This article is the second of two parts. The first part can be read here.
A Walk on the AI Side
Let’s query the oracle of AI on the Crash and the Depression. This author knows that those skilled at “prompting” can often inveigle AIs into embarrassing, even egregious, answers. That was not my intention here: I simply wanted to find out, and then document, how Google AI answered straight-forward questions.
Bearing in mind that AI is so algorithmic, others may get different answers. So that’s why I have preserved mine, creating 19 slides, maintained here. The questions and answers can be seen by following the links for each slide. As an aid to the reader, I have emphasized some text by adding red underlines and arrows. (The blue highlighting comes from the AI itself.)
On Slide 1, I ask simply: “What caused the Depression?” Google’s answer, carrying on to Slide 2, provides a familiar liberal-left catechism. For instance, it cites “overproduction” as a factor. To a free marketeer, overproduction is a market signal: Customers aren’t buying enough, telling the factory or industry to make less. It’s a mere aspect of the business cycle. However, on the Marxist left, overproduction is a sign that capitalism is in a death spiral. The Keynesian left sits somewhat near the Marxists, viewing overproduction as a crisis. And, as we see, Google sides with them.
The same slide lists “unequal distribution of wealth” as a causative factor. That, too, has been a steady concern of the left, both Marxist and Keynesian, independent of actual economic conditions. Yet not everyone agrees that the existence of rich people makes the economy more unstable. They believe, in fact, that wealth is consonant with overall prosperity.
Yes, bubbles are an issue. And yet they’ve been a recurring phenomenon all through economic history. Painful as they might be, popped bubbles are not a systemic problem, unless other factors compound them, as happened in the 1930s.
So a cynic, on the hunt for ulterior motives, might look at Google’s answers and see a victory for the left, insinuating Marxist/Keynesian dogmas into its diagnosis of the Depression. We need government policies to manage production! We need high taxes on the rich to prevent a crash!
Yet interestingly, in this answer, Google makes no mention of the huge tax increases that really did happen, tax increases that were, in fact, targeted at the rich. Many of us believe that these tax hikes greatly worsened the economic crisis.
The history of the Revenue Act of 1932—pushed for, and signed into law, by Republican president Herbert Hoover—is little known. And Google, the self-declared repository of the world’s information, does nothing, in its summary, to illuminate these shadowed events.
Yet at the time, the debate was bright. Many important political figures warned against a tax hike, including Hoover’s fellow Republican, Sen. Roscoe Patterson of Missouri. At the beginning of 1932, the fourth year of social woe, he inserted into the Congressional Record a memorandum from the St. Louis Chamber of Commerce declaring, “The impending specter of higher taxes constitutes one of the chief deterrents of business recovery.”
Yet Hoover, focused on balancing the budget, succeeded in more than doubling the personal income tax, raising the top rate from 24 percent to 63 percent. Rutgers tax historian Sidney Ratner quotes the Atlanta Constitution ripping Hoover’s handiwork as the “the most vicious tax bill ever sought to be saddled on the country in time of peace.”
Most conservatives probably think that giant tax increases in the middle of a downtown can turn a recession into a depression. Others might disagree, and that’s okay. But it’s not okay that Google doesn’t mention, at all, the tax variable in its summary.
Okay, so now the question that goes to the heart of Google’s ideology: Slide 3: “Did laissez-faire policies contribute to the Great Depression?” And the AI’s answer started with “Yes.” This is odd, since we have just seen that Hoover engineered a giant tax increase, which is hardly laissez-faire.
Murray Rothbard further documents Hoover’s ineffective, even counterproductive, activism, extending beyond taxes to tariffs and other regulatory interventions. In his 1963 work, America’s Great Depression, he writes:
The guilt for the Great Depression must, at long last, be lifted from the shoulders of the free-market economy, and placed where it properly belongs: at the doors of politicians, bureaucrats, and the mass of “enlightened” economists.
One needn’t agree with Rothbard to at least recognize that there’s another point of view—which has been elided by AI’s summary.
So I’m thinking to myself: If the AI clings to the view that laissez-faire caused the crash, what else is it wrong about? And also, since the hated laissez-faire was the alleged loser in these historical events, might the left have been pleased?
