Thank you for writing this. I use AI, and it’s an effective tool, basically being a large team that works for me. That evens the playing field for small operations like mine.
Can it be misused? Sure, any tool can be. Will it cause mass unemployment? No, for it needs human overseers. Will it end the world? It has as much of a chance ending the world as MS Office does.
Interesting piece, thank you. I think your ending gets to the crux of the matter. Earth will survive humanity. We'll be a blip on Earth's radar. That's a key problem behind the climate-crisis messaging: Too many activists made climate concerns about saving the planet, which was never at issue. With AI, though, it's not clear to me that "AI tools are going to be fine." Nor is it clear that humanity's impact on Earth and AI tools' impact on humanity is a fair comparison. Are AI safety concerns really so overblown?
Fair enough. What would need to happen for your perspective to change on the AI safety front? Just as a thought experiment. Would we need to see something akin to a full-blown AI takeover of a frontier lab? Serious, good-faith question.
Well there's already information online to do whatever you might be paranoid with about AI enabling, the hard part is getting the equipment and building the lab, which you need quite a bit more than just a chatbot to do (and some specialized skills). If you want to be really anxious you'd realize it could in theory be done without AI, but it mostly hasn't outside things like Fentanyl production (also a bioweapon tbh). The cyber security stuff is concerning but it's not existential and anyway that side of the trade is already aware of the risks.
An Inconvenient Truth did not predict a 20-foot sea level rise in the 2010s. No responsible scientists were saying that. I do agree that the movie was less than clear about timelines, and should have been more explicit about dates rather than shifting from one time horizon to another, particularly when it comes to the sea level map Gore displayed.
The 1-2 inches per decade we're seeing now isn't all THAT dramatic, but it's unevenly distributed, modestly accelerating, and pretty much unstoppable. It's going to cause problems eventually.
I gave them to the 2010s given how dire they sounded and how the movie presented the threat but fair that the 20-foot figure was a hypothetical scenario, I updated that sentence to clarify. If you look at what some people have documented on this topic there's a loooong laundry list of specific bad predictions here I just chose that one since I lived in Florida.
But still does get to my larger point. “this will cause problems one day” is an infinite goalpost prediction, similar as the stock market bears who called 100 of the last 3 corrections (and even those were fine and we lived). If the timeline goes forever and severity threshold continually shifts, at what point do we say the original alarm was wrong rather than simply moving goalposts?
I’ve been promised THE END since I was born. 🙄 I’m tired of being disappointed. 😞 Anyone’s promise of doom that’s accompanied by a mandatory new lifestyle change, increased taxation, or subscription fee shall be heretofore summarily invalid and ignored. We should make such pointless prophesying a capital offense.
I’ve talked to countless Xers over the decades about how different our life choices might’ve been if we hadn’t believed Mutually Assured Destruction was going to end the world before we escaped our twenties. 🙄 Mandatory hiding under your desk in elementary school to be prepared for nuclear sodomy leaves an emotional scar. 🤯
- You are trying to cast the alarmists as the entire "environmentalist" movement. That is false, as the environmentalists include everyone working on behalf of the environment! That includes organizations like NASA who warn us about climate change: https://science.nasa.gov/climate-change/ You need to stop casting the most extreme, fringe parts of any group as representative of that group! Especially because you only do that for groups you don't like, which is really just showing bias.
- You oddly include my reference (from Twitter) to the Ozone layer hole without introducing it or even explaining it's relevancy. It's a great example, just like Y2K of big problems that were solved by the hard work of experts and definitely not instances of unjustified alarmism. In these cases (Ozone, Y2K, Climate Change) there are real, qualified and measured threats that present very significant dangers and require coordinated effort to resolve.
- Davos is the World Economic Forum which, while they talk about the environment, is not an environmental conference. It attracts a lot of folks talking about a lot of topics! Its been running since 1971 and while it has many structural problems, claiming it's an environmental conference is false.
- None of those are similar to AI alarmism which is both economic and social. The concerns about AI are not specific or measured, and depend on your perspective.
I am starting to tire of being the resident fact checker here. You can write better pieces that are better argued with better research, and you used to do just that! We would all love for that to return.
