WEBVTT
00:00:00.299 --> 00:00:04.790
There is at least one super niche AI tax that I would like, which is, I don't know.
00:00:04.910 --> 00:00:06.775
I'm sure you guys have seen this, is that there are.
00:00:07.655 --> 00:00:12.095
podcasts and as a podcast host, I find myself vaguely concerned about this.
00:00:12.095 --> 00:00:18.394
We created an outline for this podcast and I was like, oh man, I wonder what would that podcast would be like if we fed it to them?
00:00:18.815 --> 00:00:21.425
And I would like, maybe we just need to tax that.
00:00:21.425 --> 00:00:29.135
Like all AI podcasts get heavily taxed and we maybe just push those down for a little while so I can at least retire.
00:00:29.195 --> 00:00:29.285
Yes.
00:00:30.214 --> 00:00:30.964
But that's actually a real.
00:00:32.107 --> 00:00:33.728
but I, I think we can compete.
00:00:33.728 --> 00:00:37.148
I, I think people will like us better than an AI SLAP podcast.
00:00:38.759 --> 00:00:39.420
I'd like to think so.
00:00:40.094 --> 00:00:44.689
I believe, Kyle, you have many, many valuable skills that would, would make you, um.
00:00:45.045 --> 00:00:52.770
Valuable, uh, uh, uh, contributor in other forms if such, if such a, a technology, um, was spread.
00:00:53.895 --> 00:00:54.304
Thank you.
00:00:54.424 --> 00:00:55.325
I appreciate that.
00:00:56.374 --> 00:00:59.643
hello and welcome to the Deduction of Tax Foundation Podcast.
00:00:59.643 --> 00:01:02.884
I'm your host, Kyle Houlahan, and we are back today with another episode.
00:01:02.884 --> 00:01:09.754
And I'm joined by my co-host, Erica York, and we're joined by Alex Muresianu, senior Policy Analyst at the Tax Foundation.
00:01:10.013 --> 00:01:10.763
all right, so here's the deal.
00:01:12.548 --> 00:01:14.257
AI is everywhere right now.
00:01:14.528 --> 00:01:17.378
It's like, I think you guys saw the Super Bowl, right?
00:01:17.617 --> 00:01:20.707
Like the a, it was all ai, which is kind of crazy.
00:01:20.707 --> 00:01:23.347
It's in the headlines every day, all the time, 24 7.
00:01:24.367 --> 00:01:26.617
but now it's part of the tax policy conversation.
00:01:26.947 --> 00:01:36.578
And so, you know, Alex, I'm wondering, know, do people need to know about this and, and where the debate stands on taxing ai.
00:01:36.968 --> 00:01:42.643
Yeah, so I think there are a lot of different angles that people, uh, um.
00:01:43.367 --> 00:01:46.158
Come to the AI policy conversation.
00:01:46.608 --> 00:01:46.908
Um.
00:01:47.522 --> 00:01:52.203
And there are a lot of different proposals out there that we can, we can get into.
00:01:52.203 --> 00:02:00.783
But generally the angle that at least I'm sort of most interested in is a sort of purely economic one.
00:02:01.263 --> 00:02:12.302
Uh, about how AI will, uh, or will not, uh, change the future of work and how it affects sort of basic, uh, uh.
00:02:12.992 --> 00:02:16.473
How we produce goods and services, sorts of questions.
00:02:16.473 --> 00:02:20.913
And also what does that mean for, for, uh, employment in the way, uh, work works.
00:02:21.362 --> 00:02:38.883
Um, and so I think the, the driving concern, uh, for a lot of these proposals relates to how, um, artificial intelligence will change, uh, or potentially replace, uh, human workers in various jobs.
00:02:39.242 --> 00:02:40.832
Uh, and, uh.
00:02:42.198 --> 00:02:48.407
Usually that's the focus of policy proposals, uh, to specifically target ai.
00:02:48.798 --> 00:02:57.798
The point that I, I think I'll, I'll probably make a, a couple times in various forms, is that like a lot of things that are AI might already be taxed.
