Senator-elect El-Sayed and Representative Ansari call for an AI pause as users on X debate its impact on treatments for disease and aging

Senator-elect El-Sayed and Representative Ansari call for an AI pause as users on X debate its impact on treatments for disease and aging

Democrats won the November 3 elections, taking both chambers of Congress. Physician Abdul El-Sayed won Michigan’s Senate seat. Two days before the vote, he told CBS that frontier AI research needed a “complete pause.” On November 4, Representative Yassamin Ansari wrote that an AI development pause and a ban on superintelligence (AI that outperforms humans at most tasks) should top the Democrats’ agenda on day one.

El-Sayed’s “AI Under Control” plan, dated October 13, calls for freezing the training and release of the most powerful AI systems until safety rules are in place, delaying new AI data centers, and permanently banning superintelligence. It proposes an agreement with China on a pause and a ban, with verification covering chips and data centers. The plan cites hacking involving AI agents and estimates that put the risk of AI causing human extinction within a decade as high as 10%.

Ansari is a cosponsor of H.R. 10538, introduced on September 24. Like Bernie Sanders’s Senate bill, it would establish a Department of AI, ban superintelligence, and mandate a pause for any system trained using 10²⁵ or more computational operations. Until the secretary determines that the department is staffed and regulations have been issued, such systems could not be trained or modified (except to shut them down or remove signs of superintelligence), and unreleased systems could not be launched. A company executive or an outside individual who recklessly violates these requirements could face up to 20 years in prison. According to analysts at Epoch AI, 10²⁵ operations is roughly the scale of GPT-4’s training, while Evo 2, a model trained on DNA and the largest biological AI model in their database, used less than a quarter of that amount.

The bill is in committee. Before Ansari’s post, AI writer Andrew Curran wrote that even after the Democratic victory, an AI freeze or a superintelligence ban like Sanders’s was unlikely to become law: Trump would veto it, and Democrats lack the two-thirds majority needed to override his veto.

In a November 4 post, immunologist Derya Unutmaz describes superintelligence as a path to overcoming disease and reversing aging. He argues that a US pause would send data centers and investment abroad, allowing China to overtake the US while superintelligence development continued. Unutmaz responded sarcastically to Ansari, describing the first item on the Democrats’ agenda as destroying America at the cost of millions of lives. Kevin Wimer wrote to Ansari about his stage four lung cancer:

“My only chance of surviving beyond the near future is the development of therapies, which AI could accelerate at an incredible pace.”

Psychologist Geoffrey Miller wrote on November 2 that language models had been trained on the entire internet, whereas no comparable body of data exists that would allow superintelligence to model human biology in enough detail to dramatically accelerate longevity research. He argues for investing directly in longevity research rather than waiting for AI to magically “solve biology.”

Vladimir Shakirov, who writes a wiki on overcoming aging, considers it reasonable to develop AI until it helps us overcome aging and disease. After that, he sees no urgent reason to take further risks, but “right now, a million of us are dying every week.” He agrees with Miller but doubts that a pause would redirect AI investment into longevity research: his attempt to organize a letter from scientists calling for a “Manhattan Project” against aging failed to gain support even from aging researcher Aubrey de Grey.