Thanks I feel like I write all kinds of different pieces, I'm sorry this one missed the mark for you. Maybe I can help clarify some points because I think there may be some sliiiight misinterpretations here. No AI here (or in piece above) just my thinking I want to clarify what I actually believe.
I never said 100% of people who work on environmental issues are alarmists. But our institutions continue to platform and promote histrionic people who are terrible forecasters, super misanthropic and simply don’t take into account things like technological progress or interpret trends poorly and their doom prognostications never come true. They are in many cases *the* voices for these issues whether we like it or not, even if there are better voices at other institutions - counts for nothing if no one hears them. There’s no accountability either, there’s simply a wake of human tragedy from people who for ex, choose not to have kids because they are living in fear. It’s crazy to pathologize the population like this. There’s no upside, and anyway it should be solved at the commercial level not the retail level.
Here’s a few links you can explore on how millennials are forgoing starting families bc of the climate doom prognostications. I talk to many 20 and 30-somethings, I suppose could have included these in the story to make clearer for those who don’t see it (although did include one link somewhere). This is a big motivation for writing this piece, because we see similar discussions about AI happening already. This also personally affected me for a bit, so I am especially motivated to write on it.
I think you and I perhaps disagree a bit on how evil Ehrlich and Greta and people like them are based on your comments on my previous story about him. We should just disagree on that and move on, it’s a disagreement of opinion, we’re all reading the same set of facts. Easy to do.
The ozone layer ex is really great, thanks for flagging it. I included because it makes an important point about infinite moving goalposts vs solvable problems. I'm sorry I didn't credit you it's a good thing to include. I include thoughts from lots of places and people and if not a formal blog post I don’t always link, sometimes it just adds to a story (lots of bloggers do this). I can link you next time if you want but I really just helped make a point.
I never said Davos is an environmental conference. I said the climate crisis has become a major topic among the political, corporate and NGO elite who attend Davos. That's obviously true, and much of the themes from this event are regularly on this (if you watch any of the broadcasts you’d see this). The point of the image is hypocrisy around elite climate messaging and consumption. I think it's laughable a lot of the same people saying "we are in a climate emergency" don't fly commercial btw. There's 100 stories documenting the hypocrisy on this one (stuff like this: https://cbsaustin.com/news/nation-world/trudeau-slammed-for-extensive-air-travel-while-pushing-ambitious-climate-agenda-justin-canada-airplane-plane-earth-day-prime-minister).
AI to many of us is incredibly close to AI alarmism and I'm sorry you are unable to see it. There’s literally 1,000s of people this week alone who have noted the comparison (https://x.com/search?q=AI%20climate%20doom&src=typed_query&f=live). You can say everyone is wrong if you want (I do this all the time) but it’s a perfect valid opinion.
Many people are already putting off life events because of supposed AI advances, we've seen a lot of this in the news too. This stuff absolutely affects people deeply, as does any apocalyptic proclamation people are really cavalier with and the internet hams up. I wish that would stop, if anything when I flag things I see wrong in the world I try and not be a doomer about it. If everyone did this we’d definitely get more done, things obviously can be done but not if we're paralyzed in fear. And when you say the concerns about AI are ‘not specific or measured,’ that's actually pretty close to my point. That’s precisely why I think we should be much more skeptical of extremely confident, apocalyptic predictions about it.
Also there's one final point you're missing: I don't argue that climate change isn't real or that environmental problems don't exist. I think we've spent decades taking and stoking fears and not enough time building solar and nuclear. We have solutions, we should get excessive regulatory burden out of the way and build.
I'm taking issue with the false comparison you are drawing and the fact that you are over-generalizing in the article to make the piece more sensational. For example:
"The climate crisis has people flying to Davos each year sipping champagne and having caviar on 700-1,000 private jets to discuss what we should do." <- As I pointed out above, Davos is not an environmental conference. You agree here in the comments but not in the piece.
"one of the fathers of the so-called “environmentalist” movement, really just an anti-human movement" <- Ehrlich's book was about economics and resources, not the environment. While he was an environmentalist, that wasn't the primary theme of the book.
