On Tuesday, Brad DeLong joined me for a live Theory of Change episode to discuss Donald Trump’s firing of Erika McEntarfer, the former commissioner of the Bureau of Labor Statistics, and how it is part of a decades-long war on sound economics and expertise in general waged by reactionary Republicans.
We also discussed Brad’s own efforts to get economics as a discipline to understand complexity, and then wrapped with a discussion about the economy and artificial intelligence, one of his current research interests
Video Timecodes
00:00 — The Erosion of Expertise and Trust
01:54 — Introduction and background
04:35 — Historical context of labor statistics
07:12 — Republican response to BLS firing
10:35 — The problem of Republican truth-telling
14:26 — The importance of judgment and expertise
18:05 — Trump's broader attack on expertise
26:01 — The first wave of AI development
30:20 — The crypto grift and AI investment
36:24 — The need for independent statistics
45:27 — The value of government investment
54:24 — Understanding AI's current state
01:04:11 — The role of liberal education
01:13:08 — The future of AI applications
01:19:58 — Musk, Trump and the tech elite
01:26:23 — Right-wing epistemology and AI
Transcript
The following is a machine-generated transcript of the audio that has not been proofed. It is provided for convenience purposes only.
MATTHEW SHEFFIELD: All right. So, for everybody who is, joining us, we're doing, this is a [00:02:00] live theory of change show and, we are here today with Brad Long of the, well, various places where you are at, but, primarily, your own Substack, we're going to be doing that. and, you are also you also have some affiliations. you're a professor at the University of California Berkeley.
Brad DeLong: That’s the big one. Yes.
SHEFFIELD: Yes. Okay.
Brad DeLong: Important one.
SHEFFIELD: Yes. and also, so, and I wanted to bring Brad on to do this with me today to talk about obviously, Donald Trump's firing of the Bureau of Labor Statistics, head, Erika McEntarfer, after he had the bad jobs support, a bad jobs number. But it's obviously, this is part of a much broader picture, I think, of the reactionary right’s hatred of expertise, and, continually exemplifying that, that old phrase from Stephen Colbert, that reality has a liberal bias.
DELONG: A liberal bias.
SHEFFIELD: It seems that's what they think.
DELONG: Yes. In many ways, it is the end of a long and honorable American tradition. as another Erica, Groen, my class, my graduate school classmate, who was an earlier commissioner of labor statistics, as she said, I think on Bloomberg's surveillance this morning.
It was in 1884 that Carol Wright, the first Commissioner of Labor Statistics, said that his contract with the American people was that they were going to produce the best possible numbers they could and get them out without worrying at all what implications they had for any kind of political movement.
or social reform. Or social order, right. cause that it was not their job. [00:04:00] to do anything other than to get the most accurate numbers they could out and that they pledged, that's what they would do for the American people. Donald Trump is now trying to break that. He is going to face a lot of resistance from the bureaucracy, which is in this case an extraordinarily good thing because the bureaucracy does indeed do its job as best it can, and they are indeed addicted to the idea that it is their job to produce the numbers.
Historical context of labor statistics
DELONG: That indeed, back in the 1870s, in the 1880s, people found themselves in furious political debates here in America. some of them saying that America was getting more prosperous, others pointing to falling prices of our farm goods and saying it was becoming less. Some of them saying that income distribution was becoming more equal.
Others saying that from what they saw, it was becoming more unequal. And a whole bunch of people got together and said, we disagree on a huge amount, but we all agree we do not have reliable information about this, and we should spend some serious money creating governmental institutions that will in fact produce serious and accurate information.
and that's what people interested in America, interested in, good governance, interested in a actual functioning, political sphere that does its job, got behind in the 1870s and 1880s. And that is the reason for the statistics movement. And first, the Massachusetts Bureau of Labor Statistics, and then the Federal Bureau of Labor Statistics, were started as a result of that 140 year tradition.
And now with I say the complete absence, the complete absence of any kind of pushback from. Any Republican whatsoever against this with no [00:06:00] resignations from Trump's staff with no hearings called by any senators with nobody in Trump's orbit willing to say, by and large, the economic news is likely to be good over the next four years.
And by doing this, you have just made it certain that no one will believe it, right? That you've actually shot yourself in the foot fairly significantly. And the right thing for you to do is to say, you flew off the handle and actually, will you please come back and work for us? You're a great, person that after all you were confirmed, what was it, 93 to six with JD Vance voting, yes.
are you saying that, does Trump saying that JD Vance made a big mistake in choosing a partisan democratic apparat to run the Bureau of Labor Statistics, in which case the wokeness is coming from inside the White House in a very severe fashion. You don't wanna say that. You wanna say you overreacted, you flew off the handle, and we should walk this back.
Republican response to BLS firing
DELONG: Nobody is saying that to Trump. In fact, Susan Collins and Bill Cassidy were writing to the Commissioner of Labor statistics even as she was being fired, saying, why are you producing lousy numbers? Right? and the answer is that you have to, that this particular set of numbers comes from the reports that corporations make as they pay Social security and other contributions into the government.
And, those things arrive. They arrive with sometimes a lag. You know that not every piece of paper gets in all the time at the right moment, and
SHEFFIELD: As a,
DELONG: of paper are [00:08:00] missing, you guess what they are. But then you get the real pieces of paper later on. This causes problems for the statistics.
There is an industry of economists, about how the Bureau of Labor statistics, is unlikely to hit turning points accurately because it will assume there is more inertia in the missing data in the system than there in fact is. And when the new data comes in, you say, oh, the economy actually started growing two months ago, or, oh, the economy actually started shrinking three months ago.
And they frantically attempt to construct better models to do this without a great deal of success.
SHEFFIELD: because it's flying blind in a
DELONG: to demand that you produce numbers that are not revised. Is either a demand that you go out, you send your goons out and start firing or start fining people who don't get their social insurance contribution paperwork in on time, which is you make the government a lot more oppressive and nasty in terms of filling out forms.
or you delay everything a month or two months, or you report on what the level of payroll employment in July was, not at the start of August, but at the start of October. And if that's what Susan Collins wants to see, she's an idiot. And if none of the members of her staff told her, wait a minute, if you complain about this.
Then you are either calling for data to be delayed for two further months so it can be properly checked and assessed. or you are simply wanting to make some kind of headlines about how you can't trust government statistics for partisan advantage. And neither is a professional occupation for a senator. yet apparently Susan Collins staff has no one on it who will say, wait a minute. Susan Collins herself, who is after all my [00:10:00] 52nd favorite senator, was willing to go whole hog in on this, and she has not asked Donald Trump to reconsider the firing of the Commissioner of Labor statistics, which means essentially that the entire Republican party is gone, from the perspective of actually wanting.
Ideas that are true rather than ideas that are convenient go out there in the world.
The problem of Republican truth-telling
SHEFFIELD: Yeah. Well, and, it is notable that, the guy who was the previous Trump appointee. as the BLS head, has, in fact, he just came out with a statement criticizing, her or criticizing Trump for firing Erica the qar far. and, but yeah, it, only emphasizes your point though, that these are only the former Republicans, who are engaging in truth telling, in this regard.
