I generally support the idea of 80k Hours putting more emphasis on AI risk as a central issue facing our species.

However, I think it's catastrophically naive to frame the issue as 'helping the transition to AGI go well'. This presupposes that there is a plausible path for (1) AGI alignment to be solved, for (2) global AGI safety treaties to be achieved and enforced in time, and for (3) our kids to survive and flourish in a post-AGI world.

I've seen no principled arguments to believe that any of these three things can be achieved. At all. And certainly not in the time frame we seem to have available.

So the key question is -- if there is actually NO credible path for 'helping the transition to AGI go well', should 80k Hours be pursuing a strategy that amounts to a whole lot of cope, and rearranging deck chairs on the Titanic, and gives a false sense of comfort and security to AI devs, and EA people, and politicians, and the general public?

I think 80k Hours has done a lot of harm in the past by encouraging smart young EAs to join AI companies to try to improve their safety cultures form within. As far as I've seen, that strategy has been a huge failure for AI safety, and a huge win for immoral AI companies following a deeply cynical strategy of safety-washing their capabilities development. OpenAI, DeepMind, Anthropic, xAI have all made noise about AI risks... and they've all hired EAs... and they've carried on, at top speed, racing towards AGI. 

Perhaps there was some hope, 10 years ago, that installing a cadre of X-risk-savvy EAs in the heart of the AI industry might overcome its reckless incentives to pursue capabilities over safety. I see no such hope any more. Capabilities work has accelerated far faster than safety work. 

If 80k Hours is going to take AI risks seriously, its leadership team needs to face the possibility that there is simply no safe way to develop AGI -- at least not for the next few centuries, until we have a much clearer understanding of how to solve AI alignment, including the very thorny game-theoretic complications of coordinating between billions of people and potentially trillions of AGIs. 

And, if there is no safe way to develop AGI, let's stop pretending that there is one. Pretending is dangerous. Pretending gives misleading signals to young researchers, and regulators, and ordinary citizens. 

If the only plausible way to survive the push towards AGI is to entirely shut down the push towards AGI, that's what 80k Hours needs to advocate. Not nudging more young talent into serving as ethical window-dressing and safety-washers for OpenAI and Anthropic.

Strongly endorsing Greg Colbourn's reply here. 

When ordinary folks think seriously about AGI risks, they don't need any consequentialism, or utilitarianism, or EA thinking, or the Sequences, or long-termism, or anything fancy like that.

They simply come to understand that AGI could kill all of their kids, and everyone they ever loved, and could ruin everything they and their ancestors ever tried to achieve.

Alex - thanks for the helpful summary of this exciting new book.

It looks like a useful required textbook for my 'Psychology of Effective Altruism' course (syllabus here), next time I teach it!

Well, the main asymmetry here is that the Left-leaning 'mainstream' press doesn't understand or report the Right's concerns about Leftist authoritarianism, but it generates and amplifies the Left's concerns about 'far Right authoritarianism'.

So, any EAs who follow 'mainstream' journalism (e.g. CNN, MSNBC, NY Times, WaPo) will tend to repeat their talking points, their analyses, and their biases.

Most reasonable observers, IMHO, understand that the US 'mainstream' press has become very left-leaning and highly biased over the last few decades, especially since 2015, and it is functioning largely as a propaganda wing of the Democratic Party. (Consider, for example, the 'mainstream' media's systematic denial of Biden's dementia for the last several years, until the symptoms became too painfully obvious, to everyone, to ignore. Such journalists would never have run cover for Trump, if he'd been developing dementia; they would have been demanding his resignation years ago.)

In any case, the partisan polarization on such issues is, perhaps, precisely why EAs should be very careful not to wade into these debates unless they have a very good reason for doing so, a lot of political knowledge and wisdom, an ability to understand both sides, and a recognition that these political differences are probably neither neglected nor tractable.

If we really want to make a difference in politics, I think we should be nudging the relevant decision-makers, policy wonks, staffers, and pundits into developing a better understanding of the global catastrophic risks that we face from nuclear war, bioweapons, and AI.

Yelnats - thanks for this long, well-researched, and thoughtful piece.

I agree that political polarization, destabilization, and potential civil war in the US (and elsewhere) are worthy of more serious consideration within EA, since they amplify many potential catastrophic risks and extinction risks.

However, I would urge you to try much harder to develop a less partisan analysis of these issues. This essay comes across (to me, as a libertarian centrist with some traditionalist tendencies) as a very elaborate rationalization for 'Stop Trump at all costs!', based on the commonly-repeated claim that 'Trump is an existential threat to democracy'. And a lot of the rhetoric, and examples, are basically repeating highly partisan Democratic Party talking points, which have been promoted ad nauseum by CNN, MSNBC, Washington Post, NY Times, etc. And, many of which have been debunked upon further investigation.

 EAs tend to lean Left. We know this from EA surveys. Rich EAs (such as SBF) have donated very large sums of money to Democratic candidates. That makes it very important for us to become more aware of our own political biases, when we address issues such as polarization.