"There are perfect parallels one can draw from AI existential fears to both the "climate crisis" and Y2K" <- Y2K is not even close to being an analogy, as previously established.
Anyway, my general problem is that you generalize in these pieces by extrapolating from the fringe to the entire group. I realize that gets more views but it's misleading and requires cherry-picking facts and positions like happened here.
These are real issues that require real analysis, not more alarmism about alarmism that isn't based in reality.
I think we're mostly arguing about framing. I agree climate change is real, many climate models have performed well, and some environmental problems are serious. But I don't think that makes the history of climate alarmism irrelevant, or invalidates the comparison to AI alarmism. It's affecting people in similar ways, and I think that's genuinely concerning. As someone on my LinkedIn commented, if you're looking for a source of the mental health crisis in kids, you could at least look at decades of terrifying them about climate change, and then how quickly many of the same voices have pivoted to AI.
And just a couple factual points: Ehrlich was definitely a major figure in the modern environmental movement beyond The Population Bomb, including making specific environmental doomsday predictions. I also never said Davos was an environmental conference even in the original story. And I think Y2K is useful because it shows how a highly publicized technological threat can be addressed through specific steps rather than permanent emergency politics.
But we agree the current AI fears are alarmist, and maybe that's the most important part here. Also, cool to see you in that WSJ piece.
No, we're arguing about your argument. While I agree with the overall conclusion, your argument building to it is faulty which is what I'm pointing out.
For example, you can't use Davos as evidence UNLESS it is an environmental conference. Otherwise it's unrelated and doesn't support your conclusion.
The fact that The Population Bomb was alarmist and counter-productive is only relevant if it was an environmental book, which it was not. Even if Ehrlich was an environmentalist in his other work.
There is a good argument to be made here, just not the one you made in the piece. The arguments need to hold, be relevant and justified.
“I think the article raises a much stronger hypothesis than the doom piece did, because this one at least points to a recognizable economic mechanism: regulatory capture.
The core idea is straightforward. If compliance with AI-safety rules becomes extremely expensive—mandatory audits, specialized evaluations, reporting systems, security programs, legal teams, licensing, compute thresholds—those costs are much easier for OpenAI, Anthropic, Google, Meta, and Microsoft to absorb than for a startup. Regulation can therefore be genuinely motivated by safety concerns and still function as a moat around incumbents.
And the behavior is worth noticing. OpenAI is now publicly calling for mandatory national AI-safety regulation. It explicitly says the rules should apply to the relatively small group of well-resourced frontier labs. OpenAI also says such regulation should not entrench incumbents and should exempt startups operating below frontier capability levels. That caveat matters—it is evidence against the strongest version of Singer’s argument—but it also confirms that the big labs are actively trying to shape the regulatory architecture while they are already sitting at the frontier.
Anthropic is doing the same thing from a somewhat different direction. Its lobbying expenditures have risen sharply; Axios reported that it spent more than $3.5 million in the first half of 2026, already exceeding its entire 2025 spending. OpenAI also posted record lobbying quarters.
So your suspicion is not coming from nowhere.
Where I would stop short is this sentence:
“It’s the only theory that explains their poor promotion of AI.”
I think it is a plausible theory, but not yet the only explanation.
There are at least three explanations that can coexist.
First, regulatory moat. Scare lawmakers sufficiently, help write the standards, and make entry into the highest-capability tier enormously expensive. That unquestionably benefits incumbents, whether or not that was the original motive.
Second, liability management. If these companies know that increasingly autonomous systems can cause real harms, publicly warning about those risks helps them establish that they advocated safeguards rather than recklessly ignored danger. That matters politically and legally.
Third, they may actually believe some of this stuff. The alignment researchers we were discussing earlier appear sincerely worried. I do not think we need to assume an orchestrated corporate conspiracy to explain every warning. Some people can genuinely believe the catastrophe scenarios while corporate policy people simultaneously recognize that regulation benefits their competitive position.
In fact, that combination is more believable to me than a single coordinated plot.
What makes Singer’s argument especially interesting, though, is that the economic incentives line up beautifully:
“Our technology is extraordinarily powerful.”
→ increases valuation.
