Các nhà lãnh đạo toàn cầu tranh đấu để quy định tương lai của trí tuệ nhân tạo

There is no doubt that the pace of AI development has accelerated over the last year. Due to rapid advances in technology, the idea that AI could one day be smarter than people has moved from science fiction to plausible near-term reality.

Geoffrey Hinton, a Turing Award winner, concluded in May that the time when AI could be smarter than people was not 50 to 60 years as he had initially thought — but possibly by 2028. Additionally, DeepMind co-founder Shane Legg said recently that he thinks there is a 50-50 chance of achieving artificial general intelligence (AGI) by 2028. (AGI refers to the point when AI systems possess general cognitive abilities and can perform intellectual tasks at the level of humans or beyond, rather than being narrowly focused on accomplishing specific functions, as has been the case so far.)

This near-term possibility has prompted robust — and at times heated — debates about AI, specifically the ethical implications and regulatory future. These debates have moved from academic circles to the forefront of global policy, prompting governments, industry leaders and concerned citizens to grapple with questions that may shape the future of humanity.

These debates have taken a large step forward with several significant regulatory announcements, although considerable ambiguity remains.

The debate over AI’s existential risks

There is hardly universal agreement on any predictions about AI, other than the likelihood that there could be great changes ahead. Nevertheless, the debates have prompted speculation about how — and the extent to which — AI developments might go awry.

For example, OpenAI CEO Sam Altman expressed his views bluntly during a Congressional hearing in May about the dangers that AI might cause. “I think if this technology goes wrong, it can go quite wrong. And we want to be vocal about that. We want to work with the government to prevent that from happening.”

Altman was not alone in this view. “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war,” read a single-sentence statement released in late May by the nonprofit Center for AI Safety. It was signed by hundreds of people, including Altman and 38 members of Google’s DeepMind AI unit. This point of view was expressed at the peak of AI doomerism, when concerns about possible existential risks were most rampant.

It Is certainly reasonable to speculate on these issues as we move closer to 2028, and to ask how prepared we are for the potential risks. However, not everyone believes the risks are that high, at least not the more extreme existential risks that is motivating so much of the conversation about regulation.

Industry voices of skepticism and concern

Andrew Ng, the former head of Google Brain, is one who takes exception to the doomsday scenarios. He said recently that the “bad idea that AI could make us go extinct” was merging with the “bad idea that a good way to make AI safer is to impose burdensome licensing requirements” on the AI industry.

In Ng’s view, this is a way for big tech to create regulatory capture to ensure that open source alternatives cannot compete. Regulatory capture is a concept where a regulatory agency enacts policies that favor the industry at the expense of the broader public interest, in this case with regulations that are too onerous or expensive for smaller businesses to meet.

Meta’s chief AI scientist Yann LeCun — who, like Hinton is a winner of the Turing Award –– went a step further last weekend. Posting on X, formerly known as Twitter, he claimed that Altman, Anthropic CEO Dario Amodei and Google DeepMind CEO Demis Hassabis are all engaging in “massive corporate lobbying” by promoting doomsday AI scenarios that are “preposterous.”

The net effect of this lobbying, he contended, would be regulations that effectively limit open-source AI projects due to the high costs of meeting regulations, effectively leaving only “a small number of companies (that) will control AI.”

The regulatory push

Nevertheless, the march to regulation has been speeding up. In July, the White House announced a voluntary commitment from OpenAI and other leading AI developers — including Anthropic, Alphabet, Meta and Microsoft — who pledged to create ways to test their tools for security before public release. Additional companies joined this commitment in September, bringing the total to 15 firms.

U.S. government stance

The White House this week issued a sweeping Executive Order on “Safe, Secure, and Trustworthy Artificial Intelligence,” aiming for a balanced approach between unfettered development and stringent oversight.

According to Wired, the order is designed to both promote broader use of AI and keep commercial AI on a tighter leash, with dozens of directives for federal agencies to complete within the next year. These directives cover a range of topics, from national security and immigration to housing and healthcare, and impose new requirements for AI companies to share safety test results with the federal government.

Kevin Roose, a technology reporter for the New York Times, noted that the order seems to have a little bit for everyone, encapsulating the White House’s attempt to walk a middle path in AI governance. Consulting firm EY has provided an extensive analysis.

While not having the permanence of legislation — the next president can simply reverse it, if they like — this is a strategic ploy to put the U.S. view at the center of the high-stakes global race to influence the future of AI governance. According to President Biden, the Executive Order “is the most significant action any government anywhere in the world has ever taken on AI safety, security and trust.”

Ryan Heath at Axios commented that the “approach is more carrot than stick, but it could be enough to move the U.S. ahead of overseas rivals in the race to regulate AI.” Writing in his Platformer newsletter, Casey Newton applauded the administration. They have “developed enough expertise at the federal level (to) write a wide-ranging but nuanced executive order that should mitigate at least some harms while still leaving room for exploration and entrepreneurship.” 

The ‘World Cup’ of AI policy

It is not only the U.S. taking steps to shape the future of AI. The Center for AI and Digital Policy said recently that last week was the “World Cup” of AI policy. Besides the U.S., the G7 also announced a set of 11 non-binding AI principles, calling on “organizations developing advanced AI systems to commit to the application of the International Code of Conduct.”

Like the U.S. order, the G7 code is designed to foster “safe, secure, and trustworthy AI systems.” As noted by VentureBeat, however, “different jurisdictions may take their own unique approaches to implementing these guiding principles.”

In the grand finale last week, The U.K. AI Safety Summit brought together governments, research experts, civil society groups and leading AI companies from around the world to discuss the risks of AI and how they can be mitigated. The Summit particularly focused on “frontier AI” models, the most advanced large language models (LLM) with capabilities that come close to or exceed human-level performance in multiple tasks, including those developed by Alphabet, Anthropic, OpenAI and several other companies.

As reported by The New York Times, an outcome from this conclave is the “The Bletchley Declaration,” signed by representatives from 28 countries, including the U.S. and China, which warned of the dangers posed by the most advanced frontier AI systems. Positioned by the UK government as a “world-first agreement” on managing what they see as the riskiest forms of AI, the declaration adds: “We resolve to work together in an inclusive manner to ensure human-centric, trustworthy and responsible AI.”

However, the agreement did not set any specific policy goals. Nevertheless, David Meyer at Fortune assessed this as a “promising start” for international cooperation on a subject that only emerged as a serious issue in the last year.

Balancing innovation and regulation

As we approach the horizon outlined by experts like Geoffrey Hinton and Shane Legg, it is evident that the stakes in AI development are rising. From the White House to the G7, the EU, United Nations, China and the UK, regulatory frameworks have emerged as a top priority. These early efforts aim to mitigate risks while fostering innovation, although questions around their effectiveness and impartiality in actual implementation remain.

What is abundantly clear is that AI is an issue of global import. The next few years will be crucial in navigating the complexities of this duality: Balancing the promise of life-altering positive innovations such as more effective medical treatments and combating climate change against the imperative for ethical and societal safeguards. Along with governments, business and academia, grassroots activism and citizen involvement are increasingly becoming vital forces in shaping AI’s future.

It’s a collective challenge that will shape not just the technology industry but potentially the future course of humanity.