In my opinion, both current US political parties are showing some highly authoritarian tendencies. You mentioned some authoritarian tendencies from the Republican side. But you seem to have overlooked many authoritarian trends on the Democratic/Leftist side, which have included:

1. outsourcing government censorship to social media and Big Tech, especially from 2015 through today
2. demanding lockdowns, school closures, mandatory vaccinations, and public masking during Covid
3. using organized 'lawfare' against political opponents, including Trump
4. promoting pro-Jihadist, pro-Hamas, anti-semitic protests on college campuses, 
5. promoting racially and sexually divisive identity politics in public K-12 schools and universities,
6. threatening the independence and integrity of the Supreme Court (e.g. by planning court-packing, AOC threatening to impeach SCOTUS justices, and Biden ignoring SCOTUS decisions, e.g. prohibiting 'forgiving' student loans)
7. promoting infringements on Second Amendment rights through unconstitutional 'gun control' legislation and executive orders
8. promoting 'regulation by enforcement', e.g. Gary Gensler & Elizabeth Warren weaponizing the SEC to harass the crypto industry and infringe on American rights to hold digital assets
9. undermining states' autonomy to pass their own laws regarding controversial social and sexual issues, such as abortion
10. promoting a Central Bank Digital Currency, which would create a 'financial panopticon' in which the federal government has information about every economic transaction that every citizen makes
11. trying to protect the permanent, unelected, unaccountable, federal bureaucracy (aka 'deep state') from legislative oversight (e.g. objecting to SCOTUS recently overturning 'Chevron deference')
12. turning federal law enforcement (e.g. the FBI) into a partisan weapon for demonizing political opponents (e.g. claiming the 'white nationalism' is the 'biggest terrorist threat to America')
13. treating the outcome of the 2016 presidential election as illegitimate for years afterwards, e.g. blaming it on 'Russian interference'
14. defending a manifestly senile president (Biden) whose executive branch seems to be run by an unelected, unaccountable, shadowy set of advisors who remain unknown to most citizens, and who have concentrated a huge amount of authoritarian power behind an aging, incompetent figurehead.

Many on the Left think of 'authoritarianism' as a purely Right-wing phenomenon, following the Frankfurt School Leftists such as Adorno et al. publishing 'The authoritarian personality' (1950). However, more recent work in political psychology shows that there are plenty of Leftist authoritarians. Also, history reveals plenty of examples of authoritarian socialists, such as Lenin, Stalin, Mao, Pol Pot, Castro, etc -- who are responsible for tens of millions of deaths.

Moreover, the standard 2-D graph of political orientation, which includes a Left-vs-Right dimension, but also an Authoritarian-vs-Libertarian dimension, reminds us that the Right does not have a monopoly on authoritarianism.

So, I would urge you to continue this work, but to re-examine your own political biases, and perhaps to collaborate with researchers who hold more diverse political views, such as Centrists, Libertarians, Conservatives, Neo-Reactionaries, Nationalists, Populists, etc. 

I expect this comment to be downvoted into oblivion by EAs who reflexively think 'Trump bad, Progressives good'. 

But I beseech you all, consider the possibility that the Democrats are just as much of a threat to American democracy and liberty as the Republicans have ever been. 

Peter -- This is a valuable comment; thanks for adding a lot more detail about this lab.

Vasco -- understood. The estimate still seems much lower than most other credible estimates I've seen. And much lower than it felt when we were living through the 70s and 80s, and the Cold War was still very much a thing.

Raemon -- I strongly agree, and I don't think EAs should be overthinking this as much as we seem to be in the comments here. Some ethical issues are, actually, fairly simple.

OpenAI, Deepmind, Meta, and even Anthropic are pushing recklessly ahead with AGI capabilities development. We all understand the extinction risks and global catastrophic risks that this imposes on humanity. These companies are not aligned with EA values of preserving human life, civilization, and sentient well-being. 

Therefore, instead of 80k Hours advertising jobs at such companies, which does give them our EA seal of moral approval, we should be morally stigmatizing them, denouncing them, and discouraging people from working with them.

If we adopt a 'sophisticated', 'balanced', mealy-mouthed approach where we kinda sorta approve of them recruiting EAs, but only in particular kinds of safety roles, in hope of influencing their management from the inside, we are likely to (1) fail to influence management, and (2) undermine our ability to use a moral stigmatization strategy to slow or pause AGI development. 

In my opinion, if EAs banded together to advocate an immediate pause on any further AGI development, and adopted a public-relations strategy of morally stigmatizing any work in the AI industry, we would be much more likely to reduce AI extinction risk, than if spend our time trying to play 4-D chess in figuring out how to influence AI companies from the inside. 

Some industries are simply evil and reckless, and it's good for us to say so. 

Let's be honest with ourselves. The strategy we've followed for a decade, of trying to influence AI companies from the inside, to slow capabilities development and to promote AI alignment work, has failed. The strategy of trying to promote government regulation to slow reckless AI development is showing some signs of success, but is probably too slow to actually inhibit AI capabilities development. This leaves the informal public-relations strategy of stigmatizing the industry, to reduce up its funding, reduce its access to talent, and to make it morally embarrassing rather than cool to work in AI. 

But EAs can only pursue the moral stigmatization strategy to slow AGI development if we are crystal clear that working on AGI development is a moral evil that we cannot endorse. 

Michael -- I agree with your assessment here, both that the CEARCH report is very helpful and informative, but also that their estimated likelihood of nuclear (only 10% per century) seems much lower than seems reasonable, and much lower than other expert estimates that I've seen. 

Just as a lot can happen in a century of AI development, a lot can happen over the next century that could increase the likelihood of nuclear war.