
Interested in #climate and #energy policy. Motivated by #effectivealtruism. Works at @FoundersPledge, views are my own.
Berlin, Deutschland
Joined February 2013
- Tweets1.8K
- Following1.2K
- Followers675
- Likes1.4K
Pinned Tweet
1/10 What if philanthropy worked more like thoughtful investing—making bets based on evidence, not emotion? @HeatmapNews chronicles how this approach unlocked $100M+ in climate funding: heatmap.news/climate/founder…🧵
Johannes Ackva retweeted
Incredible news to wake up to: Obsolete is one of the Financial Times' best economics books of 2026!
"What is astonishing, as this book makes clear, is how complacent we are about the multiplicity of risks that this creates."
Writing a book with a strong perspective, there was always a risk Obsolete would be categorized as a polemic and discounted as such. But I aimed to make my argument with as much rigor and evidence as possible, and it's really nice to see it pay off!
obsoletebook.org/
ft.com/content/45fcd5a0-54d2…
Johannes Ackva retweeted
If you care a lot about AI risks, and are “high-context” about AI safety, and are persuaded by @CharlieBull0ck’s (correct!) appeal that working as a Congressional staffer for the next couple years is a high-impact way to make a difference, DM me!
AI safety twitter please share this post, to increase visibility!
If you're in AI policy in DC, or if you know about AI policy and you can move to DC, you should strongly consider trying to work as a congressional staffer. Senior people should consider doing this as well as junior people, if it's the right role in the right office. Many, many offices are staffing up and looking to hire AI people. We all have a duty to lock in and get serious and solve the problems we need to solve.
If you do this and work hard and accept the sadly inadequate pay, you will be an American hero who is on the front lines of society's response to the most important challenge humanity has ever overcome. It is hard for me to put into words how much better of a thing that is to do than going to work for a lab, and it's also a lot better than the kind of cushy think tank gig that I currently have.
Working for a state government in an important role implementing and enforcing the AI regulations that we actually have or will have is equally important.
Johannes Ackva retweeted
If you're in AI policy in DC, or if you know about AI policy and you can move to DC, you should strongly consider trying to work as a congressional staffer. Senior people should consider doing this as well as junior people, if it's the right role in the right office. Many, many offices are staffing up and looking to hire AI people. We all have a duty to lock in and get serious and solve the problems we need to solve.
If you do this and work hard and accept the sadly inadequate pay, you will be an American hero who is on the front lines of society's response to the most important challenge humanity has ever overcome. It is hard for me to put into words how much better of a thing that is to do than going to work for a lab, and it's also a lot better than the kind of cushy think tank gig that I currently have.
Working for a state government in an important role implementing and enforcing the AI regulations that we actually have or will have is equally important.
Johannes Ackva retweeted
it's still hard to imagine that USG lets this happen; this version of RSI is too fast for frog-boiling, but too slow for unilateral lab takeover.
if it unfolds, I think enough lab people will be too skeptical of handoff come crunchtime, sound the alarm, and USG shuts it all down
Johannes Ackva retweeted
Barack Obama: AI is reaching the point where it no longer needs humans to teach it. The era of exponential intelligence growth is upon us.
“The machines are starting to be able to teach themselves. They don't have to wait for humans to start figuring things out.”
“We're now at a stage where it's 50-50, and if the trend lines continue, the expectation is it might be 90-10.”
“We're now hitting the hockey stick trajectory where the models are going to get smarter and smarter at an exponential pace.”
“If they start setting their own agendas, you may get a misalignment between what they want to do and what we want them to do.”
“And that gap can be dangerous.”
Johannes Ackva retweeted
I think more donors should consider funding grassroots AI risk advocacy (e.g. Humans in Control, PauseAI UK).
It's not the first method I'd choose for tackling AI risk.
But it's by some margin the most underfunded part of the broader AI risk ecosystem today.
That's no small matter. Individual donors often find it tough to find AI projects where money is still the key bottleneck. But for these groups it's a huge issue.