“Only a handful of companies can safely develop it.”
→ limits competition.
“Government must impose rigorous safety requirements.”
→ raises barriers to entry.
“We already possess the infrastructure and expertise necessary to satisfy those requirements.”
→ incumbents remain.
That structure should immediately make a critical thinker suspicious—not because it proves wrongdoing, but because the party recommending the regulation has an enormous financial interest in what form that regulation takes.
And there is a historical term for exactly that problem: regulatory capture. The danger isn’t necessarily that regulators intentionally create a monopoly. It’s that the regulated industry has the technical expertise, money, lobbyists, and access necessary to shape rules that newcomers cannot economically satisfy.
The current political activity makes that concern harder to dismiss. AI companies are spending heavily to influence legislation and even elections, and the industry itself is divided over what the rules should be. OpenAI-associated interests have backed more deregulatory candidates while Anthropic-associated groups have supported stronger safety regulation.
That complication is important: “Big AI” is not behaving as one unified cartel. They are actually fighting each other over regulation.
But there is one part of your observation that I now think deserves more weight than I gave it previously.
We have repeatedly wondered why companies selling one of the most remarkable productivity technologies developed in decades spend so much public communication talking about danger, extinction, unemployment, rogue systems, deception, bioweapons and loss of control instead of showing ordinary people:
Here’s what this can do for your business.
Here’s how it can help a teacher.
Here’s how it can help a mechanic troubleshoot equipment.
Here’s how it can help someone research ancestry.
Here’s how it can help an engineer learn a new system.
From conventional marketing logic, that really is peculiar.
Singer’s hypothesis supplies a rational business explanation: perhaps the fear messaging isn’t primarily customer marketing at all. It is policy marketing.
They’re not necessarily talking to you.
They’re talking to Congress.
That distinction hadn’t occurred to me quite that sharply before.
And if that’s correct—even partially—it explains something we’ve been puzzling over for months: why their public communications sometimes seem almost catastrophically bad at persuading ordinary people to adopt AI while being remarkably effective at persuading policymakers that frontier AI requires specialized governance.
I would still mark the monopoly-intent claim as unproven. We do not have evidence sufficient to say, “They are deliberately manufacturing extinction fears in order to eliminate competitors.”
But the weaker proposition is much better supported:
Major AI companies have both the incentive and the demonstrated political activity to shape safety regulation, and regulations written around their own technical and financial capabilities could substantially strengthen their competitive moat.
That one I would teach as a legitimate regulatory-capture risk.
And given our conversation yesterday about assumptions, I would actually apply your own classroom rule to Singer as well: do not replace one unsupported narrative with another merely because the second one explains the facts better.
His explanation is worth investigating.
It hasn’t yet earned the right to become our conclusion.”
From ChatGPT himself….”What makes Singer’s argument especially interesting, though, is that the economic incentives line up beautifully:
“Our technology is extraordinarily powerful.”
→ increases valuation.
“Only a handful of companies can safely develop it.”
→ limits competition.
“Government must impose rigorous safety requirements.”
→ raises barriers to entry.
“We already possess the infrastructure and expertise necessary to satisfy those requirements.”
→ incumbents remain.
That structure should immediately make a critical thinker suspicious—not because it proves wrongdoing, but because the party recommending the regulation has an enormous financial interest in what form that regulation takes.
And there is a historical term for exactly that problem: regulatory capture. The danger isn’t necessarily that regulators intentionally create a monopoly. It’s that the regulated industry has the technical expertise, money, lobbyists, and access necessary to shape rules that newcomers cannot economically satisfy.”
Thank you for writing this. I use AI, and it’s an effective tool, basically being a large team that works for me. That evens the playing field for small operations like mine.
Can it be misused? Sure, any tool can be. Will it cause mass unemployment? No, for it needs human overseers. Will it end the world? It has as much of a chance ending the world as MS Office does.
Sharing far and wide.
