This is an update to my post from Tuesday about the “existential risk” panic. A good friend of mine reached out to me and said, in so many words, yes, all that techno-babble is weird. But that doesn’t mean these risks aren’t real. And as I tried to make clear in my piece, I agree with that, at least to the extent that if Group A isn’t made up of reliable narrators, I’m not clear who the reliable narrators are. He pointed me to this recent report put out by the Rand Corporation, which is kind of the original think tank and has always been closely tied to the U.S. government and for lack of a better word the U.S. military-intelligence-industrial complex. It’s an interesting report. Or at least I found some interesting and I think sensible prescriptions in it once I got through the first half or two-thirds of it, which contained very Randian decision trees, definitions and a lot of what seemed like assumptions about existential risk and coexistence with AI that I didn’t understand. It eventually sets out a sensible if hard-to-execute approach which generally amounts to a safety-first regime of governing AI (make sure new things don’t endanger us before we build them or make them widely available) and, critically, open the decision-making up to public institutions with public accountability.
This is the weekly, free edition of The Backchannel. If you’d like to receive the free edition of The Backchannel in your inbox, you can subscribe here.
Donald Trump’s manic building and naming spree has been a core feature of his second term in office. And that’s saying something. Because if nothing else, you can’t say Trump hasn’t been active. There are wars, blowing up the international trade regime, the on-going attempted authoritarian takeover of the state, the bust-out rush to wrest as much money out of the presidency as he can, his significant role energizing the AI boom (and possible bust). But with all these things, the building and naming spree stands out, especially in 2026. I focus on this today. I return to this today because of news that Trump wants to turn his West Palm Beach golf club into a new Camp David, a permanent presidential retreat.
The details here are important. If Trump had some secluded property he wanted to donate to the federal government for this purpose, there might be some logic, though it’s hard to imagine Democrats wouldn’t put it up for sale at their first opportunity. What he wants to do is different. His plan is to donate a parcel of land at the private club for the Florida Camp David. So it’s really that future presidents would have there own private bungalow at Trump International Golf Club in Palm Beach. As an additional, rather comical, sweetener the club’s legal counsel, who is presenting the idea to Palm Beach County officials, says that future presidents will even be offered comped memberships to the club.
For years, and especially in the last month or so, we’ve been collectively inundated with an AI industry which, uncannily and bizarrely, tells us simultaneously that AI is essential, inevitable and runs a significant chance of exterminating all of us in as few as two or three years. If you’ve dipped into this debate at all, you’ve quickly come upon discussions of so-called “rationalists” or “long-termists” or supporters of “effective altruism” who are simultaneously ubiquitous among those developing AI and in the AI safety (i.e., the folks sounding the alarm) community as well. (For the sake of simplicity and brevity, I’ll refer to these folks as the “REL nexus”: rationalists, effective altruists and long-termists. This is my coinage since I don’t know a better one.) I’ve spent a fair amount of time familiarizing myself with this world and how it relates to current debates about AI. So I wanted to share my basic impressions.
It’s hard to know what to say about former Republican Speaker of the House Denny Hastert, whose death at the age of 84 was reported this morning. Few if any political leaders in modern American history have experienced a fall quite so ignominious and total. The obvious point of comparison is Richard Nixon. But post-Watergate Nixon was never without his defenders and supporters. All but his most adamant critics granted significant successes of his presidency. And by the time of his death in 1994, he’d made significant progress in being treated by many as a sage elder statesman, albeit a greatly tarnished one. Hastert remains almost a test case of a top political leader whose fall was immediate and absolute. He came as close as one can come to have to having zero public defenders (though a significant number of former colleagues wrote leniency letters on his behalf at sentencing in 2016).