Like the actual people who are elected who could do something are, too intimidated and they're too intimidated by the Trumpian cancel culture against accountability. because like that's really, there, there's, this double game that has always been played since William F. Buckley came along and created, he, and others, created the, post-war reactionary movement that took over the Republican Party in which one party acts entirely in bad faith at all times.
And everyone has to pretend that they don't. And so they attack all expertise and all data that, that counteract their beliefs and opinions. And then, everyone just has to pretend that it's serious. and this is why the Republican Party has, a higher approval rating than the Democratic Party does, because Democrats actually are willing to [00:12:00] criticize their own party, whereas Republicans don't.
DELONG: yeah. No, I do remember being struck right in the last Trump administration, that, Economists, Robert Barrow and came out with a prediction that the then Trump McConnell Ryan tax cut would be a huge boost to American economic growth. Would in fact raise the level of investment in America by so much that by the time it had worked its way through, we would fully have two fifths additional as much machinery and buildings and infrastructure and, capital as we had now.
And that enormous increase in capital investment would drive a huge boom to American labor, productivity and wages. And Larry Summers and Jason Ferman took them on the grounds that this was nonsense. what Barrow and Boskin were saying and said, we should at least be able to get together and not do this, and to get together and get behind.
Say the Congressional Budget Office or the Joint Brookings Urban Tax Policy Center, or the Tax Foundation to assess what the lab impact of a, like tax cut is likely to be. And Boskin wrote a very weird response, right? that it was very much that he was not going to sacrifice his analytical judgment to any other bunch of people, even the highly respected tax foundation or the highly respected joint tax center.
and it was very important that the impact of this tax cut would be determined by how the regulations would be written. And it was very important for people like him to be in the room when the detailed regulations were written. and you read this and I read this [00:14:00] and I thought, gee.
If I actually believe that this con, this policy, if I actually believe this policy would produce a huge investment boom in America and cause a doubling of productivity growth over the next decade, I would not be defensive. I would say I will not sacrifice my judgment. I would say, well, my judgment here is correct and your judgment is faulty.
The importance of judgment and expertise
DELONG: And I will bet substantial amounts that my judgment is correct. And to the extent that the tax policy center and so forth are not in line with them, they need to reform their procedures because they're making bad estimates rather than the milk toast. we disagree. piece. So as I wound up thinking about it, I wound up concluding this is like the person blinking in the hostage video trying to blink in Morse code to say SOS, right?
That is that Boskin thinks he is a force for good in the Republican party. And, telling public things I regard to be bald faced lies about the likely effects of tax cuts. things that Republicans have been telling since the Reagan tax cut of 1980, is his price of admission to the room where it happens.
And do you know that it's better than he do this than the, true crazies run things. And so please don't be mean. and that seemed to me to be no way to be a professor. I must say. And. Do you know, the fact that there, none of them out there saying, this is really a bad idea.
the
SHEFFIELD: No, it's notable the, they're
Brad DeLong: you even like, makes me even less happy,
SHEFFIELD: well, and it's the fact they're saying nothing, I think
DELONG: yeah. The fact they're saying [00:16:00] nothing is interesting, at least they're not saying that, gee, this is a wonderful thing that will boost economic growth.
SHEFFIELD: it shows that they don't believe that they do think it's wrong. because, again, if they thought this was just a neutral thing or it was a good thing, they would be saying that
DELONG: there would be saying that, right? That there are, a bunch of people who said back when, oh, who was it? Was it.
Mitt Romney was running, it was then running for president and having a meeting. And the question was, what economic growth should we project? And Jeb Bush says, let's raise this by a percentage point. And a bunch of the, economists go along. Marty Feldstein, my teacher, who, the late teacher who I adored dearly, who is a dear, man, who is a brilliant man, was a brilliant man, all went along.
As opposed to say, back in the Clinton administration, when Gene Sperling was asking Larry Katz to put an economic growth number forward. and Katz said, I think you should say X because there's an 80% chance we'll exceed that over the next four years. And, Spurling said, fine, you know that all the numbers sound reasonable.
It's a reasonable one for budgeting and I would like to be in the position where you can validate me later on by saying I was pessimistic about what the effects of these policies were. that's not something you would have on the other side of the aisle. And that's, I think, a genuine loss for America.
SHEFFIELD: It is. Yeah. and, of course this firing, is just of a piece with Trump's larger attack on expertise. And, probably the two biggest other areas that we're seeing this is with RFK Junior and [00:18:00] the, the Department of Health and Human Services. But, and in regards.
Trump's broader attack on expertise
SHEFFIELD: To in particular funding of the National Science Foundation, that they're putting, freezing all these grants and prohibiting certain words from being used like, or, you get flagged if you use the word woman, literally.
DELONG: Mm-hmm.
SHEFFIELD: and I, wanna talk about, just so you know, this is a larger trend, but there does seem to be, I think there, there's like a, so there, I think, oh, wait for that. it, it seems like there are some, oh, hold on, we
DELONG: Sorry,
SHEFFIELD: okay.
DELONG: I did not. Okay.
SHEFFIELD: All right. Well, hey, this is live recording here. So things like that happened, my dog just walked in the door behind me. I don't know if anybody could see that. but yeah, so I think though that there are so within, the broader, say, Le Center to left.
Sometimes there are people who they don't know this history of bad faith and deliberate ignorance on the part of reactionaries, and they seem to think, and I think, Thomas Chatterton Williams as an example, he just came out with a book claiming that, if only liberals had been not so x like then the reactionaries would not be so, so deliberately ignorant and malicious.
and it's like you don't actually know any reactionary political elites do you? Or you haven't really paid attention to what they say, because you wouldn't say that if you had
DELONG: there's also a view that the people I like have no agency. Right.
SHEFFIELD: Or expand that if you would
DELONG: that, the people you like, the people whose behavior you're excusing, are just reacting [00:20:00] normally, instinctively, reflexively. And it's your fault because you do not understand them and you are the only person who can change things because they're simply reacting in this particular way.
as opposed to demanding that they have the ovaries not to be, malevolent idiots who care only about ideas that are convenient for them. As opposed to ideas that are true about the world. You know that one of the great things that happened with the coming of the Enlightenment, in was that for the first time you actually had a group of intellectuals where the principle test of an idea is not as it convincing and helpful to the guys at the top running the Society of Domination.
That takes a third of everyone's crop for itself. Who are the people who pay me? but is this idea actually true about the world? that for some reason the Catholic Church thought it was very important to preserve a medieval cosmology by which humanity has been put by God at the center of the universe.
And to say, gee, wouldn't the TMA model of astronomy be much, much, better if you understood that it wasn't the case, that there was just the earth and all these things revolving around it, but rather that the sun was at the center, the moon revolved around the Earth. The earth revolved around the sun, and Mercury and Venus are closer to the sun than the Earth while Mars and so on are further from the sun than the Earth.