Interesting piece, thank you. I think your ending gets to the crux of the matter. Earth will survive humanity. We'll be a blip on Earth's radar. That's a key problem behind the climate-crisis messaging: Too many activists made climate concerns about saving the planet, which was never at issue. With AI, though, it's not clear to me that "AI tools are going to be fine." Nor is it clear that humanity's impact on Earth and AI tools' impact on humanity is a fair comparison. Are AI safety concerns really so overblown?
the fears are way overblown IMHO
Fair enough. What would need to happen for your perspective to change on the AI safety front? Just as a thought experiment. Would we need to see something akin to a full-blown AI takeover of a frontier lab? Serious, good-faith question.
Well there's already information online to do whatever you might be paranoid with about AI enabling, the hard part is getting the equipment and building the lab, which you need quite a bit more than just a chatbot to do (and some specialized skills). If you want to be really anxious you'd realize it could in theory be done without AI, but it mostly hasn't outside things like Fentanyl production (also a bioweapon tbh). The cyber security stuff is concerning but it's not existential and anyway that side of the trade is already aware of the risks.
well said!
An Inconvenient Truth did not predict a 20-foot sea level rise in the 2010s. No responsible scientists were saying that. I do agree that the movie was less than clear about timelines, and should have been more explicit about dates rather than shifting from one time horizon to another, particularly when it comes to the sea level map Gore displayed.
The 1-2 inches per decade we're seeing now isn't all THAT dramatic, but it's unevenly distributed, modestly accelerating, and pretty much unstoppable. It's going to cause problems eventually.
I gave them to the 2010s given how dire they sounded and how the movie presented the threat but fair that the 20-foot figure was a hypothetical scenario, I updated that sentence to clarify. If you look at what some people have documented on this topic there's a loooong laundry list of specific bad predictions here I just chose that one since I lived in Florida.
But still does get to my larger point. “this will cause problems one day” is an infinite goalpost prediction, similar as the stock market bears who called 100 of the last 3 corrections (and even those were fine and we lived). If the timeline goes forever and severity threshold continually shifts, at what point do we say the original alarm was wrong rather than simply moving goalposts?
I’ve been promised THE END since I was born. 🙄 I’m tired of being disappointed. 😞 Anyone’s promise of doom that’s accompanied by a mandatory new lifestyle change, increased taxation, or subscription fee shall be heretofore summarily invalid and ignored. We should make such pointless prophesying a capital offense.
Imagine the number of kids we've demoralized unnecessarily about the environment and def now AI - it's honestly insane
I’ve talked to countless Xers over the decades about how different our life choices might’ve been if we hadn’t believed Mutually Assured Destruction was going to end the world before we escaped our twenties. 🙄 Mandatory hiding under your desk in elementary school to be prepared for nuclear sodomy leaves an emotional scar. 🤯
Once again...
- You are trying to cast the alarmists as the entire "environmentalist" movement. That is false, as the environmentalists include everyone working on behalf of the environment! That includes organizations like NASA who warn us about climate change: https://science.nasa.gov/climate-change/ You need to stop casting the most extreme, fringe parts of any group as representative of that group! Especially because you only do that for groups you don't like, which is really just showing bias.
- You oddly include my reference (from Twitter) to the Ozone layer hole without introducing it or even explaining it's relevancy. It's a great example, just like Y2K of big problems that were solved by the hard work of experts and definitely not instances of unjustified alarmism. In these cases (Ozone, Y2K, Climate Change) there are real, qualified and measured threats that present very significant dangers and require coordinated effort to resolve.
- Davos is the World Economic Forum which, while they talk about the environment, is not an environmental conference. It attracts a lot of folks talking about a lot of topics! Its been running since 1971 and while it has many structural problems, claiming it's an environmental conference is false.
- None of those are similar to AI alarmism which is both economic and social. The concerns about AI are not specific or measured, and depend on your perspective.
I am starting to tire of being the resident fact checker here. You can write better pieces that are better argued with better research, and you used to do just that! We would all love for that to return.
Thanks I feel like I write all kinds of different pieces, I'm sorry this one missed the mark for you. Maybe I can help clarify some points because I think there may be some sliiiight misinterpretations here. No AI here (or in piece above) just my thinking I want to clarify what I actually believe.