What stands out in my mind are two moments, one from his moment of ascent to power and another at the denouement but certainly not the nadir of career. Go back to 1998. The so-called Lewinsky scandal had been trundling forward for most of the year. Republicans are giddy about the Red Tsunami that was about to supercharge the existing GOP majority. Predictions of a 30- or 40-seat pick up were common and assumed. (I was one of few people I knew to basically predict no wave at all; it was simply a matter of watching the polls rather than the bipartisan hype.) When Democrats ended up gaining five seats, House Speaker New Gingrich was summarily tossed overboard — a mix of the failure of his wild overpromising and slowly building frustrations with his erratic and high drama leadership style.
We’ve been following this conversation here in The Editors’ Blog about what AI is, what it can do and whether it can possibly live up to its hype. It’s worth considering the question from what we might loosely call not only the supply side but also the demand side. From a very broad perspective, the history of the last three decades has been one defined by the fact that you have too much capital chasing too few productive investments. That’s the driver of the recent history of cycles of booms, bubbles, over-investment and busts: the Asian Financial Crisis, the Dot Com Bust, the Global Financial Crisis, perhaps soon the AI Bubble Collapse, along with many smaller of regional ones.
A friend sent me this new Bain Capital study on the growth required to match the current levels of capital expenditure on AI, essentially breaking down the cost of the data center and other associated build outs. It’s an interesting document, not simply for the number-crunching and the predictions, but as a document in the more literary and analytic sense. Needless to say, Bain isn’t coming at this from any “boo capitalism!”/”this is insane lol” viewpoint. But the conclusions aren’t that far from the “this is insane lol” position. It’s very much, “hey! we can do this” but also, “um … wow, there’s quite a lot to do.”
The broad stroke numbers are these. The study says that to keep up with the current levels of capital expenditures AI will need to become a $6 trillion dollar market by 2031. (To give some perspective on that number, a recent Gartner study predicted that global IT spending will be just under $6.4 trillion this year. So devices, hardware, software, services, kind of everything.) It then posits that we can see the beginnings of, with some level of out-year predictability, something on the order of a $1.5 trillion market — which would be made up of a greatly expanded consumer market (Claude apps, ChatGPT, etc.) of between $200 and $400 billion and then an enterprise market (software, sales, marketing, business optimization) of between $1 and $1.4 trillion. So we’re left with about $4.5 trillion to go. That’s where the crash course in innovation comes in.
This is the weekly, free edition of The Backchannel. If you’d like to receive the free edition of The Backchannel in your inbox, you can subscribe here.
In my recent writing about LLMs and AI, I’ve mainly focused on what it can do, what it is, what its potential dangers are. I haven’t focused as much on the pretty substantial evidence that we’re in the midst of an AI bubble. The entire U.S. economy is heavily dependent on the AI boom. Much of the rest of the economy is in a slump. That boom is based on cheap money and very high expectations for AI profits, just as the Fed is facing irresistible pressure to raise interest rates, which is to say, raise the cost of money. This whole question of an AI bubble gets talked about a lot. You probably know the basic outlines. What’s caught my attention is how just in the last week or so these questions, verging into assumptions about unsustainable spending, are bleeding into the tech and the Beltway political press.
As I noted a week ago, we can’t know whether the current polls will be predictive of the November election results. But the polls themselves are at least speaking with a clear voice. I want to note two things that appear to be moving in unison: consumer confidence and the Republican position on the national generic ballot.
Here is a short update on the so-called Broadview Six case — one of the high-profile prosecutorial abuse cases coming out of ICE/CBP Midway Blitz occupation of Chicago in the fall and winter of 2025-26. To catch everyone back up, this was a case involving four candidates or elected officials, one campaign staffer and one activist. It was tied to ongoing protests outside the ICE Broadview facility outside Chicago. Federal prosecutors took a situation that involved protestors briefly making contact with a vehicle unexpectedly trying to drive through a protest line and jacked it up to a felony case. As one of the lawyers on the case told me at the time, the facts supported at best a misdemeanor disorderly conduct charge, and even that would been a wild overreaction to what actually happened.