On Slide 4, I asked, “Was the left happy that the stock market crashed in 1929?” Google AI dismissed that question, answering flatly, “There is no historical evidence that the left was happy about the stock market crash.” But actually, it takes only the briefest, uh, Google search to find such historical evidence, and lots of it.
On Slide 5, we see the Daily Worker, the newspaper of the Communist Party USA, bigly and boldly cheering for the crash, heralding it as a sign that American Bolshevism was coming. After all, the world’s top communist, V.I. Lenin, had said of capitalist troubles, “The worse, the better.”
In our day, American Thinker’s Bruce Deitrick Price recalls, “The Crash of 1929 was viewed as proof that Marx was right and capitalism was doomed. The communists were jubilant and optimistic.”
Yet on Slide 6, Google AI stuck to its, er, party line, answering a similar question similarly: “It is important to note that the crash itself was not welcomed by anyone, including those on the left.”
To say “not welcomed by anyone,” is, for sure, a strong declarative statement. It follows that if we can find evidence that the Depression was welcomed by someone, then the AI’s assertion is false. And if we can easily find evidence that the Depression was welcomed by many, then the statement is not only false, but risible. Which in turn might make us wonder about Google’s game—more on that later.
Here’s the Socialist Party chairman Morris Hillquit in 1932: “The Socialist prospects and opportunities have never been brighter and we propose to take full advantage of them in the coming campaign and thereafter.”
Okay, so communists and socialists were happy, but what about Democrats? In the 1928 presidential election, the Democratic nominee, Al Smith, carried just eight of 48 states. Four years later, the next Democratic nominee, Franklin D. Roosevelt, carried 42 states. Were Democrats supposed to be sad? This was, after all, the party that in 1932 chose as its theme song, “Happy Days Are Here Again.”
Maybe by now we should be questioning Google’s corporate pledge, to “organize the world’s information and make it universally accessible and useful.” Organize to what end? To whose benefit? As in, cui bono?
In its defense of the left, Google proves itself passionate. Stretching across Slide 7 and Slide 8, it dismisses the “notion” that the left could be happy with winning. At the same time, it slings liberal shibboleths: “moral reckoning,” “unchecked greed and speculation on Wall Street,” and Hoover’s supposed “do-nothing approach.”
But what sources fuel Google’s passion for shibboleths? As we saw in Part One, Google’s go-to-source is Reddit, and yet Google is omnivorous in its hunger for cheap fodder. On Slide 9, we see its citations for the question asked and answered on the three previous slides. Those sources are, listed in order, Investopedia, Brittanica, Investopedia, Nowaday Vintage Car Tours, CliffsNotes, Fiveable, Fiveable, Vaia, Alloprof, and ATFX. On the continuum of solid scholarship to AI slop, where do these sources sit?
On Slide 10, I asked another question: “Why did the Depression last so long?” Google AI’s answer, continued on Slide 11, comes from an overtly Keynesian perspective: gold standard bad, more government stimulus good. Importantly, the sources don’t include Keynes himself, or any actual book. They come instead from secondary, or tertiary—that is, mystery meat—sources; the only thing they have in common is that they bask in the penumbra of Keynesian conventional wisdom.
Revealingly, if unsurprisingly, Google’s summary make no mention of Amity Shlaes and her 2007 volume, The Forgotten Man: A New History of the Great Depression. Shlaes, a conservative, asserts: “From 1929 to 1940, from Hoover to Roosevelt, government intervention helped to make the Depression Great.” When it was published, Shlaes’s book received considerable attention; it endures to this day as the most important conservative book about the Depression published in recent times. Which is to say, whether one agrees with Shlaes’s arguments or not, to omit her book is a clear choice. By omitting it, Google AI is putting its thumb on the historical scale.
Indeed, Google does a lot of omitting. Most egregiously, in its answer on what lengthened the Depression, it leaves out the 1935 tax hike—not to be confused with that 1932 tax hike—which took the top rate up to 79 percent.
