Es gibt eine Lösung für das existenzielle Risikoproblem der KI: einen bedingten KI-Sicherheitsvertrag

„Hundreds of AI scientists [think](https://www.safe.ai/work/statement-on-ai-risk) we might lose control over AI once it gets too capable, which could result in human extinction. So what can we do?

[Nobel Laureate](https://time.com/7065011/nobel-prize-2024-winners/) [Geoffrey Hinton](https://time.com/collection/time100-ai/6309026/geoffrey-hinton/), Turing Award winner Yoshua Bengio, and many others have provided a piece of the solution. In a policy paper published in *Science* earlier this year, they [recommended](https://www.science.org/doi/10.1126/science.adn0117) “if-then commitments”: commitments to be activated if and when red-line capabilities are found in frontier AI systems.

Building upon their work, we at the nonprofit Existential Risk Observatory propose a Conditional AI Safety Treaty. Signatory countries of this treaty, which should include at least the U.S. and China, would agree that once we get too close to loss of control they will halt any potentially unsafe training within their borders.

Once the most powerful nations have signed this treaty, it is in their interest to verify each others’ compliance, and to make sure [uncontrollable AI](https://time.com/6258483/uncontrollable-ai-agi-risks/) is not built elsewhere, either.

One outstanding question is at what point AI capabilities are too close to loss of control. We propose to delegate this question to the [AI Safety Institutes](https://time.com/6962503/ai-artificial-intelligence-uk-us-safety/) set up in the [U.K](https://www.aisi.gov.uk/)., [U.S](https://www.nist.gov/aisi)., [China](https://beijing.ai-safety-and-governance.institute/), and [other countries](https://www.commerce.gov/news/press-releases/2024/09/us-secretary-commerce-raimondo-and-us-secretary-state-blinken-announce). They have specialized model evaluation know-how, which can be developed further to answer this crucial question.

Also, these institutes are public, making them independent from the mostly private AI development labs. The question of how close is too close to losing control will remain difficult, but someone will need to answer it, and the AI Safety Institutes are best positioned to do so.“

Excellent, now solve how to enforce treaties. We’ve seen in the past few years that most treaties don’t mean diddly to some countries.

Wait, are we seriously discussing this? Wow, I feel like I’m in a sci-fi movie.

Why wait until we are on the verge of losing control? Why not stop now?

God this anti-strong AI hysteria is getting so tedious. We _need_ strong AI to save us from ourselves, one way or another.