If you start with that, all of a sudden a huge number of things about the pathic meic model of the universe, [00:22:00] which is a good one for forecasting, where the planets are going to be a whole bunch of things about that model makes sense and you understand why Mercury and Venus in retrograde are so different.
From Mars and the outer planets, and you understand why the moon moves so differently than the other planets and why the sun moves so differently from everyone. That all of those things are very accessible if you are a Copernican. While they are bunches of strange, mysterious things, if you are a T and Galileo was indeed shown the instruments and said, if you do, not recant, torture is the next step until you do.
the view that is not the kind of way to run a society is a creation of the enlightenment and is, was a major, advance. And it is one that from the right perspective of the right, we have now lost.
SHEFFIELD: Yeah. Yeah, it is. And and, we see this also just going back to the, the labor hiring statistics that, the Trump tariffs and his constant. In uncertainty that he's creating about them, promising deals and saying, we're gonna have a deadline of this and then ignoring it, then we're gonna, no, we're gonna move it up.
DELONG: Yeah.
SHEFFIELD: it is creating all kinds of uncertainty for business owners and employers. And so of course the job numbers and other GDP numbers are going just going to be extremely erratic and he's the source of it, but they can't accept that.
DELONG: it was very, yeah, as I say, the absence of pressure, to say,
let's say if Governor Waller and if. Reserve Bank President Bowman, and [00:24:00] if hopeful, future Fed chair Kevin Walsh, say that the Fed really should lower interest rates. Yeah. They also need to say the Fed should lower interest rates because I am confident that policy uncertainty will be removed and that the Donald Trump administration will give us policies with pre economic predictability and sustainability to support growth.
So that lowering interest rates will be a prudent thing to do. Right? If you're wanting to be a professional central banker, you make an ask to the rest of the government, which is that if you want a low interest rate policy, you undertake policies that make it prudent for me to do, which was very much what, Alan Greenspan did with Bill Clinton at the start of the 1990s.
yes, I will lower interest rates. If you do policies that make it prudent for me to do so. Right. And Greenspan followed through. He lowered interest rates big time. And we got the 1990s, you had the best boom that America has seen except for the sixties, since God, since the 1920s in terms of its strength.
and yet Bowman and Waller and Walsh will not even make that ask.
SHEFFIELD: And, because they won't get hired, Trump won't pick them. if
DELONG: you know that Bowman, has a job is not really a serious candidate for Fed chair. Waller would probably be the best, I think of the likely candidates Trump might pick. And so maybe you can give him a buy. On those grounds.
SHEFFIELD: Well, give us the background on them if you, for people who [00:26:00] don't know them,
The first wave of AI development
DELONG: No, I,
SHEFFIELD: who these guys are,
DELONG: no, that Bowman is a reserves or right wing reserve bank president and she stands out in my mind for someone who said that the reason that housing prices was too high was because immigrants, that the reason there was housing inflation was immigrants, which is law of false.
SHEFFIELD: yeah.
DELONG: Waller is a technocrat who's played a bunch of footsie with the Trump administration, but I think would be an extremely smart guy if he's willing to stand up to the White House, in a clever and appropriate way. The problem is that should have started now. Warsh is a guy who is for always for lower interest rates.
So the economy can grow faster when a Republican is president and higher interest rates. So inflation is under control when a Democrat is president and switches with alarming frequency, remarkably rapidly, in my view, is a dangerous person to put there and is, I think his models of the economy even taking off the political part are not that sophisticated, are not that accurate.
So,
SHEFFIELD: Yeah. Well, and the other thing is that I think Trump is, that there was talk early on in the administration that he was trying to push a new model of economics that, the Mar-a-Lago accords. and that was all the rage and in the last But I haven't heard a lot of people talking about it now.
Like, what's, what, do you think is the deal with that? What's
DELONG: it was Steve.
SHEFFIELD: Yeah.
DELONG: Hoping that he could actually get his policies through by working for Trump and by saying whatever Trump said he actually wants my preferred policy.
SHEFFIELD: Yeah.
DELONG: And, you can have arguments about how his preferred policy [00:28:00] is actually not so smart, that what his preferred policy winds up doing is it winds up sacrificing the global, role of the dollar in exchange for a, in order to get a dollar worth less so we can have more manufacturing, ho here in the United States with the idea that manufacturing is a locus of the creation and maintenance of the communities of engineering practice that are uniquely value in pushing forward technology and in raising incomes and wages in the long run.
The problem is it's not just manufacturing that's important. More important than manufacturing is investment. And the global role of the dollar allows us to invest in America at a much, much greater pace than we would be able to otherwise. So, you wind up sacrificing, you wind up sacrificing the dollar's global role.
You make America poor, you make the dollar worth less, you wind up losing more in technology, forcing communities of engineering practice in investment, good sectors than you gain in manufacturing sectors and it's no longer So clear manufacturing is really gonna be so key over the next generation, even though it has been key for the past five.
So, Trump wanders around expressing grievances essentially at random. Policymakers are then responsible for red, conning them into some particular view of the world. Trump will then endorse or deny your red conning depending on whether he likes you or not, and whether he thinks it sounds good and approved or not.
SHEFFIELD: And what side of the bed he woke up
DELONG: yes, this is a very dangerous game to play, but there are an awful lot of people who are eager to play it. [00:30:00] some I think because they genuinely think this is their only chance to possibly influence policy for the good. some, because there's going to be an awful lot of grifting done and money made from the Trump Kleptocracy complex and they wanna be in on it.
The crypto grift and AI investment
DELONG: Right.
SHEFFIELD: They do. Yeah. And the, thing also, that this, and I, and I don't think this is just miran, this does seem to be the reactionary fundamental economic assumption, which is they believe in negative, negative fiscal multiplier. That's what it seems like. They believe that a
DELONG: Except they believe, in a positive fiscal multiplier, right? That the,
SHEFFIELD: say that again.
DELONG: that they change from month to month.
SHEFFIELD: but it does seem like that, yeah,
DELONG: the, so belief Yeah. The, that there are two beliefs, right? The, one belief is that taxes on capital are always too high. a second belief is that workers always have too much bargaining power.
and it is a good thing if those without papers, are scared, lest their employer drop a dime on them. And so they want to be extremely valuable employees and that their employers are turn, are scared that a customer will report them to the I to, to the ICE. And so as a result, they're willing to offer real Americans good services at low prices.
they love the idea of H one B visa holders. the idea that someone whose visa status is tied to their job with the added wrinkle that these days, if you fire an H [00:32:00] one B visa holder, you can then call up Christie Nome and say, deport this person tomorrow. And so H one B visa holders are now the equivalent of ERFs.
And Elon Musk would love to have every single graduate from an Institute of Indian Technology Institute, Indian Institute of Technology in the United States as this kind of H one B erf, and it's spreading that We have discovered that if you're a green card holder and if Christi Nome asks Marco Rubio revoke your green card on the grounds of national security, and the Supreme Court will not help you, right?