Throughout the case, it was clear that the case was part of the Trump administration’s larger strategy of targeting peaceful protestors with wildly excessive charges. Eventually the case fell apart when major prosecutorial misconduct was uncovered on the part of the prosecutors in the case. All charges were eventually dropped in May with prejudice, just before the trial was to take place. The remaining mystery was tied to the original prosecutor in the case, who committed the misconduct securing the indictment the previous fall. (There was other misconduct tied to turning off doctored transcripts of that grand jury session.) That prosecutor, Sheri Mecklenburg, had a generally apolitical reputation and certainly was no right-winger. Defense attorneys had the impression she didn’t have a lot of confidence in the government’s position. She left the case with little explanation for a DOJ assignment working on the Senate Judiciary Committee under Sen. Dick Durbin (D-IL) in February. Meanwhile, the U.S. Attorney for the Northern District of Illinois — Andrew Boutros — and Mecklenburg’s replacement — William Hogan — were both known to be strongly onboard with the Trump/ICE program. (Hogan also had a tattered reputation on the professional ethics front.)
This is the weekly, free edition of The Backchannel. If you’d like to receive the free edition of The Backchannel in your inbox, you can subscribe here. If you’d like to upgrade to the members’ edition for full-access to The Backchannel and everything else we publish at TPM, this is the final week we are offering a special promotion for 50% off an annual Prime membership.
This won’t come as huge news to most. But I’m getting asked about it a lot: can the polls be wrong? What do you think?
So let me put an answer here. They can obviously be wrong and you should be prepared for them to be wrong. If you’re running a campaign or SuperPAC, they can give you intel about where to deploy resources. If you’re just observing, they’re more about human nature and anxiety management. But whether you get excited or pushed into despair, we’ll know the actual results in about six weeks. Watch them or don’t. It’ll will be fine either way. There are no poll hubris gods. They can’t strike you down.
Let’s talk specifics.
The polls we’re seeing now are almost certainly directionally accurate. But they could be overshooting on their margins, which are quite large now. If they are we should expect to see that most in red states. (Even for someone like me who’s been bullish about this midterm cycle since mid-2025, some of these numbers are hard to believe.) States have partisan muscle memory that polls can miss. So for instance, yesterday Sabato’s Crystal Ball changed the Kansas Senate race from Likely Republican to Leans Republican, an astonishing development. The polls there are pretty close given that it’s Kansas, and the most recent, an Emerson poll, has Democratic challenger Adam Hamilton up by 1 point. But this is a classic case where you have to weigh a substantial amount of GOP muscle memory into the equation. Polls can miss that, or it can simply mean that undecideds break heavily for the state’s dominant party. To be clear, if I were a Kansas Democrat canvassing I’d be pretty pumped right now. A win is possible. But this is the cautious/realistic way to look at it. The same applies to the handful of other seemingly competitive races in deep red states.
Daniel Drezner says we are now in the Wile E. Coyote phase of the 2026 midterm, the phase of six weeks or so in which everyone knows the GOP is toast but we pretend we don’t. Some of this is simply the political press’s collective suspension of disbelief. It’s neither fair nor fun to act as though the outcome is mostly known. Nor is this new. G. Elliott Morris notes that election odds makers consistently overstate the odds of the president’s party in the midterms, which is sort of another way of saying the same thing since the president’s party is usually on the losing end of the proposition. I would argue that a significant amount of this remains the structural, though not necessarily ideological, bias in favor of the Republican party — two decades on, Washington, D.C. remains wired for the GOP. But what all of this really reminds me of is the aftermath of the 2024 presidential election and what was then referred to as the “vibe shift” — both the hidden reality Donald Trump’s narrow but clear victory had purportedly revealed but also the shift in attitudes about cultural power and the future driven by that victory.
In elite circles the argument and belief went something like this. We thought that MAGA was a movement of reaction fueled by those left behind or discarded by the forward march of history and culture trying to wrench society back to an imagined 1950s. Not at all. In fact, it was the movement of the young and the most vital elements of the culture and the future. It was even expanding in the non-white constituencies which were supposed to be the anchors of the political left, the communities which were supposedly the targets and victims of MAGA, not its latest converts. Most of all, MAGA now had cultural sway to match its electoral power. And both were growing.
