the relationship." The Economist views this, although seemingly arcane, as really "quite a concession." Remember that the China White Paper 2050 aspired to "make a difference"—a very subtle understatement.
Regulation of AI is not likely to bond the US and China closely anytime soon. The world, as viewed in the above-mentioned summit and executive order, is working it hard at it. The Economist, Of Evils and Evals, in late October still believed that learning how to regulate around the globe is quite a challenge. While it is possible that these global efforts at regulation could succeed, an AI Safety Summit organizer noted that despite the commonality of importance in AI regulation, vice climate change and other "big policy debates," "…we still don't know what the right answer is." The EU is working on an end-of-year AI Act; the White House is already moving forward; the G7 is drafting an AI code of conduct this year; and China, on 18 October, unveiled a "Global AI Governance Initiative."
The Economist cites a sort of "one-upmanship" (your author's verbiage) competition to not be outdone by the Chinese, cast as "the Beijing effect" What is identified as a greater surprise is that the tech industry is no longer fighting regulation, but rather lobbying for it. The proviso, however, is that "…it must be narrow and target only extreme risks." The SEC (Securities and Exchange Commission) chairman, Gary Gensler, is concerned that an "AI-engineered financial crisis was nearly unavoidable without swift intervention."
In "AI Is Already at War: How Artificial Intelligence Will Transform the Military," Foreign Affairs' Michèle Flournoy, (former Undersecretary of Defense for Policy) notes that slowing down to refit, regarding AI, is impossible. Some of the reasons are AI progress in military applications, the ubiquitous nature of AI's mathematical foundations, the wide proliferation of AI models, and AI resource and development power, based on human creativity and the commercial driver. Her bottom line is that the Pentagon needs to speed up, not slow down to reset. The Department of Defense (DoD) has established a Chief Digital and Artificial Intelligence Office to deal with this.
Ms. Flournoy notes that Congress is aware of the situation, but has not yet acted, and the Department of Defense (DoD) is still working on it. Under the rubric "High Risk, High Reward," she notes that the Pentagon, for its part, is heading in the right direction, but the Defense Advanced Research Projects Agency (DARPA)—the entity that DID create the internet—needs to be more involved. The Senate and House are trying to educate themselves on AI and attendant technologies and understand that regulation at some level is essential. But the details remain to be resolved.
Time will tell if the November sessions not only draw various approaches together, but also do so fast enough. The tech world sets a high standard, but the unification of intent from various sectors across the globe makes AI regulation a good challenge. Perhaps the coming together of disparate parts creates the superpower to regulate AI worldwide.
To see previous articles, please visit the Cyber Scene Archive.