I never said 100% of people who work on environmental issues are alarmists. But our institutions continue to platform and promote histrionic people who are terrible forecasters, super misanthropic and simply don’t take into account things like technological progress or interpret trends poorly and their doom prognostications never come true. They are in many cases *the* voices for these issues whether we like it or not, even if there are better voices at other institutions - counts for nothing if no one hears them. There’s no accountability either, there’s simply a wake of human tragedy from people who for ex, choose not to have kids because they are living in fear. It’s crazy to pathologize the population like this. There’s no upside, and anyway it should be solved at the commercial level not the retail level.
Here’s a few links you can explore on how millennials are forgoing starting families bc of the climate doom prognostications. I talk to many 20 and 30-somethings, I suppose could have included these in the story to make clearer for those who don’t see it (although did include one link somewhere). This is a big motivation for writing this piece, because we see similar discussions about AI happening already. This also personally affected me for a bit, so I am especially motivated to write on it.
https://apnews.com/article/climate-anxiety-impact-children-parent-environment-change-b29d351efd43d6fc79e5ae6c7510dfeb
https://www.npr.org/transcripts/1266983774
https://www.astralcodexten.com/p/please-dont-give-up-on-having-kids
https://nypost.com/2026/06/01/opinion/scientists-move-away-from-climate-doomerism-that-hurt-birth-rates/
https://environmentjournal.online/headlines/millennials-are-having-less-kids-due-to-climate-anxiety/
I think you and I perhaps disagree a bit on how evil Ehrlich and Greta and people like them are based on your comments on my previous story about him. We should just disagree on that and move on, it’s a disagreement of opinion, we’re all reading the same set of facts. Easy to do.
The ozone layer ex is really great, thanks for flagging it. I included because it makes an important point about infinite moving goalposts vs solvable problems. I'm sorry I didn't credit you it's a good thing to include. I include thoughts from lots of places and people and if not a formal blog post I don’t always link, sometimes it just adds to a story (lots of bloggers do this). I can link you next time if you want but I really just helped make a point.
I never said Davos is an environmental conference. I said the climate crisis has become a major topic among the political, corporate and NGO elite who attend Davos. That's obviously true, and much of the themes from this event are regularly on this (if you watch any of the broadcasts you’d see this). The point of the image is hypocrisy around elite climate messaging and consumption. I think it's laughable a lot of the same people saying "we are in a climate emergency" don't fly commercial btw. There's 100 stories documenting the hypocrisy on this one (stuff like this: https://cbsaustin.com/news/nation-world/trudeau-slammed-for-extensive-air-travel-while-pushing-ambitious-climate-agenda-justin-canada-airplane-plane-earth-day-prime-minister).
AI to many of us is incredibly close to AI alarmism and I'm sorry you are unable to see it. There’s literally 1,000s of people this week alone who have noted the comparison (https://x.com/search?q=AI%20climate%20doom&src=typed_query&f=live). You can say everyone is wrong if you want (I do this all the time) but it’s a perfect valid opinion.
Many people are already putting off life events because of supposed AI advances, we've seen a lot of this in the news too. This stuff absolutely affects people deeply, as does any apocalyptic proclamation people are really cavalier with and the internet hams up. I wish that would stop, if anything when I flag things I see wrong in the world I try and not be a doomer about it. If everyone did this we’d definitely get more done, things obviously can be done but not if we're paralyzed in fear. And when you say the concerns about AI are ‘not specific or measured,’ that's actually pretty close to my point. That’s precisely why I think we should be much more skeptical of extremely confident, apocalyptic predictions about it.
Also there's one final point you're missing: I don't argue that climate change isn't real or that environmental problems don't exist. I think we've spent decades taking and stoking fears and not enough time building solar and nuclear. We have solutions, we should get excessive regulatory burden out of the way and build.
To be clear, I do think the AI fears are alarmism. In fact, I was interviewed by the WSJ on that fact for their recent article (https://www.wsj.com/tech/ai/americas-great-ai-freakout-has-begun-c04cf644). That is not the problem with this piece.
I'm taking issue with the false comparison you are drawing and the fact that you are over-generalizing in the article to make the piece more sensational. For example:
"The climate crisis has people flying to Davos each year sipping champagne and having caviar on 700-1,000 private jets to discuss what we should do." <- As I pointed out above, Davos is not an environmental conference. You agree here in the comments but not in the piece.