00:02:57.798 --> 00:03:00.467
In other words, if you're thinking about profits from ai.
00:03:02.237 --> 00:03:04.367
Those are covered by the corporate income tax.
00:03:04.367 --> 00:03:22.758
you know, the, the particular concern here is, is sort of bespoke, uh, policy making that's sort of targeting, uh, whether it's, uh, investment in, you know, data centers or just, or revenue perhaps from AI products, no, I'll use the word again.
00:03:22.788 --> 00:03:26.657
Bespoke, bespoke policy specifically related to, um.
00:03:27.527 --> 00:03:28.758
Yes, it's a good word.
00:03:28.758 --> 00:03:29.268
It's a good word.
00:03:29.268 --> 00:03:39.828
Although I felt like saying it again, reminded me that there's a Veep episode where everybody uses the word, um, robust, um, in like three different meetings and then it becomes like a news story.
00:03:39.828 --> 00:03:42.437
So what are people in particular most worried about?
00:03:43.276 --> 00:03:49.967
A lot of science fiction stories, of course, uh, uh, talk about how machines will, will, like replace humans.
00:03:49.967 --> 00:04:10.516
And in the most extreme case, that has been covered in a few sort of, um, theoretical economics papers, uh, recently, and it is an interesting sort of theoretical exercise, at the very least, um, is that artificial intelligence will replace like all human labor as like an input in the economy.
00:04:10.997 --> 00:04:22.697
That's a, a sort of pretty extreme example, and I think people have predicted that technology would end, you know, uh, uh, human labor for a very long time like that.
00:04:22.697 --> 00:04:28.187
That's a, a, a prediction that has kind of failed, um, many, many times before.
00:04:29.026 --> 00:04:38.656
Um, to be fair, you only have to be right once, uh, for that one, but at a slightly less extreme level.
00:04:38.836 --> 00:04:58.276
The concern is more that that large portions of, uh, what people might call, like knowledge work or, or white collar employment, um, will be, will be replaced because all this technology, um, makes the work sort of much, much easier to, to, uh, a finish.
00:04:58.276 --> 00:05:04.733
And that this rapid productivity increase will replace, uh, a lot of human workers and the.
00:05:05.033 --> 00:05:11.963
The thing is that like, well, at the very high level, we don't know what the future holds.
00:05:12.204 --> 00:05:24.504
Um, but at a more specific level, we don't know how technology that makes a sector or, uh, a certain type of work, um, more productive.
00:05:25.043 --> 00:05:25.704
More productive, or.
00:05:26.338 --> 00:05:27.269
More efficient.
00:05:27.478 --> 00:05:33.059
We don't actually know what that means for employment, uh, in that particular sector.
00:05:33.389 --> 00:05:40.798
Um, there are story that there, there are sort of two, uh, terms to think about here, uh, that people sometimes use.
00:05:40.798 --> 00:05:45.629
One's bowel, mals cost, uh, cost disease, and the other one's Jevons Paradox.
00:05:45.959 --> 00:05:48.899
Um, those are just, just term good, good terms to know.
00:05:48.899 --> 00:05:53.699
But like the, the bowel mo cost disease story, is that like as.
00:05:54.298 --> 00:06:00.091
Some sectors get more productive, they kind of shrink, as a share of employment.
00:06:00.451 --> 00:06:06.752
fewer people work in them because new technology means you can get like a lot more, uh, uh, done with fewer people.
00:06:06.752 --> 00:06:09.182
So like a good example here might be like agriculture.
00:06:09.512 --> 00:06:17.072
Um, like we've gotten a lot more productive in producing, uh, uh, food and.
00:06:18.107 --> 00:06:25.966
The share of people who work in agriculture has declined dramatically, uh, over the past, you know, 200 years.
00:06:26.206 --> 00:06:29.447
That's like one end where we're getting better means fewer workers.
00:06:30.226 --> 00:06:34.487
The other end of the spectrum, I think would be like software or data analysis.