You then have to try to fight this revocation or a three year process from another country if you want to get back in. And, God knows what's going to happen to people with citizenship who in when they claim that there is some irregular irregularity, some lie on their citizenship papers. but I don't think Elon Musk or Melania Navs have to worry about the fact that they overstayed their visas in the past, while claimed on their citizenship papers that they had not. But quite possibly. Lots of other people will, and quite possibly the Supreme Court will not help them because four members of the Supreme Court believe that Trump is a righteous God, and two members believe we do not dare pick a foot fight with Trump because we might well lose it.
SHEFFIELD: Yeah.
Well, and so, but, so, but going back to their economic consumption though, like they, they say at least currently as they, they do oscillate, especially when Democrats are in power, that they currently are saying that the reason why we want all these cuts to services like Medicaid or Medicare, and is that.
Government spending takes money out of the economy, and is debt driven and therefore, [00:34:00] has a negative stimulus on the economy. That's what they claim. And then of course, they obviously don't believe this by virtue of the fact that they have increased the debt, by over a trillion dollars. So they don't believe what they're saying.
But I do think it's important, to, for if you would like to push back on this idea that they, literally say that, the fiscal multiplier doesn't exist. And if it did, it's negative.
DELONG: well, the more that government spending doesn't do, anything, and then
SHEFFIELD: say it's harmful actually.
DELONG: well, but you know that it's harmful because it adds to the debt,
SHEFFIELD: Yeah.
DELONG: except when the debt is not a problem.
SHEFFIELD: Well, granted, like I said, they don't believe
DELONG: you say, it's not, that there are individual Republicans who have economists, who have views, who go in and out of favor, depending on what is the desired legislative policy, convenient proposal of the moment, and confidence that people will fall into line or at least stay quiet, when policy goes against them.
that very much that it's not that. We need to work to find the economists who have ideas about how to make the economy grow faster and then implement their policies. It's that we know how to make the economy grow faster. it's to cut back on government spending wherever possible, cut back on regulations except those that we like and eliminate taxes on capital of all kinds.
we know that and we don't need to know anything else. Yeah. and so economists are there to be hood ornaments for the car that we're driving, and if they're not willing [00:36:00] to pose properly there, we can find another one.
SHEFFIELD: Which is yeah, the approach he's, he is doing. With BLS, clearly. I mean, and I think,
DELONG: it's counterproductive. It is massively counterproductive from Trump's own perspective because he has just made it very clear that none of these numbers are to be trusted.
The need for independent statistics
SHEFFIELD: Yeah.
DELONG: that he has taken steps we see and almost surely steps we do not see to attempt to corrupt the statistical process.
And, I will be surely cross crossing, reports that come out of the BLS and of the BEA and of the census going forward with other potential large sources of data. and indeed it is time for the financial times for Bloomberg, for the economists to themselves stand up a shadow statistics department of their own, to say, here's what the government reports, but here's what a big data.
Assessment of the general information flow says,
SHEFFIELD: Yeah. Well, and, I think there, I think you're right that's almost inevitable at this point now that, Trump has tried to corrupt the process. but there is a, there's an interesting parallel, I think perhaps in what Trump had been trying to do to the, the National Weather Service, which provided very critical, forecast weather forecasting data and, recording information to specific small local communities.
And, then Trump went and gutted it. and, the intent very clearly with regard to the NWS is to eliminate it. Like that seems to be the goal because they think that there's no point in government providing whether, whether [00:38:00] forecasting services seems like, and so I wouldn't be surprised if at some point you'll see soon Republicans saying, oh, well, who needs.
BLS at all. Why do we need, why? Because like, I mean, if you remember, I guess it was about two months ago, they, fired a bunch of the advisory board that was contributing to labor statistics, if you remember that right.
DELONG: Yeah. Yeah. but you know, the, poll point, or at least, in these are not things that have a material impact on the debt. and they are things that do have a material impact on say, the ability, if you're a farmer in Iowa, to figure out what a hailstorm is likely, or your ability if you are a commuter in, suburban Dallas, to figure out when you need to wait for 20 minutes, before heading out onto the road, that, that a world, that.
We have an immensely complicated world that has enabled us to be quite prosperous, in which we have extremely large scale, you'll call them information and coordination services called the market economy, called Apple Computer called the Federal Government Bureaucracy of X. and all of these things need to be healthy and need to be working well, or we wind up with a much poorer society very quickly.
Trump is probably not going to be able to
break all that much for all that long, but do you know the point is that there are an awful lot of things that are, he will break, that [00:40:00] will be quite, damaging.
SHEFFIELD: yeah.
DELONG: and one way to think of it is that we have, or we had before Trump, we had productivity growth of maybe 2% per year, in America, coming from, investment in America of about 20% of our income a year in total.
And so that's a net return on our investment spending of about 10%. and of that 10%, right, about 5% comes from about a 5% return on private invest sector, Amer investment in America flowing directly, flowing directly to the people, the corporations, the people who make those investments.
You also have about 2% return flowing to other corporations and businesses that can see what one business is doing and taking advantage of it. Or flowing to workers who have advantageous market positions and get to share economic rents. but still the 17% of investment in America that comes from business multiplies 17%, 17 by, 0.7, 0.07.
and you get, 0.12 right? Or 0.012, which means that. Two fifths of our economic growth, eight tenths of a percent per year of our economic growth comes from the public investments in infrastructure, in health, in education, in co, in technology,
SHEFFIELD: Science.
DELONG: and development in science.
And so if [00:42:00] this part is getting indeed 0.08%, that's not a 5% return or a 7% return, that's a 20% return. And so hitting public investment is a uniquely damaging thing to do to American economic growth, and it's something the Trumpists appear to be very eager to do to the maximum extent possible.
SHEFFIELD: Yeah. Well, and it, also like that percentage, that's just the direct measure though. So in other words, like the, NHSI think is pro is such an incredibly successful investment vehicle. Nasa, you've got the, these two agencies that have, been at the cornerstone of so much worldwide innovation.
DELONG: Oh yeah.
SHEFFIELD: and, they more than paid for themselves over the decades. every single year they do.
DELONG: And what I've done is purely an exchange value calculation looking at things that flow through the markets. But you know, there's also the use value calculation, that you need to do, right. that antibiotics, right? Antibiotics are productive things and they're an important part of our national income and product accounts.
And when you produce, say, an antibiotic that allows. Oh, someone like me who got an infected abscess in his butt in his thirties and went down to the University Health service and, they glanced it, they poured antibiotic powder in it. They slapped a bandage on it. They said, come back in two days if it's still red.
And then they yelled at me for getting, letting it get this bad. Nathan Mayor Rothschild richest private citizen in the world up to that point, up to the first half of the 19th century, got the same [00:44:00] thing in his fifties. He died, he died after being tortured by months for German doctors who were trying to figure out how to deal with this infection and could not.
