This is a special post for quick takes by Maxime Fournes. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Comment Permalink
Crunch Time is Now. Drop Everything.
Epistemic status: Hopeful, stressed, terrified, ecstatic.
tl;dr This could be it. For the next few days, do nothing but contact journalists and policymakers (ideally via existing orgs), attend/share/organise protests near you, and use your public profile to speak out.
This is Our Chance
The explosion of interest in AI safety over the last week has been both inspiring and dizzyingly fast. We might actually win this thing.
Events since the Jacob Coxon resignation have blown the Overton window as wide open as it is likely to get. This is the kind of chance we have dreamed of, and we have a responsibility not to waste it. We may not get another.
What You Can Do
This window will not last forever. We need to act over a matter of days, not weeks or months.
This is the time to call in favours, work overtime, and put all your strength of heart and brain into keeping our world safe.
If you work in AI safety, we think the most impactful thing for you to do is to drop everything you are doing and pour your energy into policy and media outreach.
Politicians and journalists want to be able to speak on this issue but are keenly aware of their lack of technical knowledge. Cold-emailing and cold-calling them and offering to brief them has a surprisingly high likelihood of success.
This especially includes conservative ones. We don’t care what your personal politics are, survival is a non-partisan issue.
There are many ways in which people of all skillsets and levels of experience can contribute, but we suggest three key actions:
- Contact journalists and members of congress/parliament, This is best done by volunteering for an existing organisation. We are capacity constrained and need your help. You do not need to be an expert: there are plenty of jobs that need doing.
- PauseAI Global
- People for a Pause
- ControlAI
- Humans in Control
- Humans First
- Evitable
- Any other group you have a line into
- Share and attend protests in your area and invite your friends and family. For instance:
- This PauseAI protest in London
- This People for a Pause protest in DC
- Occupy OpenAI (in San Francisco and New York)
- This PauseAI protest in Berlin
- Speak out: Twitter, Instagram, wherever. Be grounded in facts, but also convey the right mood. Also, we’re in the Overton window - speak accordingly.
We need to put out clear, actionable asks for policymakers to translate this momentum into real solutions. We suggest:
- US - The people really want Donald Trump to negotiate a pause treaty with Xi Jinping and he might get a Nobel Peace Prize if he makes the deal. National security depends on it.
- UK - Andy Burnham needs to announce that the UK will convene a pause treaty summit and thereby protect its citizens and honour its legacy of global influence and world-leading safety talent (e.g. Bletchley Park Summit, UK AISI).
- Elsewhere - Every government can act within their own borders and put pressure on the US government to get on board. We cannot let unaccountable US tech giants put us all at risk.
Good Luck and Godspeed
For the first time in a very long time, we actually think things might go well for humanity. Whether that happens is up to us.
See you on the other side.
I think that bipartisanship is a terrible strategy that is completely disconnected from reality right now, the focus needs to be on impeaching Trump and Vance if we want to have any hope for AI regs, restoring USAID or to save ourselves from being robbed blind by corruption. Trump still denies global warming after years of CO2 caused temperature increases and disasters. He has no regrets for USAID and global health cuts that have led to the spread of Ebola and screw worms. He is never going to care about speculative AI risks and is obviously not either a rational actor or someone concerned by the welfare of others.
Trump will be President until 2029. He will not sign AI safety legislation passed on Democratic party lines. Overriding his veto (or an impeachment conviction) would require 1/3 of GOP senators to cross party lines, which will never happen. Making AI safety partisan would ensure we have a 0% chance of federal AI safety legislation until at least 2029. Trump’s “paced the frontier” before by limiting Fable access when it was in his interest, and he may do it again if he perceives it to be in his interest. I understand your frustration with Trump, but there is nothing to gain and a lot to lose from partisanship here.
I would say the opposite. There is nothing to gain and everything to lose from non partisanship here.
The 1/3 of Republicans in the Senate number is not fixed it is dependent on the midterms. If by some miracle Dems won every seat that is up for grabs they would get 67 seats and be able to impeach Trump with only their own votes. A far more likely scenario is to get less than that but be able to get Republican votes for impeachment, every seat matters for that. The difference between 50-50, 51-49, 52-48 results are all massive.
If impeachment is unreachable bills can be passed through vetoes, again at the 2/3 margin.
