Never trust a T-Rex
In 2024, at the same time as I was being called a genocide apologist, Zionist baby killer, etc. etc., I was also being hounded by my right-wing, pro-Israel readers, who demanded of me: knowing what you know, understanding what you understand, how could you possibly vote for Kamala Harris? How could you donate to Kamala’s campaign, urge your readers to vote for her, when she (like most Democrats) is obviously beholden to young left-wing activists, and young left-wing activists’ central unifying cause has become the death of “Zionists” like yourself and your children? Can’t you see that, even if Trump is a raging, lying, bullying, incoherent monster, at least he’s our monster?
This view, I confess, gave me more pause than “just accept that the liberation of the world’s oppressed requires Israel’s annihilation and hence the death of much of your family,” which my far-left critics considered their most persuasive argument. And yet I also rejected, with extreme prejudice, the right-wingers’ constant invitations to join their MAGA brigade. I replied: I’ll simply continue being a moderate Enlightenment liberal, transported here in a block of ice from the 1990s, even if I’m the last such on earth, even if I’m condemned by everyone on either side for it.
Why? Because, I said at the time, while I’m honest enough to admit when a rampaging T-Rex happens to do something that aligns with my interests—when, e.g., it chases away some velociraptors ready to slice me open, or cracks down on antisemitic fanatics trying to dominate the universities where I spend my life—still, I’ll never delude myself into imagining the T-Rex my ally. I understand that it would just as soon devour me. If the monster goes after my enemies not because of liberal principles or recognizable moral emotions, but just raw self-interest and ego gratification, then what happens the instant its perceived self-interest changes?
It gives me no pleasure that this week Trump proved my instincts here 3,000% correct, by fully capitulating to the Islamic Republic of Iran, far more so than Barack Obama ever did, or than President Kamala Harris ever would have. Trump has now abandoned both the Iranian and the Israeli peoples to suffer and die at the hands of the IRGC, as he abandoned the Ukrainians and the Venezuelans before them, as Neville Chamberlain abandoned Czechoslovakia.
On the most charitable reading, Trump gambled that taking out Ayatollah Khamenei would lead to a cowed and compliant puppet regime, as basically happened in Venezuela, with no need for a ground invasion or arming the Iranian resistance. His gamble predictably failed, because (alas) the Iranian government actually believes in something—horrifying and medieval though it is. Trump was unable to process that fact, because he’s never believed in anything beyond himself, and he wrongly assumes that everyone else is the same.
With the $300 billion and control over the Strait of Hormuz, I expect that Iran will rebuild its military and proxies to stronger than they were before Trump’s idiotically mismanaged war. I expect that Iran will then launch attacks against Israel that make October 7 look like the Little League—and that, when it does so, a large fraction of the Western world will ecstatically cheer, as it did on October 7 itself. Netanyahu is a fool for expecting otherwise from Trump, a man who’s betrayed everyone who’s ever trusted him outside possibly his immediate family.
I salute those Israelis who will choose to stay and fight even after their abandonment and betrayal, in the spirit of the Bar-Kochba revolt and other desperate battles of Jewish history. Despite the existential danger that Israel will soon be in, facing a victorious and emboldened Iran essentially alone, I also see it as possible that Western countries will rapidly become even more dangerous for Jews than Israel. If that happens, I’ll be grateful that Israel still exists, that it considers itself unbound by America’s surrender, and that I’ll be able to seek refuge there, as was the idea when Israel was founded.
At the same time, I wouldn’t begrudge any Israelis who moved to the US, or Switzerland, or whatever other country will take them. As in the 1930s and at countless other points in Jewish history, the priority now is physical survival, wherever that turns out to be possible.
With hindsight, I spent the first half of my life in a strange interregnum, wherein Jewish history seemed to have finally ended. Now, fueled by fading memory of the Holocaust and by the greatest lie-amplification technologies the world has ever seen, the history I learned as a child has come roaring back. Jews, as they have for millennia, look out on a world of murderous enemies and fickle friends. It’s just the restoration of a norm.
Incredibly, abandoning both the Iranian and Israeli peoples to the Revolutionary Guard might not have been the most shortsighted or catastrophic thing Trump did in the last couple weeks. Another candidate would have to be Trump’s attempt to destroy Anthropic (and as collateral damage, American AI development more generally), transparently to punish Anthropic for the crime of having any principles that it was willing to put ahead of obedience to Trump and Pete Hegseth. Specifically, the White House forced Anthropic to pull Fable, its new flagship model (and the “safe” version of Mythos), off the market just a few days after customers had started using it, by subjecting it to export controls that it knew were impossible to enforce. Even the many foreign nationals who work at Anthropic are no longer allowed to use their own model (!). A plausible consequence is that those foreign nationals will stop working at American AI companies altogether, and will move to China or whichever other country rolls out the red carpet for them.
For AI accelerationists, you’d think that this would be a worst-case scenario: a direct government crackdown on AI that goes beyond anything the AI safetyists had proposed, and that indeed would’ve sounded like fantasy even a year ago. And yet many of the accelerationists are gleeful. Why? Because Anthropic, in supporting reasonable AI regulations, had made itself the accelerationists’ enemy. So, the accelerationists’ attitude now is quintessentially Trumpian: “Haha, Anthropic, you say you like regulation? Then take that!” Never mind that whatever dangerous behavior can be elicited from Fable, can almost certainly be elicited as well from GPT 5.5 Pro, and yet there’s no talk of any similar crackdown against OpenAI. Sam Altman, after all, donated $1 million to Trump’s inaugural fund. No one finds it remarkable anymore, in Trump’s destroyed and recreated United States, that your rights depend entirely on your standing with the Don.
And so, just like the question of whether Trump would side with the isolationists or with the hawks who wanted to liberate Iran, was resolved by his worst-of-all-worlds choice to surrender to Iran, so too the question of whether Trump would hit the brakes on the race to dangerous AI, or accelerate in order to beat China, has been resolved by his worst-of-all-worlds choice to lose the race to China. I.e., we’re still in full race mode, but we’re also going to do whatever we can to lose the race—by, for example, letting NVIDIA sell its chips to China, and now, scaring away our top researchers and punishing our AI firms with capricious and arbitrary crackdowns.
It’s disconcerting to reflect that, while the prognosis of the world is arguably the worst it’s ever been in my lifetime, my own life is pleasant. Intellectually I know that the Titanic has already hit the iceberg, but the band is still playing and I’m still being served delicious food.
Last week I visited Paris and the French countryside with my wife and kids. In addition to sightseeing, I spoke at a workshop celebrating 50 years of Aumann’s Agreement Theorem (where I got to meet the 96-year-old Aumann), and gave a quantum computing talk at the Sorbonne. Next week I’m going with my family to a science camp in California, then to STOC in Salt Lake City, where I’ll accept the Trevisan Prize and give an after-dinner speech, then to Epsilon Camp where I’ll again teach theoretical computer science to 11- and 12-year-olds.
Like I said, life is good here on the Titanic, if you ignore the rapidly rising seawater at your feet.