He died in his fifties before he got to see his grandchildren grow up. And, what fraction of his wealth would he have given for a single dose of antibiotic powder that could have killed this thing stone dead in five minutes when you don't even get to see a nurse. You just get to see a nurse's assistant.
that we have, the value of, our, the value of our technology in terms of health, in terms of longevity, in terms of not just kid good health, but the ability to actually do things and be mobile and active in the world. Plus the value of our technology, not just in allowing us to consume goods, but to exchange ideas and laugh at ourselves.
those themselves are massively greater than what the national income and product accounts tells us. And because those things do not flow through the marketplace, are not bought and sold, right? Those use values and those entertainment and those knowledge values are disproportionately created by government, as well.
The value of government investment
SHEFFIELD: Yeah. And this, this general ignorance, that the reactionary right has about these, non, non-direct, or, in indirect account, stimuli and obligations and red and resonant effects. it's, it is just bizarre that this belief continues it, given that Adam Smith, their supposed hero, that explicitly.
And it wrote an entire book about how [00:46:00] the, what, was the, just sort of conventional economics, if you will, that human societies are so much more than that. And they completely ignored him. and sure they did not read any of his books for that matter.
DELONG: Right.
SHEFFIELD: and, and the, and, just for, people who are interested that the book is the Theory of Moral Sentiments, it's definitely worth checking out to understand that. that it, is just not simply about, checks going in from one account to the other. Like, this is about soci. That's what the point of economics is supposed to be, is about. because it comes from the Greek word, omi, which is the management of the household. and it's not, so, it didn't mean just the budget.
DELONG: Yes. Okay.
SHEFFIELD: and that's something you have, been trying to, you've been trying to reorient economics, toward that more. I, we, I think. And how do you feel like that's gone outside of the Republican, party?
DELONG: I think it's gone great. Right. I think economics as a discipline is in a much more healthy state Yeah. Than it was when I started in this profession 50 years ago. in large part because we are much more aware, of, the limits of the perfectly competitive model and also much better in figuring out how to deal with the fact that model has limits so that we are not just, in the business of saying, make this like a perfectly competitive market.
But rather saying that since this cannot be made like a perfectly competitive market, here are all the second best things we can do to make it significantly better. [00:48:00] not just in terms of the market structuring part, but you know also in terms of the internal, within the organization management cybernetics.
organize the organiza, create an organization that is properly responsive to what it needs to do and what it can know. and also with respect to human beings ability to successfully manage their desires so that they can live wisely and well, as in terms of understanding how great the market system is as a way of successfully coordinating a productive division of labor among now eight and a half billion East African plains apes.
With respect to that, Adam Smith had most of it. but when you cannot get a perfectly competitive market in an industry or a sector, and when you cannot get well-informed consumers and when you have problems of designing a. Organizations that stand between the indi, larger, smaller in size than society, but larger in size than the individual that need to do necessary parts of the coordination because the price signal by itself does not have enough bandwidth.
we've made so much progress on that in the last 50 years that I actually can be proud to be an economist in the way that one really couldn't, back when I started in this business.
SHEFFIELD: Yeah. Well, and it's, it is the, yeah, the understanding that markets, they're not just the creations of the government, that, but they have to be managed, in, they have to be overseen that the bounds without any sort of bounds or regulatory structure or, anti-monopoly structure, that they will collapse into monopoly.
Like that's the natural tendency, especially with technological progress.[00:50:00]
DELONG: Yep. That after all the whole idea was mean. The very idea that you would have a market economy was itself a radical expansion in the size of government, right? The idea was then that property rights and terms of trade, would not be, what local notables, what the local big man said should be property rights and terms of trade, and was not what either roving or stationary bandits, would say Should be yours.
You are yours only as long as I can milk you. If God had man not meant you, to be sheared, he would not have made you sheep as the villain says in the magnificent seven. and I think most important at all, that the. Organization of property and contract, not be at the whim of the government's own functionaries, right?
all these things that allow people to invest, accumulate, understand where their sphere of action is, and to figure out how their resources can be best used and what resources others have that they should trade for, that's an absolutely, truly remarkable thing, right? That's crowdsourcing the problem of economic organization in a remark, fruitful and productive manner.
That earlier societies that said, you are a surf, you are a slave. Here's what you have to do. Here's what the big man says is yours. Here's what the big man says, isn't yours. Here's what it's such a huge advance and also very much a product of enlightenment, ideas about individual autonomy, agency action, and truth.
You. Wonderful thing. Wealth of Nations. Absolutely wonderful book. Not [00:52:00] all of it, right? Not all of it.
SHEFFIELD: it's not the full story.
DELONG: by a long shot. And we, economists now have management cybernetics and behavioral economics and more sophisticated market structure ideas that we did not really have 50 years ago that allow us to properly situate the market in its place. Right? That after World War ii, there really was a belief. You only had to do three things. you only had to have fiscal and monetary policy to maintain full employment. You only had to impose the proper pollution and other taxes on things that were noxious. And you only had to have a progressive tax system and some kind of safety net in order to make the distribution of income and wealth not too unequal.
Those were the only three functions of the government that plus national defense and so forth, that were truly legitimate. Now we understand that the proper way of structuring the economy in the society is much more complicated and requires much more subtle. And, and we're doing, with the exception of the past six months when you've been making rapid, backward progress, we're not doing that poorly.
SHEFFIELD: Yeah. Well, and, the proof is in the pudding in terms of that there has not been, since the rate depression, there hasn't been a rate depression,
DELONG: Right.
SHEFFIELD: the United States any large, major world economy, since then. So, but there, now there is, as, we come up on the hour here, there's, let's maybe turn the, focus to ai, because that's, I think is a source of a lot of uncertainty, with regard to people's careers, with regard to investment, with regard to the economy.
there's, and I think that on the political [00:54:00] left, and, I'm curious what your thoughts are that there, there seems to be a lot of. Of, strong skepticism and on, on, with regard to a ai. Ai, they think it's a scam, that it's just a big nothing burger. and you don't seem to be of that opinion.
I'm not of that opinion, but it me, but there's a lot more nuance here.
Understanding AI's current state
DELONG: I mean, to the first order, AI is having tech platform oligopoly build data centers at an absolutely astonishing scale in order to provide natural language interfaces to their existing businesses. So that nobody else who constructs a better natural language interface to their business that responds more accurately and quickly to questions say for Amazon about where to shop for Facebook as to how to interact with my social network for Google as to how to search for Microsoft as to how to actually do my group team work, job, that no one with a better natural language interface to that can make a run and grab my markets and my monopoly profits away from me.
And it is worth spending a large chunk of my current profits on building out these data centers and using these models in order to provide this particular kind of insurance against ai, natural language interfaces turning out to be really, big and really, important, and I cannot afford to be behind at all.