00:06:34.487 --> 00:06:54.863
I think where the technology for doing data analysis is so much better than it was in say, 1970 We have way more people who like do sort of data analysis or software, uh, now than did back in the 1970s as like a very direct example.
00:06:55.314 --> 00:07:04.084
Um, Tax Foundation has like software engineers, like I do not think that we had that in the eighties.
00:07:04.384 --> 00:07:06.694
Uh, for instance, like baseball teams.
00:07:07.144 --> 00:07:07.894
Baseball teams.
00:07:07.894 --> 00:07:22.954
Now, every baseball front office has like an analytics department where they have a bunch of, uh, uh, uh, you know, uh, stack geeks pouring over these sort of advan new advanced metrics that didn't even exist, um, you know, 30 or 40 years ago.
00:07:23.314 --> 00:07:29.853
Uh, and so what it boils down to is like how much more of something people want.
00:07:29.853 --> 00:07:34.564
Sometimes new technology opens up new markets and new applications.
00:07:34.983 --> 00:07:44.374
For certain types of work and that that means more employment in that sector and sometimes it replaces work because people don't actually want that much more of whatever this stuff is.
00:07:45.064 --> 00:07:57.634
So ultimately that's to say that one things are ambiguous, uh, and, and in terms of how productivity growth affects employment in a particular sector.
00:07:57.934 --> 00:08:02.014
And two, the big picture claim that.
00:08:02.853 --> 00:08:13.113
All of human labor would be replaced is quite an extreme one, and it would make sense to think about AI in more of the terms as other periods of technological change.
00:08:13.492 --> 00:08:23.781
so we're at this like early stage where we don't know how this new technology is going to affect the labor market, but there are a lot of fears that it could be extremely disruptive.
00:08:24.201 --> 00:08:29.122
We've got senators, we've got some economists, we've got even AI company CEOs.
00:08:30.307 --> 00:08:35.017
Proposing these really targeted taxes that could try to slow adoption.
00:08:35.917 --> 00:08:38.797
is that kind of getting ahead of where we need to be right now?
00:08:38.797 --> 00:08:49.236
What's the evidence for like labor market disruption today and does that justify like new tax provisions targeted AI right now?
00:08:49.236 --> 00:08:49.897
Or what?
00:08:49.927 --> 00:08:52.356
What should we be thinking of and what, what do we actually know?
00:08:52.626 --> 00:08:54.187
Like today in 2026?
00:08:54.513 --> 00:09:08.852
there are a whole host of, um, uh, labor market sort of indicators that one could, could comb through and, and probably put together a, a compelling narrative once you've gone through enough of them to say that something, something, you know, big is happening.
00:09:08.852 --> 00:09:09.003
But the.
00:09:10.352 --> 00:09:16.393
Big picture for most of the headline numbers is, not indicating a mass wave of, of automation.
00:09:16.783 --> 00:09:24.043
Um, the, you know, unemployment rate is lower than it was for the vast majority of the 2000 tens.
00:09:24.432 --> 00:09:31.692
Uh, labor force participation is at, you know, close to a, a, you know, 20 year high.
00:09:32.113 --> 00:09:32.653
Um.
00:09:33.087 --> 00:09:41.038
You, you know, so those are like the most basic high level, um, uh, employment indicators.
00:09:41.337 --> 00:09:59.158
Um, the one that people sort of most often, um, you know, rattle, the rattle the SARE about is the low rate of like entry level white collar work or, or like low, um, job hirings particularly, and, and software.
00:09:59.488 --> 00:10:01.168
Um, but the.
00:10:02.158 --> 00:10:11.998
Sort of confounding factor here is that there was this sort of massive hiring boom in those sectors in, in like 2021 into 2022.
00:10:12.717 --> 00:10:13.857
Um, and.
00:10:14.337 --> 00:10:38.217
The sort of low hiring rate is mostly I, I think a reflection of those, um, decisions that, that, you know, you had, um, you pulled a bunch of hiring to 20 21, 20 22, and now you're hiring last because you walked all those people in and it's, um, not confirm morale to just lay everybody, uh, uh, off if you, you were hired.