For varying reasons the biggest donors are not putting much into it. (While they have concerns about it backfiring, expected impact is not the only barrier.)
Why might this work be good?
Large fractions of the public in the US/UK/etc already support pacing AI development and prioritising doing it very carefully, among many other sensible ideas.
Grassroots organising can better inform people what's going on and channel those attitudes into political outcomes.
With timelines so short most good futures run through some degree of pacing/pausing. That's something the public can understand and successfully push for.
But the general public has to be actively woken up and coordinated or it'll be crushed before it can have much say.
And grassroots orgs take years to build. So if we want them to be around to take advantage of much higher salience in 2028, the money has to go in now.
Why might it be bad?
Grassroots movements tend to have low epistemic hygiene, and over time get confused about their topic of concern (audience capture, rage/purity spirals, etc).
As a result it's inevitable that some members of any mass movement will make a lot of factual errors. And amongst good work, will end up advocating for ineffective policies, and even some harmful ones.
(For instance, yesterday I pushed back on where I see some anti-AI grassroots organising verging on 'Human exclusivism' that I think is seriously misguided.)
That kind of thing can make intellectuals allergic to grassroots organising.
But while it's risky, everything in this space is risky and can easily backfire. I'm not sure grassroots organising particularly stands out in that respect.
There is a public backlash already underway, if people concerned with loss of control aren't in that conversation, it will be more confused and do worse things by their lights.
Plus, the policy status quo today is so bad that there are many more ways for things to get better than to get worse. Where the public has strong reactions to this stuff I usually think their common sense takes are solid.
I would bet against the general public coming in and tearing down the few good technocratic things that already exist (e.g. SB53). They're more likely to demand extra transparency and rules that are improvements, even if not the perfect policy.
Furthermore, if 'going off the rails' is the main reservation there are some orgs like Humans in Control or PauseAI UK that seem likely to stay focused and truthful.
ORGS IN THIS BROAD SPACE
Here are some organisations doing things of this kind and a few notes on what they are about (via Claude).
I am not very familiar with their work so do not take their inclusion as an endorsement - I am trying to list what's out there both good and bad:
• Humans in Control (US). Trains volunteer leaders, starting in early-primary states, to make AI safeguards a 2028 election issue. Connects jobs, kids and data centres to loss of control. Long-term goal: a verifiable international deal not to build AI we can’t control.
• PauseAI UK (UK). Protests, plus an open letter from 60+ UK politicians criticising Google DeepMind. Until recently run on £100k a year (!).
• PauseAI Global (Netherlands). Coordinates PauseAI chapters in about 16 countries. Is probably what you think of as PauseAI.
• Evitable (US). AI professor David Krueger’s group to inform and organise the public to end the race to superintelligence.
• Stop the AI Race (US). Started by ex-AI researcher Michael Trazzi. Marches between the labs' SF offices. The ask: each lab stops building new frontier models if the others credibly do the same.
• AISAP (US). AI safety workshops for mainstream audiences in cities around the US, facilitated by local volunteers.
• QuitGPT (US). Boycott of ChatGPT over OpenAI leaders’ political donations and government contracts. Organised by young progressive activists. Claims 1.5M+ participants. (Note, not especially focused on extreme risks.)
• Guardrails (US). A super PAC plus an affiliated nonprofit, launched by Democratic operatives with tech workers and unions behind it. Populist and small-dollar. PAC gifts are limited to US citizens and permanent residents. (Political naturally.)
These are orgs that I have seen criticised or which I expect more people will have reservations about (alas I simply don't have time to investigate myself, but you could):
• ControlAI (UK + US). Single issue: an international ban on superintelligence, not general anti-AI sentiment. Has briefed many UK parliamentarians.
• PauseAI US (US). Protests, lobbying and community-building for an international treaty to pause frontier AI. (Controversial because PauseAI Global publicly disendorsed PauseAI US’s current leadership over repeated personal attacks on people that could be counterproductive. I mostly share these worries.)