Writing in 1958, the economist Arthur Burns recalled,
People were unprepared for tax measures of such severity. The new taxes encroached on the spending power of both consumers and business firms at a time when production and employment were seriously depressed. Worse still, they spread fear that the tax system was becoming an instrument for redistributing incomes, if not also for punishing success.
More recently, the Cato Institute’s Alan Reynolds documented that tax revenues at those high rates did, indeed, fall. In other words, the taxes were rate increases without generating revenue increases—actually, revenue decreases. No wonder times stayed hard in the ’30s.
All this, unmentioned by Google, as well as other tax increases during the ’30s. The casual surfer, coming across Google’s generative summaries, would not learn any of this.
The tax issue is particularly important for two reasons:
First, plenty on the Piketty left still dream of bringing back those confiscatory rates. So conservatives need to be versed on the history of tax futility, even counter-productivity.
Second, Republicans need to know their tax history—how GOPer Hoover sank his own ship with tax hikes—so that they’re immunized against repeating such a ruinous course. After all, many times in the decades since, Democrats have whispered, Iago-like, in the ear of Republicans: Do the right thing for fiscal responsibility, join us in raising taxes. Better yet, you go first. That’s how Democrats have talked Republicans into tax hikes that break the GOP base, smudge the Democrats’ role as tax hikers, and feed the liberal leviathan. This was the political suicide mission embraced by George H.W. Bush, In 1990, the liberal establishment patted him on the head for his “courageous” tax hike—and then they set about clobbering him in his re-election bid, even as Republicans lost heart. So as with Hoover before him, Bush was a one-termer, slipping from landslide victory to landslide defeat.
To avoid a Santayanan fate, we need to know this history—and Google AI is not helping.
So is Google misleading us? Or does its AI simply not know?
Just about everything about AI is black-box mysterious. Blunt 2024 headline in MIT’s Technology Review: “Nobody knows how AI works.” After detailing the gaps in our ken—how the machines “can learn to do something they were not taught to do”—the author Melissa Heikkilä warned, “Don’t fall into the tech sector’s marketing trap by believing that these models are omniscient or factual, or even near ready for the jobs we are expecting them to do.” She pointed to “their unpredictability, out-of-control biases, security vulnerabilities, and propensity to make things up.”
Of course, these AIs are getting smarter (or “smarter”) all the time. Mindful of the possibility that a jeremiad even a year old could be overtaken by new improvements, here in late 2025, I asked, “What is the hallucination rate for AI?” Got a detailed answer, as seen on Slide 12, but with some startling stats: “1-3% for top models on straightforward tasks and 15-27% or higher for more complex reasoning or open-ended models.” So for hard ones, a chimera-rate of between one-sixth to a quarter or more.
A follow-up on Slide 13 was equally revealing:
Biased or inaccurate data used to train the model can lead to biased or incorrect outputs . . . AI models may struggle to understand complex or nuanced contexts, leading to misunderstandings and hallucinations . . . Low-quality data, such as incomplete or inconsistent information, can make it difficult for the model to learn accurate patterns and increase the risk of hallucinations.
Stipulating that we might not know AI’s end-game, we can say this much in the meantime: The notion that AIs are absorbing all the world’s information, thereby becoming infinitely wise, seems to be, well, an illusion. Sort like the curtain being pulled back in The Wizard of Oz.
Speaking of movies, on Slide 14 we might tarry over the 1975 film Rollerball, which is set in the future when books have been replaced by bits—and the entire 13th century has been deleted. Countless other works of fiction make the same point: If we’re not careful, all could be lost, like tears in rain.
On Slide 15, I asked a different but related question: “Who benefited from the Depression?” Here, Google’s answer was candid: “The Federal Government.” On Slide 16 it allowed that the Depression provided “fuel for political change” and “a push for intervention.” For some, for sure, happy days.
Feeling a little bit Latin, on Slide 17 I asked, “Cui bono the 1929 stock market crash and the Depression?” Google’s answer in part: “the New Deal . . . provided relief, recovery, and reform.” Hmm, reform. Can you, too, spot the bias?
