PauseAI US Newsletter #2 - February 2025

The Paris ‘AI Action Summit’ (on February 10 - 11) abruptly changed focus away from AI safety.

• Concerns about the Paris summit began early, when the name of the summit was announced. While prior summits were named AI Safety Summits, the Paris summit was renamed the AI Action Summit. Soon after, a leak of the AI Action Summit Statement was released, which was criticized for decentralizing AI safety. “The view within the community of people concerned about safety is that it's been downgraded,” said Stuart Russell– professor at UC Berkeley and founder of the Center for Human-Compatible Artificial Intelligence– in an interview with TIME.

• In response, PauseAI Global launched a petition calling for the Paris AI Action Summit to return its focus to safety. PauseAI Global also organized a series of worldwide protests calling for an international treaty to pause human-level AI development.

• The summit’s final official statement “highlighted the importance of reinforcing the diversity of the AI ecosystem.” Safety was briefly mentioned, but only after the first priority of “promoting AI accessibility to reduce digital divides.” The US and UK both refused to sign the statement. Per an unnamed UK government spokesperson, “we felt the declaration didn’t provide enough practical clarity on global governance, nor sufficiently address harder questions around national security and the challenge AI poses to it.”

• The US, meanwhile, criticized “excessive regulation” of AI. Vice President JD Vance stated that “AI will have countless revolutionary applications in economic innovation, job creation, national security, health care, free expression, and beyond. And to restrict its development now [...] would mean paralyzing one of the most promising technologies we have seen in generations.” Further, Vance promised that “we’ll make every effort to encourage pro-growth AI policies.” Vance did not distinguish between safe, specialized AI– which could indeed bring such benefits in science, medicine, and economic growth– and superhuman AI– which could backfire horribly and cause our extinction.

Photo: world leaders pose at the AI Action Summit after a productive few days of doing nothing to keep humanity safe.

The Trump Administration opens the floodgate to AI acceleration.

• On January 23, Donald Trump signed an Executive Order revoking Biden’s previous Executive order on AI, citing the goal of “REMOVING BARRIERS TO AMERICAN AI INNOVATION.” The order “calls for departments and agencies to revise or rescind all policies, directives, regulations, orders, and other actions taken under the Biden AI order that are inconsistent with enhancing America’s leadership in AI.”

• Trump has had a slew of meetings with AI tech company leaders, including Nvidia CEO Jensen Huang, Meta CEO Mark Zuckerberg, and OpenAI CEO Sam Altman. Altman and Zuckerberg donated $1 million each to Trump’s inauguration fund, and Zuckerberg was front-and-center at Trump’s inauguration.

• Trump also declared his support for the Stargate Project, a $500 billion OpenAI infrastructure project, announcing he would “help a lot through emergency declarations, because we have an emergency. We have to get this stuff built.” Since this statement, Trump has declared a National Energy Emergency, which grants emergency approvals to “facilitate the identification, leasing, siting, production, transportation, refining, and generation of domestic energy resources, including, but not limited to, on Federal lands.” This emergency declaration is a first step in securing energy resources that colossal AI infrastructure projects could utilize.

• Trump appeared to turn up the heat on the global AI arms race in his statement supporting the Stargate Project, saying, “What we want to do is we want to keep it in this country. China is a competitor and others are competitors.”

Deepseek’s success has sparked concerns of an AI arms race.

• In January, the Chinese AI company Deepseek released its R1 model, which boasted performance on par with OpenAI’s recent o1 model, but at a fraction of the cost.

• US companies responded quickly, with Meta reportedly assembling “war rooms” of engineers to attempt to figure out how Deepseek achieved this feat. OpenAI– who has faced lawsuits for use of copyrighted training materials– accused Deepseek of extracting data from OpenAI’s models in order to train Deepseek’s models in a process known as distillation.

• Deepseek’s success has been called a ‘Sputnik Moment,’ a sentiment that has been echoed in the media. Sputnik 1 was a Soviet satellite, the launching of which marked the beginning of the space race between the US and the Soviet Union. The main flaw in this comparison is that the space race got us to the Moon, whereas an AI arms race could threaten our existence on Earth. Any race to superhuman AI would incentivize reckless development of AI capabilities, before we understood how to keep these AI models in control – it would be a race without winners.

Contrary to what some of Trump’s inner circle– and, indeed, some Democrats– might think, there is no benefit in “beating” China to smarter-than-human AI. Their reasoning is straightforward enough – they regard superhuman AI as yet another tool (or, indeed a weapon) to win geopolitical advantage over our rivals. We are used to thinking in these terms. But when it comes to superhuman AI, this arms-race mentality breaks down.

AI-race advocates assume we can control smarter-than-human AI and wield it in our favor, to secure American dominance over our adversaries. This is wishful thinking at best. Thousands of AI researchers have warned that we don’t know how to control superhuman AI, and that these systems could go rogue and cause human extinction. In other words, we don’t even know how to prevent these systems from killing us, much less install specific American values into their hardware (of course, even if we could install such values, the question of whose values to emulate is itself a thorny one).

The United States will gain no strategic advantage from being the first country to destroy the world. Superhuman AI is not a tool to be strategically deployed; it is a doomsday device, which could blow up in our faces the minute we turn it on. Superhuman AI will not care whether it uses American or Chinese technology to usurp humanity – the final outcome will be the same regardless.

Building superhuman AI, crossing your fingers, and expecting everything to go well for your country is like unleashing an army of alien scientists within your borders and giving them access to your military secrets without knowing their intentions. It is extreme myopia and folly, and it is ultimately self-defeating.

As it turns out, building something smarter than you and asking nicely if it will comply with your wishes is fiendishly complicated! Who would have guessed?

This is a race with no winners. The only reward for first place is that you get to die slightly faster.

The only way out is through an AI Treaty – an agreement, signed by the world’s superpowers, prohibiting the development of smarter-than-human AI, along with enforcement mechanisms to ensure this technology isn’t developed behind the scenes.

Not only would such a treaty save the human race, it is in our unique national interest to negotiate it. As the world’s leaders in AI development, the ball is in our court. If the United States negotiates an AI Treaty today, we can do so in terms favorable to our economy and security, preserving our edge in safe and specialized AI development while universally prohibiting the development of AI that can kill us. If we fail to act, we will find ourselves at the mercy of an uncertain future. Our adversaries might even catch up to us in AI development and force our hand, making us agree to provisions that restrict the kind of useful, specialized AI that would actually serve American interests. Better to negotiate when you’re ahead, than risk falling behind.

President Trump has branded himself as a deal-maker. He now has the opportunity to make the deal of the century. To protect American interests– and safeguard the future of the human race– the United States must initiate negotiations on an AI Treaty.

Are you unnerved by the unchecked advance of Artificial Intelligence and its increasing integration into our everyday life? Do you wish you could do something, but feel powerless to do so? Maybe you just don’t know where to start? Here are two ways you can get involved this week and make your humanity heard.

• Voice your opinion about Trump’s AI Action Plan
  
  The American Office of Science and Technology Policy wants to hear your input about Trump’s executive order on AI. Make your case through the federal register.

  Deadline: March 15th, 2025

  
  Always remember: Your voice matters.