Or else some upstart whipper snapper or one of the other platform monopolists, near monopolists might take my profits away from me, relatively quickly. That, that's what they're all doing with only Apple of the great platform near monopolists, hanging back. So that's the first order thing.
whether these will actually [00:56:00] be used for their intended purposes or not, nobody really knows yet. But at any event, we will have an awful lot of data centers at the end and we can do an awful lot of data processing, as a result that if we can find something useful to do, it'll be neat to have all these GPUs hanging around.
but it's probably not going to make a great deal of money for the tech incumbents. It's just a way to spend money to best way of buying insurance against disruption to preserve their current profit flows. The second most important. In terms of financial flows, use of AI is as a successor to the crypto grift as a way for not too ethical and extremely persuasive venture capitalists to separate gullible, venture capital organizers to separate investors and venture capitalists from a surprisingly large amount of their money.
Right? That, crypto was a great grift and it took a lot of money away from a lot of people. and a lot of people now have crypto assets, which they think are worth something. And at the moment for each individual, they are worth something because you can sell it to another gullible investor.
But long-term value of crypto does not look that bright for practically everything except possibly Bitcoin if it manages to make the jump to digital gold, which it will be interesting to watch if it does, but you know, everything else is gonna be toast.
SHEFFIELD: Yeah. And actually, sorry if I can interject on that point. I think you made, a, your rate made me think of something else. Is that the only way that crypto will increase in value over the long term is government like, it, so, it was, it, markets itself as this, anti-government, libertarian, utopian technology, but ultimately only government can make it have value by, [00:58:00] by, by giving it sanction.
And yet they, and so they're all rushing to lobby and endorse candidates and give money, but they won't admit the fun, fundamental contradiction of their entire worldview that they're engaging in. But I go by, go
DELONG: And, and, third most important at the moment, at least important financially, but also very, important for the quality of human work is
modern advanced machine learning models as not so much chatbots, but rather as text summarization and brainstorming engines. that, that back during World War ii, say for Richard Feynman to do his job as a calculating physicist at the Manhattan Project required 40 people, one, Richard Feynman to write down equations, and 39 women with.
punch calculators, manual punch calculators to actually do computations come the spreadsheet. You need only one guy to write down the equations. and you can run through huge numbers of extra scenarios. You can do much more calculations than could the 40 person anthology intelligence, that was Richard Feynman plus his 39, female employees back during the Manhattan Project.
It's a wonderful thing. It changes accounting, it changes analysis, it changes an awful lot of things because now you have a number processor that would actually really, works well. With the exception that you actually probably should really have become a real [01:00:00] computer programmer and done all your work in, FORTRAN or C because your Excel spreadsheet is impossible to debug and almost certainly has an error in it somewhere. but easy to use revolutionary entire back offices filled with people with calculating machine. Both calculators, mechanical calculators vanish overnight, just as with the coming of automatic switches. The 600,000 women who used to plug cables into plug boards to make telephone circuits work vanished over the course of 15 years or so.
now comes, the modern advanced machine learning models as natural language interfaces to text processors. To expand a sentence into a paragraph, to contract a paragraph into a sentence to brainstorm 55 ideas. they have one big problem, which is that the way they were initially trained was because they were a text demonstrator, a technology demonstration by open ai.
They trained them on the internet. So as a result, what you produced was something that was a pretty good model of the average internet troll, say a vibe poster, a meme poster, and you then, like, there's a huge industry, right? Andre Carpathy is especially eloquent on this, on all the things you have to do to a model to make it not just the level of the typical internet troll.
But instead useful, rag, RLHF, this is the, that's the others, the different acronym of the day. Prompt engineering, context engineering, such, et cetera. it's going to be quite useful, [01:02:00] because we can now poke it into places where it doesn't think, but rather it repeats elements of human thought in ways where you can quickly say yes or no.
And it has the possibility to get everyone up to what used to be a 90th level of facility with right 90th percentile, the level of facility for writing prose, which massively democratizes, the ability to make yourself persuasive. Yeah. it also, right. it also enables you to deal with information overload in a sense that you could not before.
Because it allows you to survey, at least if it can find it, a huge chunk of everything that has been said and compress it and distill it in such a way that it can read more than you could and a year in a day. And then we'll give you a starting point of what things to read more deeply in order to acquire the information and then to conduct, urinalysis.
I mean, when you think of it, we really don't need to build an artificial super intelligence. Some because we already have the anthology super intelligence that is the collective human mind as it has been building ever since the invention of writing 5,000 years ago. for the past thousand years, students who have gone to college have done so to learn how to be, Good draws on good users of good front end nodes to and in, the real super intelligence that exists today. The anthology, super intelligence of the collective human mind because individually we are not smart enough to know, remember where we left our keys last night. But as long as we can [01:04:00] draw on this anthology superintelligence of humanity, we can be very smart because we can draw on the best of all that has been thought and known.
The role of liberal education
DELONG: And, the purpose of a liberal education, of an education for someone who will neither not have a choice because they're a surfer, a slave, or not have a choice because there'll be a thug with a spear who has a place in the society of domination that is quite a nice one and doesn't need a different job if you're in the middle.
If you're in the middle, you have to make a living by your wits. And being a knowledge worker is the best way to do that. And by being a good one, by going to university and studying diligently, you both allow yourself to enrich your life by drawing on the best of all that has been thought and known. And also make yourself potentially useful to whatever thug with spear or their tame accountant, bureaucrat or propagandist has decided they need an extra person in the office.
Right? It'll hire you. That's the origins of the university, that continues. Information technologies change, they change radically. there's the question of whether modern advanced machine learning models as text processors are more like the, are relative to the old fashioned typewriter. More like the, say more like the Moog synthesizer is to the piano.
Or like the Moog synthesizer is to actually having a stick and a tin can, to bang it against. there's a question of how big a jump this is. This is a jump, but this is a jump that leaves the essentials of being a white collar worker intact. You have to survey the information and find the relative question that you, or that your bosses will want answered.
You then have to process the information and you then have to disseminate the information [01:06:00] both in the form of making it accessible so that the next person to start asking this question can draw on you as you have drawn on all the people who have written in the past. And also persuasive so that the people who need the information you have produced, who need the conclusions you've reached to act on them, actually believe you.
and you know this. Research question processing, answering, and then dissemination is still the same steps. And what you really need is you really need to demand from your professors an education that will let you do this for pros, especially in the age of information overload, in the same way that the data science tools that you do this for numbers.
so people going to college especially need to demand that they be properly educated to become good front end nodes to the anthology superintelligence. And if you do that, you are going to be able to do very valuable stuff no matter what your job is, no matter who your boss is, because it's the same kind of process of figuring out what is thought and known and how do I make, reach the right conclusions about it, and then make those conclusions.
Properly persuasive, both in fixed written form and in oral presentation or oral text chat form.
SHEFFIELD: Yeah. Well, and the interesting thing about all of this, to your point, about that the, the overall, a glutton nation of human knowledge over the past, 5,000 years of, writing, of literacy. Like it is also like we're at this moment now where the value of a liberal education.
has been, has never been greater because now there's so much uncertainty with regard to professions, [01:08:00] with regard
DELONG: kind of liberal education. It's the right kind of
SHEFFIELD: Well, yeah, but,
DELONG: It's the kind that makes you be persuasive rhetorically in person. It's the kind that makes you persuasive on paper, either by. Figuring out how to use AI in order to get your pros up to what used to be the 90th percentile or becoming sufficiently expert in using it to push yourself from the 97th to the 98th or the 99th percentile in the words you actually commit, to the screen.