One of the key arguments for a great power — other than the great power itself wanting to be one — is that they act as a guarantor of stability and the status quo. That is far from the only public or global good. And a great power will ensure that status quo on its own terms and mostly to its own advantage. But if you have any acquaintance with history, you know you can do a lot worse than relative peace and the status quo. Trump’s Iran War has been a bracing illustration of all these realities. Once the great power starts letting things break, those breakages can start multiplying very fast. Eventually you’ll find a new equilibrium, of course. But eventually can take a long time. And again, even a passing acquaintance with history tells us that things can suck without end for extended periods of time.
There’s an interesting New York Times piece from over the weekend that illustrates all these things. We know the outline. But it provides new details. As a knock-on effect of the U.S. war with Iran, the ongoing battle between Saudi Arabia and the Houthi rebel government in Yemen has again become hot. In recent weeks, the Houthis have partly cut off Saudi Arabia’s other path to the ocean. Put simply, the Saudis can export oil through the Strait of Hormuz or the Red Sea. The former has been mostly closed for months. Now the Houthis are in de facto or near control of the second. That’s a big crisis for the Saudis. This is part of the larger U.S.-Iran battle because the Houthis are clients of Iran. And Iran’s core strategy has been to create enough pain for the U.S.’s regional allies that they will force the U.S. to end the war. Remember, each time the U.S. attacks Iran, they respond by attacking the various Gulf emirates.
I want to flag this poll for you. It’s a poll of the Florida Senate race by St. Pete Polls, a mid-tier regional pollster. It shows appointed incumbent Senator Ashley Moody (R) and Democratic nominee Angie Nixon tied at roughly 45% of the vote. There’s not much polling of this race. There are two other recent polls, one by Change Research a respected but generally Democratic-aligned pollster, and another from Quantus Insights, a GOP-aligned pollster with an iffy reputation. Change Research had a tied race a week ago. Quantus had Moody up by eight points a bit less than a week ago.
I want to be clear that I am not predicting or expecting Nixon to win this race. What I do expect is that Republicans will likely need to fight to hold it. And that is significant in itself. There are other signs that Republicans are facing a rough political environment in Florida this year, at least one that looks very different from where Florida has been trending since 2018-20 when the state went from a GOP-tilting purple state to something approximating bright red. Yesterday, Nicole LaFond highlighted the video put out by South Florida Rep. María Elvira Salazar (R-FL) which basically had Salazar telling Donald Trump: We didn’t know you meant mass deportation when you promised us mass deportation.
We can joke about the the substance of Salazar’s video but it speaks for itself as a measure of the political climate in the state.
As I’ve suggested in recent posts I’m finding it quite challenging to make sense of this roiling debate about AI safety. When I write about American politics, American history and a few other topics I have a body of knowledge that I draw on and try to share with readers — explain how it might illuminate present-day questions. I may be right or wrong. But I’m speaking from a background of some real knowledge and study. When I delve into questions like AI or AI dangers I’m seldom trying to come up with definitive answers. I lack the qualifications or foundational knowledge. What I’m trying to do is first try to educate myself, but in my writing I at least try to frame the basic questions, provide a framework for considering the issue, point to what are at least good faith and knowledgable voices if not necessarily people who are “right.” But while I’m still uncertain about dangers posed by the companies doing the most advanced and often reckless research on artificial intelligence I’ve been increasingly stuck on the prominence in the AI world of people who are in the overlapping communities of “effective altruism,” “rationalism” and “longtermism”.
Just to point to the latest examples, the Journal has a write-up on Jacob Coxon, the AI researcher whose Twitter thread quite literally triggered the current freakout/reckoning. It turns out his formative years were spent in the various parts of the “rationalist” subculture in London before he migrated to the U.S.
