Something I keep harping on, the public being misaligned, e.g. anti-datacenter, is close to your best situation. You got “AI bad.” This is ideal. You even have a bad guy. You don’t get much better than that in politics.
Some bridge-building is still necessary but getting to the negative valence around AI is definitely the hard part. I’m a little concerned about some “stochastic parrot” style complaints but I think those have thinned out, and I think a capable politician can still work with that.
---
Alex Bores is actually solid evidence that anti-AI sentiment translates into AI safety policy. He got a lot of AI safety endorsements for his NY-12 race because of his RAISE Act he passed in state legislature. The industry attacked him—crucially, they attacked him transparently, wanting everyone to know it was the AI industry attacking Alex. And Alex lost −4 compared to prior expectations of −10 or −15.
Alex’s political framing was that he stands up to “tech oligarchs.”
Alex’s policy framing is about transparency and accountability for the AI industry.
This is just how politics works. You harness voter anger and try to do something useful with it. Or do something populist and destructive with it, either way, it’s decoupled.
---
We keep doing something more like, going on interviews and talking about how AI doing racism doesn’t concern us, because our real concern is superintelligence. I think that was Bengio who said that and it just makes us sound racist.
Instead you could say “of course AI is racist, but here’s what they don’t want you to know: they don’t know how to make it NOT be racist even if they wanted to. They literally can’t control how this thing behaves and they want it in every computer system in the planet.”
They are concerned about racism, I am concerned about x-risk, but I am happy to talk about why we need to control AI because of untold levels of racism (actually we are concerned about that too, we call it s-risk in the authoritarianism context) and set aside my x-risk lecture. This is bridge-building and either way I wind up educating on, and framing the problem/solution around the control problem.
---
Anyway, your problem is super PACs going after AI safety candidates and distracting from AI. Bores’ case above shows talking about AI backfiring for the AI industry. Crypto and AIPAC distract from the core issue and they’ve been more successful. But also note the dam is breaking on the pro-Palestine side among Democrats.
(The last sentence above is a bad “conclusion” sentence; I’m interested if someone runs with it though and places AI within modern Dem populism.)
“of course AI is racist, but here’s what they don’t want you to know: they don’t know how to make it NOT be racist even if they wanted to. They literally can’t control how this thing behaves and they want it in every computer system in the planet.”
Strong upvoted because of this bit. It’s astounding that we haven’t been messaging like this the whole time; clearly people already understand “alignment” in a basic sense, and it seems straightforwardly good to work with them here instead of saying “um actually my concern is more important.”
Something I keep harping on, the public being misaligned, e.g. anti-datacenter, is close to your best situation. You got “AI bad.” This is ideal. You even have a bad guy. You don’t get much better than that in politics.
Some bridge-building is still necessary but getting to the negative valence around AI is definitely the hard part. I’m a little concerned about some “stochastic parrot” style complaints but I think those have thinned out, and I think a capable politician can still work with that.
---
Alex Bores is actually solid evidence that anti-AI sentiment translates into AI safety policy. He got a lot of AI safety endorsements for his NY-12 race because of his RAISE Act he passed in state legislature. The industry attacked him—crucially, they attacked him transparently, wanting everyone to know it was the AI industry attacking Alex. And Alex lost −4 compared to prior expectations of −10 or −15.
Alex’s political framing was that he stands up to “tech oligarchs.”
Alex’s policy framing is about transparency and accountability for the AI industry.
This is just how politics works. You harness voter anger and try to do something useful with it. Or do something populist and destructive with it, either way, it’s decoupled.
---
We keep doing something more like, going on interviews and talking about how AI doing racism doesn’t concern us, because our real concern is superintelligence. I think that was Bengio who said that and it just makes us sound racist.
Instead you could say “of course AI is racist, but here’s what they don’t want you to know: they don’t know how to make it NOT be racist even if they wanted to. They literally can’t control how this thing behaves and they want it in every computer system in the planet.”
They are concerned about racism, I am concerned about x-risk, but I am happy to talk about why we need to control AI because of untold levels of racism (actually we are concerned about that too, we call it s-risk in the authoritarianism context) and set aside my x-risk lecture. This is bridge-building and either way I wind up educating on, and framing the problem/solution around the control problem.
---
Anyway, your problem is super PACs going after AI safety candidates and distracting from AI. Bores’ case above shows talking about AI backfiring for the AI industry. Crypto and AIPAC distract from the core issue and they’ve been more successful. But also note the dam is breaking on the pro-Palestine side among Democrats.
(The last sentence above is a bad “conclusion” sentence; I’m interested if someone runs with it though and places AI within modern Dem populism.)
Strong upvoted because of this bit. It’s astounding that we haven’t been messaging like this the whole time; clearly people already understand “alignment” in a basic sense, and it seems straightforwardly good to work with them here instead of saying “um actually my concern is more important.”