Convincing swing Republican senators to vote for things that we want such as a restoration of USAID and AI regs doesn't necessarily benefit from bipartisanship either. It is more important to have intimidating voter and $ power. This requires punishing Republicans and Trump for violating all of the values that we hold dear, so that they can learn that that is a bad electoral strategy.
Always being nonpartisan regardless of the actions of parties is identical to always folding in an iterated game of The Prisoner's Dilemma. This will get us killed. We need to adopt a tit for tat strategy and find a way to be a credible threat at the ballot box or the pro authoritarian, accelerationist, pro pandemic side will win.
Apparently AI is not even a winning issue for Democrats right now:
https://x.com/DKThomp/status/2099871668411805768
So I don't see how doubling down on Democratic Party partisanship gets us anywhere.
Remember that in the US, Independents are much more common than either Democrats or Republicans.
The Trump stuff may all be true. Nonetheless, getting Trump impeached would basically need to be a bipartisan effort.
We might be able to win with Democrats + Independents. But the way to win the Independents, in my mind, is to make arguments which could, at least in theory, persuade Republicans.
On that poll I would agree with Brian Schatz reply "wait a week." The voters are not aware of the different parties actual views on AI, but right now Dems ARE heavily winning the polls for the midterms and Trump is polling terribly, so he is not someone we want to associate with for the sake of general popularity.
Yes, it's important to make arguments that can convince independents and Republicans, I certainly agree with that. This just can't extend to ignoring the fact that our single biggest lever right now BY FAR is the midterms and getting the people who support autocracy, the elimination of USAID, almost zero pandemic or climate change prevention, and zero AI regulation out.
Doing that almost certainly requires conversion and requires arguments suited for people who previously voted for Republicans, but it also requires pointing towards Republican elected as the catastrophe they are and not associating ourselves with them or aiding them.
Why is that Yud and everyone who follows him is constantly terrified of cloud software killing us all without being able to explain HOW or WHY AI would kill us all?
I'm sorry, this is science fiction cosplaying as emergency preparedness. The AI labs are preaching a "pacing" or "slow down" because the models are plateauing and they want to IPO. You really think Altman is pushing the IPO because of his grave responsibility to humanity via AI safety? The guy can't IPO because their finances are a dog's breakfast.
AI is powerful and can cause real-world harms. Hacking, AI psychosis, environmental harms are all real consequences of unchecked AI development. But this ain't Terminator.
Crunch Time is Now. Drop Everything.
Epistemic status: Hopeful, stressed, terrified, ecstatic.
tl;dr This could be it. For the next few days, do nothing but contact journalists and policymakers (ideally via existing orgs), attend/share/organise protests near you, and use your public profile to speak out.
This is Our Chance
The explosion of interest in AI safety over the last week has been both inspiring and dizzyingly fast. We might actually win this thing.
Events since the Jacob Coxon resignation have blown the Overton window as wide open as it is likely to get. This is the kind of chance we have dreamed of, and we have a responsibility not to waste it. We may not get another.
What You Can Do
This window will not last forever. We need to act over a matter of days, not weeks or months.
This is the time to call in favours, work overtime, and put all your strength of heart and brain into keeping our world safe.
If you work in AI safety, we think the most impactful thing for you to do is to drop everything you are doing and pour your energy into policy and media outreach.
Politicians and journalists want to be able to speak on this issue but are keenly aware of their lack of technical knowledge. Cold-emailing and cold-calling them and offering to brief them has a surprisingly high likelihood of success.
This especially includes conservative ones. We don’t care what your personal politics are, survival is a non-partisan issue.
There are many ways in which people of all skillsets and levels of experience can contribute, but we suggest three key actions:
We need to put out clear, actionable asks for policymakers to translate this momentum into real solutions. We suggest:
Good Luck and Godspeed
For the first time in a very long time, we actually think things might go well for humanity. Whether that happens is up to us.
See you on the other side.
Why is that Yud and everyone who follows him is constantly terrified of cloud software killing us all without being able to explain HOW or WHY AI would kill us all?
I'm sorry, this is science fiction cosplaying as emergency preparedness. The AI labs are preaching a "pacing" or "slow down" because the models are plateauing and they want to IPO. You really think Altman is pushing the IPO because of his grave responsibility to humanity via AI safety? The guy can't IPO because their finances are a dog's breakfast.
AI is powerful and can cause real-world harms. Hacking, AI psychosis, environmental harms are all real consequences of unchecked AI development. But this ain't Terminator.