I’ve been loosely aware of the first and the second for a while, but didn’t have much pressing need to learn more because they didn’t seem connected to much that I care about. But they’re very big in the AI world and maybe even bigger in the AI will wipe out humanity worlds. So given the global freakout getting a handle on this becomes more pressing. Just as an aside, I will note that I don’t know how the dorks operating under the “rationalism” label secured exclusive branding rights to that term. Like, rough day for Descartes and quite a few other philosophers. But I digress. What I’m going to argue here is that at least the popularized versions of what I’ll call ERL (“effective altruism”, “rationalism” and “longtermism”) is made up of a mix of absurd and nonsensical ideas. The fact that so many AI doomers are part of that world gives me a lot of pause. But we can’t go from ERL is absurd to AI is fine.
This is the weekly, free edition of The Backchannel. If you’d like to receive the free edition of The Backchannel in your inbox, you can subscribe here. If you’d like to upgrade to the members’ edition for full-access to The Backchannel and everything else we publish at TPM, this week we are offering a special promotion for 50% off an annual Prime membership.
I just noticed this Politico piece which contains a poll they just conducted about AI and the odds that AI will lead to the extinction of humanity. 63% of Americans think there is at least a “moderate risk” that AI means our species is toast. 20% see a “significant risk” and an another 17% think our demise is “almost certain.” As with data centers, there’s little difference between Republicans and Democrats.
Needless to say, this is not an issue on which we want to assume that majority opinion is a good proxy for reality. But it is a potent reminder of the almost unfathomable depth of negative opinion about AI, ranging from hate to fear and basically all the other negative emotions in between. Meanwhile, this question seems pretty distinct from the even more overwhelming opposition to data centers. They’re not two expressions of the same animus but two distinct channels of opposition directed at the same thing.
Conventional wisdom holds that a midterm election only fully comes into focus after Labor Day. That’s when a critical slice of the electorate starts focusing on the election and the choices before them. More or less in line with that, we’ve had a spray of polls out in the last few days showing the congressional generic ballot shifting decisively in Democrats’ favor. Looking only at generic ballots fielded entirely in September by pollsters the New York Times rates as trusted we have D+7, D+9, D+9 and D+11. If that trend or average is representative, it shows Democrats opening a big lead on this measure where the earlier average had been in the 5-7 point range. As Ron Brownstein notes here, history also tells us that in the final weeks of a campaign, this generic ballot measure will bend toward the president’s approval number. That appears to be happening. One of those, from the Times itself, was released today (D+9) with an accompanying write-up which, simply because it’s the Times, will probably draw attention to this trend even though the Times poll is only one of several.
Polls and trends can of course be wrong. But these numbers are extremely bad for the GOP, and if anything in this ballpark is accurate Republicans will likely lose the Senate as well as the House. Indeed, the variability in possible outcomes may be even higher in the Senate since it’s not amenable to gerrymandering. Even the Times, usually eager to find possible downsides for Democrats, frames their poll with declarative statements like this: “Voters see Republicans broadly as more extreme, more elitist and more corrupt than Democrats, according to the poll. And while voters continue to have doubts about both parties, they aligned more with the Democratic Party on both cultural and economic matters than the Republican Party in the survey.”
As you’ve likely seen, a strange and perhaps positive chain of events unfolded over the weekend in the world of AI and what we might call human extinction discourse. Dario Amodei, CEO and cofounder of Anthropic, posted an essay calling for a global slow-down in AI development to allow more time to prioritize safety and what the industry calls “alignment” and pledging that Anthropic would take a series of steps in that direction unilaterally. Somewhat surprisingly, OpenAI CEO Sam Altman and archvillain Elon Musk stepped forward and said they agreed and appeared to agree to join Anthropic in the proposed slowdown.
Regardless of what one thinks of AI and its possible dangers — existential or otherwise — it’s hard not to see this as at least a somewhat good development. And I say that with full cognizance of the various arguments about financial overextension or heading off sterner regulation that may make these moves self-interested. (AI critic Gary Marcus gives it a qualified endorsement along with a good cross-section of responses.) But I wanted to step back for a moment and talk more generally about the question of “externalities” which has always been the problem lurking in the background of Big Tech’s road to national and global domination.
Login