• Irreplaceable (US). Founded by climate movement veterans (Sunrise, 350org). Turns anger over data centres, bills, layoffs and kids into a push for “a say, a stake, and a slowdown."
• Humans First (US). A right wing group pushing an ‘America First’ AI policy. Chaired by RNC member and Trump ally Amy Kremer. Ran a ‘Data Center Revolt’ bus tour. Wants mandatory government testing and approval of frontier models before release.
PARTICULAR ISSUES RE GIVING TO ADVOCACY
Because they engage in political advocacy, not all of these orgs offer tax deductible giving, though some do. Some are irrationally put off by the absence of tax deductibility. For top-rate taxpayers lack of tax deductibility can increase the effective price by 2x, but one project can easily be much more than 2x as cost-effective as another, so it shouldn't dominate your thinking.
(In cases where there's both a tax deductible c3 and a political c4, it may be better to give a smaller amount to the non-charitable vehicle so they have more flexibility on spending.)
Some take money from AI company staff, especially junior staff, some don't. I don't think AI company staff should feel sheepish giving to (most of) these groups.
A FURTHER NON-CONSEQUENTIALIST MOTIVE FOR THIS WORK: AUTONOMY, FAIRNESS, DEMOCRACY, RESPECT
A separate consideration in favour of this work that resonates with me is about legitimacy and autonomy.
All of these innocent people are having their lives risked, and their desired futures torn away, without almost any consultation or consent.
If I treated the people around me this way, even in honest pursuit of good consequences, it would rightly be viewed as an appalling violation of their autonomy.
All else equal, it's wrong to knowingly keep adults in the dark about very important factors affecting their lives that they would want to know about and react to.
Even if an individual will react poorly by my lights, that should be their choice to make, not mine. That should be our default stance to the general public in a democracy also.
(Yes, this consideration can be outweighed if the reaction will be sufficiently awful, but respect for other people's autonomy is a default you should have to be argued away from.)
This also makes me allergic to the intellectual's aversion to mass politics. The line of argument that mass politics has such low epistemic standards it's best avoided, if applied consistently, would suggest that the public should be sidelined to the extent possible, and self-assessed experts allowed to decide how society evolves in their place.
While such a system has particular advantages, I don't think it's better all things considered - and history shows it's wildly less robust. A messier less accurate discussion is the price we pay to live in a society in which a wider range of people are given a chance to defend their interests and argue for their beliefs.
Johannes Ackva retweeted
in retrospect AI pause ppl should’ve clarified further when we said "prevent takeoff"
Johannes Ackva retweeted
life still happens at the end of the world, and working through hard things when you're not sure how much time you have left can be tough. with everyone's timelines getting shorter, i decided to write down some thoughts on this
longerramblings.substack.com…
Johannes Ackva retweeted
At Irreplaceable's press briefing just now, I asked about RSI. They said:
"This lawsuit is the argument that it is illegal."
"We're alleging that by racing to recursive self-improvement...that violates California law today. We're asking for an injunction that would stop them."
BREAKING: We’re suing @Anthropic, @OpenAI, @Google, @Meta + @SpaceX to demand an immediate freeze on advanced AI systems that threaten public safety.
Together with @HeatInitiative, @Public_Citizen, and @secureainow, we’re going to court to stop these companies from building AI that could escape human control or enable bioweapons and cyberattacks.
We won’t wait for a catastrophe to act. The people come before Big AI profits.
apnews.com/article/artificia…
Johannes Ackva retweeted
Unfortunately, the people who sound crazy about AI have a far better track record than the people who sound reasonable.
New post -->
Johannes Ackva retweeted
Eli Lifland is a very smart and careful forecaster (he works with Daniel K at AIFP). Last year he thought we had until 2034 before top expert dominating AI (AI better than top human experts at everything), now he thinks we have until 2028, with automated AI research going into overdrive in 2027.