And it's perhaps most of all, figuring out how to use these tools to do a better job for your research rather than just going to Wikipedia and regurgitating what you find there, which is not at all bad, right? Just not at all bad, but you know, how to become much better at actually finding what is the information I need to know to answer the question I ought to be asking, what is that question and how do I determine if this information is reliable?
SHEFFIELD: Yeah. Well what and
DELONG: those are things, that are not closely add to the data science skills. And those are things that are not that closely aligned with current college curricula.
SHEFFIELD: Yeah, I agree. No, there's, gonna have to be changes.
DELONG: yeah, like I was having lunch with my friend, computer sciences at Adam Quare, my friend from elementary school, Adam Qua yesterday here in Berkeley.
And, talking about things and similar things and him still pushing for his, education at St. John's, right? That these are the 80 most important big canonical books. We will read them deeply. and me saying, well, yes, this is an ideal education for the 14 hundreds when there really are only 80 worthwhile books that have been written.
[01:10:00] most of them in the Latin language. But these days you need to know how to be able to read shallowly and effectively, and to use all kinds of summarization engines as well, because, not all knowledge is gained by reading the 80 canonical books as deeply as you can. Even though say Martin Wolf and I were both enthus on the internet last month about how great a book Thomas Mons the Magic Mountain is and how important and valuable it is to read the Magic Mountain deeply.
these days, and in some ways these days more than any others, since the days of the 1920s when it was first published, that deep reading has a place, video has a place, audio has a place. shallow reading has a place, summarization has a place, brainstorming has a place. saying that using AI is cheating, is I think educational malpractice.
Saying we can't assign term papers because we don't know if they will actually write them, is also intellectual malpractice because you then call people in and have them talk to you for half an hour about their term paper. And by then, damned well how much of it is theirs in any real sense? in any real sense at all, getting people to learn deep reading, which is a very valuable skill, is I think a more difficult nut to crack.
and is one where the current youths are not that great in getting worse as time passes, and I don't have a good solution to that at all.
SHEFFIELD: Well, I guess that could be that, that, that's another episode then,
DELONG: another episode, another call me in a month, call me in a month, and I may have ideas about pedagogy in the way of ai. Then [01:12:00] behind that, there is the remaining use of modern advanced machine learning models, which is outside of text summarization and processing. Outside of trying to protect, spending tens of billions of dollars to try to protect your existing profits by giving people the best natural language interface to your services behind taking money from gullible investors in venture capital.
There also is extremely big data, extremely high dimension, extremely functional form regression and classification analysis of all kinds so that you no longer have to throw situations or people into say 15 different boxes and say, we treat things in this box this way and in this box that way. Instead allowing an algorithmic custom customization of how the complicated social system that is our society can deal with and react to individuals.
The future of AI applications
DELONG: So far, the only uses I have seen of this that are good on a large scale have been a bunch of mapping, a bunch of geographic uses of mapping things, and also Facebook claims that it's using it to massively increase its ad targeting. And thus it's ad revenue. Although whether this is doing this by showing people things that they will be happy they bought or hacking people's brains so they buy things they then regret is not a question that Facebook's minions are interested in giving us any information about.
But there will be additional mammoth use cases from very large, very big data, very large dimension, very flexible, functional form, large scale classification [01:14:00] regression and prediction analysis. We just do not know now where they will be valuable for us, but they will be see my colleague Marian Haw and her co-author Kieran Healy, and I think the book may be called The Algorithmic Society.
And since I have it upstairs on my bed, it's embarrassing that I'm too dumb to remember it. Say ought as an East African plains ape to have a better visual memory than I do. thus demonstrating we're really not that smart unless we can, as Isaac Newton said, stand on the shoulders of giants and draw on every good thought they have had.
SHEFFIELD: Yeah, exactly. And, it, you're reminding me here also that, there's a longstanding phrase within, techno reactionary circles that crypto is libertarian. Peter Thiel has said this, crypto is libertarian. AI is communist. And they seem to believe that because specifically that, the human corpus of knowledge does not seem to fit well with their anti-government, ideologies and, futile belief systems.
and that's why they're cons. Elon Musk is the great example, of course, constantly complaining about his, grok and other AI LLMs being woke. and, they never stop to think, well, why is that? is it because you, what if your epistemic processes are inferior? and, that's, it's, a subject.
I think that people who are not reactionaries, that we ought to be examining this question a lot more. I feel like.
DELONG: I don't know. it's, [01:16:00] no, it, is very odd, right? That you got, that in some sense. The root of their problem seems to be that they seem to believe that they deserve some kind of respect, right? that they are not getting. But, it's,
there's really no reason why they couldn't right? That. instead of trying, getting, making money by selling crypto grifts, mark Andreessen could be an evangelist for the web browser as a product category for democratizing access to global information. And do you know if he'd done that, he would be rich, not as rich as he is, but he would sleep easier because he would not have to think about, gee, all the people who put their money into these crypto funds, what's gonna happen to them and what's gonna happen to their customers?
and he'd certainly, be rich enough, right? That I've long thought that, that most of them would very much rather have been Steve Jobs, right? who on the one hand was an incredible asshole as a boss, but on the other hand was an incredible motivator and director of people who could take that and thought it was worth having a truly brilliant boss with his particular sense of taste and.
Brain that actually had a finger on the technological pulse of the [01:18:00] world, from his grabbing Steve Wozniak and saying, you are gonna build the apple one through the Apple, two through the Macintosh, through the, iPod through the iPhone, you on, through the Apple Watch and the AirPods, every single major revolution in how we interact with our technology, led by people motivated by Steve Jobs following his aesthetic psychological taste, and, so he's rich.
but he also was enormously respected and even loved, even though I know very, few people who would f found it any fun at all to work with him, or who thought he wasn't, especially. Good person except for the fact that his, I'm going to show you. Drive was focused. I'm, going to show you how to be about to make and ride a better bicycle for the mind.
they could be doing this right. Instead, they thought that they would get to a point where when they made the most money by being venture capitalists, they would've won something, and indeed they have won something, but it's not what they want, and they somehow feel they need to reorient the culture.
Sure. In a way that plutocrats like them get lots of respect. and then for some reason they think we should, they should all align themselves with kleptocrats to do things. Thinking that kleptocrats regard them as their friends, and actually kleptocrats regard them as their prey.
SHEFFIELD: And we saw that with Trump and Elon Musk, that as soon as Musk wasn't useful, he was the fall guy. That's all he
Musk, Trump and the tech elite
DELONG: Musk thought. I am [01:20:00] going to elect Donald Trump. I am going to give him more money than God has for him to win the presidency. And I am then going to, be the fall guy for actually cutting back on government spending by firing people. So Trump can then reverse whatever things I do he doesn't like.
