Summit ministers outreach campaign (e-mails)
The UK AI safety summit in november will be incredibly important. If we can properly prepare those who are invited, and get them to consider a pause, we increase the likelihood of the summit being actually useful.
So we need to:
- Find out who we should focus our attention on. I'm assuming ministers of foreign affairs to receive the invite, and then their chief of staff being responsible for delegating preperations.
- Send them a convincing e-mail. Should inform them about the stakes we're dealing with, and have concrete focus points / suggestions.
- Urgency. We may have very short timelines. We've been surprised before. Legal things move slowly, but this time we need them to move fast. The AI safety summit needs policy outcomes!
- Importance. If we fuck this up, we're putting all lives at risk. The difference between good and bad (or no) policy could mean everything.
- Feasibility of pause / slow down. Slowing down is very popular with normal people. It's also what the scientists have been asking for since quite a while now. The only ones who do not seem to support this, are AI company CEOs, who for some reason keep being the centre of attention in the AI governance debate.
- Clear asks. We should consider focusing on pre-training checks, instead of pause. Sufficient pre-training checks could effectively mean a pause, but if the checks are insufficient we could be doomed. In any case, we NEED a functional international body that can press the brakes if needed.
- Find out who will be invited! I don't know how.
- Research your targets / recipients
- Help writing convincing mails
So we need to:
- Find out who we should focus our attention on. I'm assuming ministers of foreign affairs to receive the invite, and then their chief of staff being responsible for delegating preperations.
- Send them a convincing e-mail. Should inform them about the stakes we're dealing with, and have concrete focus points / suggestions.
What do we want to convey
- Urgency. We may have very short timelines. We've been surprised before. Legal things move slowly, but this time we need them to move fast. The AI safety summit needs policy outcomes!
- Importance. If we fuck this up, we're putting all lives at risk. The difference between good and bad (or no) policy could mean everything.
- Feasibility of pause / slow down. Slowing down is very popular with normal people. It's also what the scientists have been asking for since quite a while now. The only ones who do not seem to support this, are AI company CEOs, who for some reason keep being the centre of attention in the AI governance debate.
- Clear asks. We should consider focusing on pre-training checks, instead of pause. Sufficient pre-training checks could effectively mean a pause, but if the checks are insufficient we could be doomed. In any case, we NEED a functional international body that can press the brakes if needed.
How you can help
- Find out who will be invited! I don't know how.
- Research your targets / recipients
- Help writing convincing mails