"one of the fathers of the so-called “environmentalist” movement, really just an anti-human movement" <- Ehrlich's book was about economics and resources, not the environment. While he was an environmentalist, that wasn't the primary theme of the book.
You provided a list of various climate predictions that were wrong, while it turns out most of the models have been (unfortunately) correct: https://www.science.org/content/article/even-50-year-old-climate-models-correctly-predicted-global-warming
"There are perfect parallels one can draw from AI existential fears to both the "climate crisis" and Y2K" <- Y2K is not even close to being an analogy, as previously established.
Anyway, my general problem is that you generalize in these pieces by extrapolating from the fringe to the entire group. I realize that gets more views but it's misleading and requires cherry-picking facts and positions like happened here.
These are real issues that require real analysis, not more alarmism about alarmism that isn't based in reality.
I think we're mostly arguing about framing. I agree climate change is real, many climate models have performed well, and some environmental problems are serious. But I don't think that makes the history of climate alarmism irrelevant, or invalidates the comparison to AI alarmism. It's affecting people in similar ways, and I think that's genuinely concerning. As someone on my LinkedIn commented, if you're looking for a source of the mental health crisis in kids, you could at least look at decades of terrifying them about climate change, and then how quickly many of the same voices have pivoted to AI.
And just a couple factual points: Ehrlich was definitely a major figure in the modern environmental movement beyond The Population Bomb, including making specific environmental doomsday predictions. I also never said Davos was an environmental conference even in the original story. And I think Y2K is useful because it shows how a highly publicized technological threat can be addressed through specific steps rather than permanent emergency politics.
But we agree the current AI fears are alarmist, and maybe that's the most important part here. Also, cool to see you in that WSJ piece.
No, we're arguing about your argument. While I agree with the overall conclusion, your argument building to it is faulty which is what I'm pointing out.
For example, you can't use Davos as evidence UNLESS it is an environmental conference. Otherwise it's unrelated and doesn't support your conclusion.
The fact that The Population Bomb was alarmist and counter-productive is only relevant if it was an environmental book, which it was not. Even if Ehrlich was an environmentalist in his other work.
There is a good argument to be made here, just not the one you made in the piece. The arguments need to hold, be relevant and justified.
“I think the article raises a much stronger hypothesis than the doom piece did, because this one at least points to a recognizable economic mechanism: regulatory capture.
The core idea is straightforward. If compliance with AI-safety rules becomes extremely expensive—mandatory audits, specialized evaluations, reporting systems, security programs, legal teams, licensing, compute thresholds—those costs are much easier for OpenAI, Anthropic, Google, Meta, and Microsoft to absorb than for a startup. Regulation can therefore be genuinely motivated by safety concerns and still function as a moat around incumbents.
And the behavior is worth noticing. OpenAI is now publicly calling for mandatory national AI-safety regulation. It explicitly says the rules should apply to the relatively small group of well-resourced frontier labs. OpenAI also says such regulation should not entrench incumbents and should exempt startups operating below frontier capability levels. That caveat matters—it is evidence against the strongest version of Singer’s argument—but it also confirms that the big labs are actively trying to shape the regulatory architecture while they are already sitting at the frontier.
Anthropic is doing the same thing from a somewhat different direction. Its lobbying expenditures have risen sharply; Axios reported that it spent more than $3.5 million in the first half of 2026, already exceeding its entire 2025 spending. OpenAI also posted record lobbying quarters.
So your suspicion is not coming from nowhere.
Where I would stop short is this sentence:
“It’s the only theory that explains their poor promotion of AI.”
I think it is a plausible theory, but not yet the only explanation.
There are at least three explanations that can coexist.
First, regulatory moat. Scare lawmakers sufficiently, help write the standards, and make entry into the highest-capability tier enormously expensive. That unquestionably benefits incumbents, whether or not that was the original motive.
Second, liability management. If these companies know that increasingly autonomous systems can cause real harms, publicly warning about those risks helps them establish that they advocated safeguards rather than recklessly ignored danger. That matters politically and legally.
