Most of the general public seem to be either against AI (because datacenters! and water usage!)
Or is it due to a misperception of AI capabilities? If someone believes the AIs to be stupid, then how can we convince one that the AIs do pose existential risk?
When I read something about AI outside of the rationalist sphere, the consensus seems to be that AIs are useless, harm the environment, violate intellectual property... and if they replace you at your job, it is not because they can actually do something better, but simply because they are cheaper and the greedy capitalists will always prefer cheap low-quality work over a more expensive one.
From that perspective, you might imagine that your childhood toys will magically gain consciousness and exterminate your family, that's how plausible it seems.
The first step is to admit that the AI can do something better than you. Most people are not there yet.
For many people another necessary step will be seeing robots around them, because the "it's all just pixels on the screen, it has no impact on actual life" feeling is very strong.
I've heard people talk a lot about how AI Safety, as a community, has a communication issue. In reality I think we, as society, have a 'difficulty caring about huge abstract problems' issue. This means that I think this issue is partially inherent to being human. But then again, we could probably do better.
As someone who has grown out of a childhood of apathy, I've wrestled a lot with how to think about the transition from not giving a shit to giving a shit. As someone who has recently gotten into the general space of AI Safety/Biosec/EA, I found myself having similar feelings around x-risk, so here's an attempt at formalizing this.
How I think giving a shit works
Phase 1. I've never heard of x- risk and would probably roll my eyes if someone mentioned it to me. It sounds a bit like a conspiracy/cult.
Phase 2. I got over my judgement and started consuming a plethora "Here is my kill-chain of how AI will inevitably lead to our destruction"-style blog posts and youtube videos. A lot of these are incredibly well written and logically sound and great food for thought, but in quite an abstract-thought-experiment way. I think I get the gist of it, but I still don't actually "get it". Not in the same way that I can sit down in my apartment during the hottest June Europe has seen since 1950 and get really anxious about climate change. It's not emotional.
Phase 3. I've engaged with this enough to be able to access a general feeling of urgency around this. I feel more comfortable communicating this to other people. The underlying assumption is that I'm probably still engaging with it in some way because I'm emotionally drawn to it.
Phase 4. I am working/volunteering in or adjacent to x-risk and contributing actual value. A lot of my time is spent on this.
Note the feedback loops. Caring about something will make you engage with it, which will make you know more and care more and engage with it more, etc.
I don't think everyone needs to work on x-risk but ideally, we'd like more people to be in phases 2-4. How to do this is an open question, but a good amount of self-censorship should probably happen talking to people in the first 2 phases.
The transitions
The arrow between every phase represents some kind of hurdle that needs to be overcome.
Phase 1 -> Phase 2
The transition from (1) to (2) is fairly straightforward (but not easy). Somehow, people need to stumble upon accessible media they can consume and consume it.
There are difficulties to this:
The target audience for this step is ideally nearly everyone, but there's an open question to how to communicate this in a palatable way. I think the current PauseAI campaign "AI is not just coming for your job" gets it: tie the story to an already existing sentiment, and expand upon it.
Phase 2 -> Phase 3
This transition is harder than the first. It definitely requires some sustained, self-motivated engagement. It might also be somehow qualitatively different from the first phase transition (1->2) in that it's only accessible to people that already regularly engage with the art of caring (eg. the people that look at something hard and don't just throw their hands in the air).
Luckily, our main target audience for this is people that, in some circles, would be referred to as 'talent' (eg. the coveted people with a specific set of skills). A lot of 'talent' already cares about something so the difficulty is to make the sustained engagement as easily accessible as possible. I think short form courses, such as Bluedot courses or even slightly longer bootcamps do an incredible job at this. The main bottleneck here is spaces without in person communities -> way harder to engage with.
I think there is a separate component to this, which is the shame of caring about something speculative that not everyone is on board with yet. People will look at you weirdly, especially if your communication around this is fear-monger-y. It's a skill to casually speak about the apocalypse but so subtly that it doesn't sound like you're a conspiracy theorist.
Phase 3 -> Phase 4
I'm currently in this transition and I think the main difficulty is simply time and patience.
Maybe I'll update this part when I have the wisdom of hindsight.
Note: of this information is new, it's just structured to help me think about it. This is a first draft on these ideas and something I'm curious to learn more about. Feel free to reach out with anything you've read or written around the topic that you think is relevant.
Crossposted from my Substack.