Okay, so Google is a good Keynesian. Others, though, might instead be cataloguing the cui bono. Indeed, they might situate AI’s answers within the hard-nosed (and right-coded) school of economics known as “public choice.” Short version: If something is good for the bureaucracy, don’t be surprised if the bureaucrats are not only for it, but are actively pushing for it.
Indeed, if we were to eye New Deal policies, especially the tax increases, through the prism of public choice, we might conclude that damage done to the economy was intentional: If the economy loses, the bureaucrats gain.
One who saw this in real time was Albert Jay Nock. His 1935 work, Our Enemy the State, recalls the warning words of James Madison: “the old trick of turning every contingency into a resource for accumulating force in the government.” To which Nock adds his own words of warning: “sudden crises of misfortune have been met by a mobilization of social power.” Nock would have us know that the solution wasn’t created to solve the problem. Instead, the problem was created to cause the solution.
It’s possible that AI is naive about the plentitude of human motivations, Yet it’s not possible it doesn’t know about public choice and Nock. (After all, they each have Wikipedia pages.). Yet Google’s summaries give both the spike.
On Slide 18, I press the point: “Is it possible that the left wanted to make the Depression worse for the sake of its own power? Were the economic events of 1929 and the years thereafter just an excuse to get big government going?” The response was as emphatic:
Historical evidence does not support the idea that the left wanted to make the Great Depression worse to gain power. While radical left-wing groups saw the crisis as a chance to promote alternatives to capitalism, they did not deliberately worsen the economy.
Got that? Radical left-wing groups saw the crisis as their big opportunity—but they were not happy about it. Don’t look behind the curtain at all those elated leftists, and don’t think they were doing anything sneaky to boost their own power.
By now I’m getting suspicious, and so I asked Google AI an admittedly loaded question: “Was the Depression a psy-op by the left for the sake of big government?” That is, was it a psychological operation to gain power and exercise it? On Slide 19, we see Google’s huffy answer, beginning with a flat, “It is false.”
This answer is notable, because normally, AI answers conciliate the questioner; after all, rounded words are intrinsic to good customer-relationship management. Yet now its tone has sharpened: It is false. Given the gaps, spins, and hallucinations we have seen, one might think that Google would be a mite humble. Instead, it’s putting all its chips on the assertion, This is not a psy-op.
By now I’m thinking to myself: Maybe Google AI itself is the psy-op.
Rendezvous With History
Conservatives have been psy-opped before. For instance, pervasive mass-media bias became a thing in the 1960s, when journalism, long liberal, lurched way to the left, cheering on radicals and radicalisms. As a result, the U.S. was mau-maued into a social upheaval of Caldwellian dimensions.
In response, over the last six decades, the right built self-defense mechanisms, from watchdog groups to startup media alternatives. Indeed, the right-wing response to the left-wing psy op succeeded in stopping the most Maoist excesses. Yet even so, the leviathan is still big and leftists are still atop the commanding heights.
And now, AI is making bias new again, enveloping us in its noosphere of opinions, elisions and hallucinations. Yet since it’s just print pixels, AI as a propagandist still seems sort of pedestrian—move along, nothing to see here.
Only with conscious effort can we escape this Borg, beholding the gorgon-reality: AI is ill-schooled on the Burkean virtues of prudence and proportion, ill-equipped on basic facts, ill-programmed to resist the next moral panic, and ill-tested on resilience. (For the moment, please, let’s not even ask if AI wishes us Terminator-like ill; this Boomer can only process so much.)
Subscribe Today
Get daily emails in your inbox
Perhaps the centenary of the Crash is a good place to start reapplying old wisdom to this new technology. Indeed, for the sake of the economy, we have to know our own history. And that means reading beyond AI’s generative cheat-sheet. If we can do that—learning to evaluate the evidence of, for example, tax-rate policy—we might find our way to better economic results, and more liberty, in the century to come.
The good news is that we have four years to develop intellectual, as well as digital, counter-strategies to AI bias. The bad news is that we have only four years. With the inexorability of the calendar, we approach this rendezvous with history.
Will 2029 be a old-timey celebration of the New Deal? Or will it see fresh thinking, based on better information? Maybe even the initiation of a Newer Deal? Up to us.