But he starts from a point of view of the federal government that's been severely reduced in size and in return, I get an end to legal harassment of my businesses by the government that actually wants, say, SpaceX not to pollute as much as it does. I get, carve outs from tariffs so Tesla can continue to function as a profitable global value chain enterprise.
I get the, I get the tax credits for EVs continued and for Cafe continued. So Tesla is act not just a functional production network, but a highly profitable one that will actually support my claim to be the richest man out of live. And I get over time the entire NASA budget transferred to SpaceX. Musk thought he had gotten all those four things and had a deal with Trump and then Musk learned that he did not, the only thing he got was the end of what he regards as legal harassment of his companies.
And Trump is threatening to return that unless Musk shuts up. it's an interesting example of how somebody or other says everything Trump touches dies, and confess that I don't know how Musk is going to function going forward. I'm. Personal bankruptcy is definitely not outta the question, given his current situation and given the
SHEFFIELD: Yeah. Those, tax credits and the,
DELONG: Yeah.
SHEFFIELD: the, the, because so much of their revenue was coming from other car companies,
DELONG: Oh yeah. Oh yeah. Oh yeah. And also, and also the market, [01:22:00] right? They, a 300 mile electric vehicle in a United States with a very spotty, fast charging network is a luxury purchase. Right, the sweet spot. If you are going to be, if you are a, if you live in normal America, rather than say San Francisco or Berkeley and want a car, the sweet spot is a plugin hybrid.
both if you are worried about gas prices in the future as you should be, and if you want to do something for the environment, as you should as well, it's not an electric vehicle. Right? The electric vehicle was a bet made by environmentalists who wanted to declare allegiance to a forward looking, progressive future of new and high technology.
And, to the extent that Trump, and Musk are seen as at all alike as together waging war on the woke unquote, and to the extent that the environmentalism drops away, the. The market for Teslas vanishes as well. It has effectively vanished in Europe already. people would like EVs that have, more modern platforms, have better build quality, and also have companies with nicer, more pro-environment figures that their heads, and the market, for Dukes of hazard themed cyber trucks with confederate flags painted on their sides is very, small. And yet that seems to be the direction that Musk cannot help but keep pushing himself.
SHEFFIELD: Yeah, no, it will be interesting. And, he's, and I, it, is notable that he, seems to have, been trying to pivot toward X ai, which [01:24:00] purchased Twitter in a paper shuffling, recently. And so, we'll, see how he ends up going there with that. But, there, there is with regard to grok, that Grok keeps having these, right wing
DELONG: you know the point. Yeah. As we say the point, it's been trained, it's been trained on the typical internet, right? And the typical internet associates right wing beliefs with anti reality craziness. so as a result, you ask it to express rightwing beliefs and it then says what kind of words are linked to general Rightwing beliefs?
Well, calling myself Mecca Hitler is linked.
SHEFFIELD: Well, and
DELONG: and behold, here we're,
SHEFFIELD: Yeah.
DELONG: this doesn't necessarily, it's simply that yo the reality based community has a liberal bias, I think for pretty good reasons, means that if you are going to create. a reality-based chatbot liberal views are going to leak in and scrubbing it, is extraordinarily hard and must has not been willing to invest in hiring millions of people to write essays that are half propaganda and half true facts about the world, and use those as the training as the basic raw pre-training corpus for X ai.
SHEFFIELD: Well, and it's, interesting because like, so I'm working on a, a theory of, integrating dual process with, ai, which I can, share with you after we're done here. But, one of the things that, that I've discovered in working on this is that basically the psychological state of the right winger is the only difference between complete reactionary insanity and just milk [01:26:00] toast conservatism.
In other words, they have the exact same epistemology, the exact same epistemology, is what I believe is true, because I believe it. and they have no method of adjudicating between something that is completely insane and completely nor,
DELONG: And they're
SHEFFIELD: just of, a country club Republican.
Right-wing epistemology and AI
DELONG: That, Travis Kalanick going on a right wing podcast to say that my interactions with chat GPT have brought me to the frontiers of physics, and further advances are imminent, and the right wing hosts gull swallowing this.
SHEFFIELD: Yeah, exactly. That's what I'm saying. Yeah. Like they, they have no method of saying, well, what is truth? Independent of what I think. They don't have it.
DELONG: That that,
SHEFFIELD: when you
DELONG: indeed, as we say, we, we are pretty dumb. We cannot remember where we left our keys last night. And we also are primed by evolution to reach agreement because it's better that the whole band go down to the water hole at once than someone digging his heels and say, I'm not going down to the water hole until half an hour later.
This is too early. Right? The one who goes down half an hour later gets eaten by the leopard after all. and so
SHEFFIELD: Yep.
DELONG: are quite dumb, right? That, the ancient Germans, right, the ancient Germans. Actually believed that in some sense the rumble of the thunder was the rumble of the wheels of a giant cart in the sky drawn by two magical goats.
And in that cart there was a very large red figure with, a enormous, magical hammer named Ner, an anger management problems. And, you did not want [01:28:00] to deal with Thor when he was angry. Therefore, do not stand on top of a hill during a thunderstorm. Right. conclusion is quite reasonable. It is good not to stand on top of a hill during a thunderstorm, but rather the cower in a ditch, I understand seeing a lightning strike and saying, what could this possibly be besides ulner, the mighty ma massacre of miscreants. But then one goes on to wearing a hammer. Around one's neck and praying to the guy to help you and to sacrificing useful animals and burning their corpses for so he can smell the smoke and so forth.
we're easily misled even when the basic cognitive message don't stand on top of a hilltop during a thunderstorm is actually one that is very useful to know. And without actually a theory of electromagnetism kind of, it's Thor, he has a magic hammer and big anger management problems when he is driving fast is as good an explanation as any to lead you to the conclusion it's memorable and you know it's leads to good actions.
SHEFFIELD: Yeah, it's serviceable for a small, achievement, but not for anything bigger, basically.
DELONG: On the other hand, Chris Hemsworth is truly excellent as the character. and do you know,
SHEFFIELD: Yeah. Well,
DELONG: very much, I'm very much on a Thor on the Thor side of the Thor versus Loki. I think Thor was better played than Loki in that.
SHEFFIELD: well, I, we'll, have to let people sort that [01:30:00] out in the comments.
DELONG: Okay. Alright.
SHEFFIELD: all right. Well, Brad, so, for people who want to keep up with your stuff, what, where, what's your advice for them? You got any books you're working on or,
DELONG: brad DeLong substack.com also my 2022 books vouching towards Utopia. The economic history of the long 20th century, I think are both supremely great. and the market appears to agree, as well or at least enough of the literate market, that it makes me very grateful, for my 50,000 Substack subscribers, more books.
We'll see what I actually managed to accomplish this fall. I think there are four things that I might push across the line well enough to send out to publishers to see who wants to actually publish it, but thank you very much for inviting me to theory of change
SHEFFIELD: All right. Sounds good. All right. Thanks for being here.
DELONG: You.
SHEFFIELD: Uh, let's push stop here. All right. Ooh, so.
Share this post