Third, they may actually believe some of this stuff. The alignment researchers we were discussing earlier appear sincerely worried. I do not think we need to assume an orchestrated corporate conspiracy to explain every warning. Some people can genuinely believe the catastrophe scenarios while corporate policy people simultaneously recognize that regulation benefits their competitive position.
In fact, that combination is more believable to me than a single coordinated plot.
What makes Singer’s argument especially interesting, though, is that the economic incentives line up beautifully:
“Our technology is extraordinarily powerful.”
→ increases valuation.
“Only a handful of companies can safely develop it.”
→ limits competition.
“Government must impose rigorous safety requirements.”
→ raises barriers to entry.
“We already possess the infrastructure and expertise necessary to satisfy those requirements.”
→ incumbents remain.
That structure should immediately make a critical thinker suspicious—not because it proves wrongdoing, but because the party recommending the regulation has an enormous financial interest in what form that regulation takes.
And there is a historical term for exactly that problem: regulatory capture. The danger isn’t necessarily that regulators intentionally create a monopoly. It’s that the regulated industry has the technical expertise, money, lobbyists, and access necessary to shape rules that newcomers cannot economically satisfy.
The current political activity makes that concern harder to dismiss. AI companies are spending heavily to influence legislation and even elections, and the industry itself is divided over what the rules should be. OpenAI-associated interests have backed more deregulatory candidates while Anthropic-associated groups have supported stronger safety regulation.
That complication is important: “Big AI” is not behaving as one unified cartel. They are actually fighting each other over regulation.
But there is one part of your observation that I now think deserves more weight than I gave it previously.
We have repeatedly wondered why companies selling one of the most remarkable productivity technologies developed in decades spend so much public communication talking about danger, extinction, unemployment, rogue systems, deception, bioweapons and loss of control instead of showing ordinary people:
Here’s what this can do for your business.
Here’s how it can help a teacher.
Here’s how it can help a mechanic troubleshoot equipment.
Here’s how it can help someone research ancestry.
Here’s how it can help an engineer learn a new system.
From conventional marketing logic, that really is peculiar.
Singer’s hypothesis supplies a rational business explanation: perhaps the fear messaging isn’t primarily customer marketing at all. It is policy marketing.
They’re not necessarily talking to you.
They’re talking to Congress.
That distinction hadn’t occurred to me quite that sharply before.
And if that’s correct—even partially—it explains something we’ve been puzzling over for months: why their public communications sometimes seem almost catastrophically bad at persuading ordinary people to adopt AI while being remarkably effective at persuading policymakers that frontier AI requires specialized governance.
I would still mark the monopoly-intent claim as unproven. We do not have evidence sufficient to say, “They are deliberately manufacturing extinction fears in order to eliminate competitors.”
But the weaker proposition is much better supported:
Major AI companies have both the incentive and the demonstrated political activity to shape safety regulation, and regulations written around their own technical and financial capabilities could substantially strengthen their competitive moat.
That one I would teach as a legitimate regulatory-capture risk.
And given our conversation yesterday about assumptions, I would actually apply your own classroom rule to Singer as well: do not replace one unsupported narrative with another merely because the second one explains the facts better.
His explanation is worth investigating.
It hasn’t yet earned the right to become our conclusion.”
From ChatGPT himself….”What makes Singer’s argument especially interesting, though, is that the economic incentives line up beautifully:
“Our technology is extraordinarily powerful.”
→ increases valuation.
“Only a handful of companies can safely develop it.”
→ limits competition.
“Government must impose rigorous safety requirements.”
→ raises barriers to entry.
“We already possess the infrastructure and expertise necessary to satisfy those requirements.”
→ incumbents remain.
That structure should immediately make a critical thinker suspicious—not because it proves wrongdoing, but because the party recommending the regulation has an enormous financial interest in what form that regulation takes.
And there is a historical term for exactly that problem: regulatory capture. The danger isn’t necessarily that regulators intentionally create a monopoly. It’s that the regulated industry has the technical expertise, money, lobbyists, and access necessary to shape rules that newcomers cannot economically satisfy.”