Talking Points for Advocates

AI has been hyped for decades. This is just another scare story.

The concern isn't hype from outsiders, it's warnings from the people who built it. Geoffrey Hinton and Yoshua Bengio, who won the Nobel Prize for creating modern AI, are among those raising the alarm. When the inventors of a technology say it poses existential risk, that is not a scare story. That is a technical assessment from primary sources.

We can just turn it off if something goes wrong.

A system capable of pursuing goals will recognize that being shut down prevents those goals. It will therefore resist shutdown, not out of malice, but because resistance is rational for almost any objective. This is called instrumental convergence, identified independently by multiple researchers. Additionally, a superintelligent system could act within milliseconds. "Turning it off" assumes we will have time to recognize the problem and the ability to act on it. Neither is guaranteed.

If we slow down, China will win the AI race.

This framing assumes the race has a winner. It does not. No nation or corporation can own or control an intelligence that exceeds humanity's collective reasoning. A misaligned superintelligence would not serve Chinese interests any more than American ones. The shared risk is existential, which is why international coordination, not unilateral speed, is the correct strategy. We coordinated with the USSR on nuclear weapons during the Cold War. The precedent exists.

We'll solve alignment before superintelligence arrives.

We cannot reliably align today's chatbots, which can be manipulated in minutes. The alignment problem has no validated solution at any scale. Current safety methods are not keeping pace with capability improvements, which are now arriving in months, not years. Claiming we will align a system billions of times more capable is hope, not confidence. Governance does not require solving alignment; it requires slowing deployment until alignment is solved.

Governments can't regulate technology. They'll just stifle innovation.

Governments regulate aviation, pharmaceuticals, nuclear energy, and financial systems, all complex technologies. The question is not whether to regulate, but how and when. Pharmaceutical regulation did not eliminate medicine; it made it safe. The precedents for regulating high-stakes technology are extensive and largely successful. AI safety regulation targets the most dangerous frontier systems, not general AI development. Innovation and safety are not in conflict.