Before AI copies their expertise, I think AI lab employees should try to coordinate an industry wide strike! I think this could be especially effective before the IPOs. If successful, it could be much more effective than quitting individually. Quitting weakens the company, but striking can influence their behavior more directly.
Ideas for demands
You might worry that there aren't any good demands a strike could make. I'm not a lab employee and haven't thought too hard about that the best possible demands, but there are some ideas. Think of these as a lower bound for what you could concretely demand!
Follow through with the merge and assist clause. OpenAI has a merge and assist clause for when they get close to AGI. This demand is met when OpenAI merges with Anthropic.
0% of compute is used for training. This demand is met when no compute is used for training models. Serving current models is fine.
10% of compute used on theoretical alignment research. AI agents have gotten quite good at pure mathematics. It might be worth using, say, this capability on agent foundations or something like that. This demand is met when 10% of compute is used to run inference on such agents.
10% of compute is used on lobbying for an international treaty.In experiments, AI is weakly superhuman at persuasion. This demand is met when 10% of compute is used for lobbying the powers that be to implement an AI pause treaty. An alternative version of this demand is to just lobby Chinese politicians.
To reiterate these are just examples! If your a lab employee I would strongly suggest coming up with your own demands. But hopefully this list shows that they are concrete demands that would be net positive; hopefully you will find even better ones.
Also, even if you aren't a lab employee, consider sharing this post with lab employees if you know any!
Before AI copies their expertise, I think AI lab employees should try to coordinate an industry wide strike! I think this could be especially effective before the IPOs. If successful, it could be much more effective than quitting individually. Quitting weakens the company, but striking can influence their behavior more directly.
Ideas for demands
You might worry that there aren't any good demands a strike could make. I'm not a lab employee and haven't thought too hard about that the best possible demands, but there are some ideas. Think of these as a lower bound for what you could concretely demand!
To reiterate these are just examples! If your a lab employee I would strongly suggest coming up with your own demands. But hopefully this list shows that they are concrete demands that would be net positive; hopefully you will find even better ones.
Also, even if you aren't a lab employee, consider sharing this post with lab employees if you know any!