Demand

Our premise

AI needs regulation. That much is set, and we say so openly.

Our base demand

Slow down AI development, so that not only research but also society can keep pace with the changes.

The instruments

• Licences for training and deployment.
• An international AI supervisory authority.
• Control of compute, the computing power behind AI.
• Liability of providers.

We orient ourselves by the demands of A Narrow Path and PauseAI Deutschland, because our focus is on informing. We adopt instruments, not their timing. This list is our own selection and wording, not every proposal covers every instrument, and “providers” is our term.

How our position arises

• Member survey. Each member will rate each class of hazard, from “harmless” to “should not be reached”, and state coarse attributes such as age group, gender, occupation and country.
• Rating corridor. For each class of hazard, we will show the 5 %, 25 %, 50 %, 75 % and 95 % quantiles of the results. This range is our official position.
• 25 % quantile. For each class, our demand follows what a quarter of members consider necessary at the latest. We choose this out of precaution, and the median stays visible next to it.
• Not one single regulation. Some measures may be wanted at once, for example to soften upheaval from mass unemployment. Others matter at other stages, for example to prevent human extinction.
• Anonymity. The token that lets you answer will not be linkable to you. Rare combinations of attributes could still identify someone, so we will evaluate coarse categories only, and groups above a minimum size. Once the list of who collected a token is deleted, outsiders cannot fully recheck how many were issued.

These are the opinions of our members, not of the population. We will verify members by e-mail and Discord, and that has limits.

Current results

There are none yet. The first survey round has not taken place, so we do not currently state any timing. Results will appear here next to representative polls.