Yes, I agree that 1 is generally best.
Regarding 3, I agree that for some people, and I can include myself among them, making sense of this untrusted advice can be fun and part of the game. However, for some people, at least in my experience, this just doesn’t work. For sure, the outside of the game, initial explanation can at least make them keep their guard higher though.
I actually didn’t know the eurogame genre (as a word) and, after reading about it, I realize why games like Catan or Carcassonne are my preferred choice when playing with a diverse mix of players. They indeed allow the strongest players to have a good game despite bad decisions from the weaker ones, and also allow the weaker ones to do their thing without being crippled too early.
In my previous comment I was mainly thinking about Risk.
EA and the AI safety community are fumbling the “Coxon-NavierStokes” opportunity to raise awareness.
The debate I see on X/mastodon/substack/traditional social networks is maybe interesting for 3% of the population:
1% finance bros that are going to write down everything we say as nerd paranoia
1% people already familiar with the problem, that either already agree or disagree
1% that maybe can start doing something about it
Often, when we talk about AI-safety with technical people already familiar with AI, it is a struggle to convince them that there are dangers in the developments of more powerful AIs.
On the contrary, when I talk with old middle/high-school friends, or family friends, it is a completely different story. Often I get asked if it is still possible to stop/kill “the AI” just when I explain that we are training it to become as good as us at our jobs. No even need to start talking about HuggingFace, rogue AIs, RSIs.
This is a big part of the population, that successfully held back / slowed down vaccines, electric vehicles, nuclear energy… It may be the unusual ally for Less Wrong, but they have a great track record for fighting their battles. Are we sure it is better to organize an AI-safety conference / a workshop at LightHaven than to just invest in tiktok/facebook ads against AGI? There are so many voters that are so ready to throw themselves against the AI labs, but just have no idea of even a tiny part of what is happening.
Of course, the things we already do are important, I am just exaggerating to convey the point, but the fact is that if we want laws and regulation, we need to provide politicians with good reasons to implement them. Bernie Sanders is already against them, and midterms are in two months: if the AI-safety debate became more widespread, this is the right timing to turn AI-safety into a bipartisan issue.