All of Daniel and his colleagues, who are professional AI forecasters who predicted numbers that seemed preposterous at the time but turned out to be correct (around frontier AI company capex numbers, revenue, capability growth etc) now think we will get AIs better than humans at ~everything - systems that make current AI capabilities look impossibly archaic - by 2028. There is variation, but they also on average think that the odds of a civilization scale catastrophe or permanent human disempowerment afterwards is something like a coin flip or worse.
(An aside - reading this report, and then watching the Anthropic roadshow video after made me feel physically ill - I know many of the people at Anthropic are concerned about these risks, but in this presentation they have a breezy tone that with Accenture and Claude’s constitution and other measures to prevent misuse risks they got it under control. I believe that people at AI companies owe the public far more candor than they have provided - for all the talk of them dooming, I find they are actually constantly minimizing what they privately believe.)
The future is not yet written. I have a great deal of faith in the ability of people collectively to rise to the occasion (if only at the last possible moment) to protect themselves and their families. But the current trajectory we are on with AI development is truly terrifying and insane. If you would prefer to not use specific percents or numbers or look at lots of slightly made up graphs that is fine - I don't want to pretend that this is a settled science or that anyone including AIFP knows for sure what is going to happen.
But at some level all they are doing is assuming that what has happened previously will continue happening for another year or two absent intervention, and when you take a detailed and precise accounting of what exactly that means the answer is epistemically violent. That doesn't mean it will happen, but it clearly means to me that taking no serious precautions or plans on that basis (urgently!) is a tremendous abdication of our collective responsibility.
Unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new METR report (though other sources of evidence are also important). If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. I am scared, sad, and angry. blog.aifutures.org/p/q35-tim…
Johannes Ackva retweeted
Unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new METR report (though other sources of evidence are also important). If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. I am scared, sad, and angry. blog.aifutures.org/p/q35-tim…
Johannes Ackva retweeted
If you or someone you know is working on developing viable leg text for an RSI ban, pls DM
Johannes Ackva retweeted
It's actually legit pretty crazy to be so pessimistic about stopping AI, given that:
- Most Americans want that.
- Politicians are actively calling for it.
- EA has a strong, coordinated network of political donors.
- EA's allocation of resources on AI Safety so far has been something like (as a charicature):
90%: technical work
9%: pushing for inadequate policy solutions (I think this work is sometimes net negative towards stopping AI)
<1%: stopping AI
I increasingly think that "can't-do attitudes" among AI Safety / EA are among the biggest obstacles we face.
Spoke to an Anthropic employee today. They have a near-100% p(doom), but they think the only chance of survival is that they're somehow completely wrong about alignment and it turns out to be solvable, and the only way to get there is by accelerating capabilities as fast as possible. They were much more confident that humanity would never have the political will to do what needs to be done to stop ASI development than that alignment would be solved, even though they think alignment will almost certainly not be solved
They have not called their elected representatives to inform them that this is what they believe
Johannes Ackva retweeted
Trump's anti-AI regulation stance might be among the least popular positions an American politician has ever taken.
Makes 'defund the police' look like an electoral powerhouse. (From a convo with @davidshor.)
Johannes Ackva retweeted
There is no better avenue for incoming House & Senate Dems to take than hearings on wtf is going on with NVIDIA + Trump corruption + AI policy in the U.S.
It has the virtue of riding the wave of anti-tech populism and directing that energy towards the productive need to stop selling these chips to China.
Johannes Ackva retweeted
Here’s a pretty important for thing for a new Dem congress to get to work on right away!
This report and reactions from people I trust (like Jake) are making me physically ill. As far as I can tell, Anthropic's central plan is to continue this acceleration of AI capabilities until we're getting decades of AI capabilities growth in a year. Given our very limited ability to understand or control AI systems, this would be extremely reckless and completely unacceptable, and it should be the single top priority of civil society and the US government to stop them and other companies from doing this.







