00:10:38.217 --> 00:10:40.648
So you just avoid hiring new people.
00:10:40.738 --> 00:10:42.613
Um, and so I.
00:10:43.182 --> 00:10:46.783
Like, like I sort of said, about like uncertainty.
00:10:46.783 --> 00:11:05.273
Like this is not like proof that, that some massive wave of automation could come in the future, but to say like, it's here and we need to like design some crazy new policy because, uh, uh, you know, the, the writing is on the wall for all this stuff.
00:11:05.423 --> 00:11:09.442
I think that is a, a pretty extreme reaction to existing evidence.
00:11:10.351 --> 00:11:11.422
Yeah, and, and obviously.
00:11:12.581 --> 00:11:14.591
We don't wanna completely invalidate the fear around this.
00:11:14.591 --> 00:11:19.211
This is obviously, is a very new thing and it's a very real massive change in all of our lives.
00:11:19.572 --> 00:11:21.522
Um, and so that, that makes sense.
00:11:21.522 --> 00:11:30.761
I understand some of the fear, but, but what do you see as like the, what are the main tax proposals out there right now and, and who or what is driving those?
00:11:31.278 --> 00:11:43.668
there's a, a category that I think of as explicitly targeted at the employment issue and or the employment issue, I think is the, the driving focus of a lot of them.
00:11:43.878 --> 00:11:48.707
The intent, uh, of the policies might vary.
00:11:48.918 --> 00:11:51.768
So, uh, the two, uh, um.
00:11:52.472 --> 00:12:09.302
Proposals that I highlighted in the, uh, piece, uh, for tax foundation that I wrote are on the one hand, uh, Senator Bernie Sanders has proposal to, um, both tax like AI profits and also, uh, to.
00:12:09.947 --> 00:12:14.118
Uh, get rid of deductions for, uh, investment in equipment.
00:12:14.118 --> 00:12:18.587
So that's covering physical automation as well as sort of digital automation.
00:12:18.798 --> 00:12:24.197
And, you know, the Sanders perspective is explicitly like, this is a problem.
00:12:24.197 --> 00:12:29.597
This is going to, you know, replace workers and like, we need to stop.
00:12:29.972 --> 00:12:39.452
The, uh, uh, replace, you know, you know, we, we just need to, to penalize these investments to, to prevent them from happening and prevent the employment disruption from happening.
00:12:40.592 --> 00:12:46.442
Then the sort of other perspective is not as, uh, sort of ideologically like motivated.
00:12:46.682 --> 00:12:47.852
It's more oh.
00:12:48.677 --> 00:12:55.817
It's good that we have this new technology, but it will come with downsides or, or, or some people will be disrupted.
00:12:55.822 --> 00:13:04.067
Or, or, or, or, you know, you, you might have some sector level unemployment and the proposed answer is to sort of fund programs, uh.
00:13:04.977 --> 00:13:09.837
Four people who are, are disrupted through some sort of tax targeted at the industry.
00:13:10.227 --> 00:13:14.337
Um, this is the, uh, Senator Mark Kelly proposal.
00:13:14.668 --> 00:13:21.227
Um, and the details of, of that proposal, at least when I, I wrote the piece.
00:13:21.227 --> 00:13:24.587
Maybe there's something that has come out that's more fleshed out that I, that I missed.
00:13:24.888 --> 00:13:31.368
Um, but his proposal, basically some sort of tax on, on industry revenue.
00:13:31.727 --> 00:13:32.148
Um.
00:13:33.038 --> 00:13:39.217
And so I think Kelly's thought process is a little bit different and a little bit more reasonable.
00:13:39.398 --> 00:13:50.258
However, I wouldn't be confident that the tax designed to just ameliorate some of the issues wouldn't itself disincentivize a new investment.
00:13:50.768 --> 00:13:59.557
Um, you know, again, the details aren't out there, but just because your intentions are a little bit different doesn't mean that you wouldn't have the same effect of, of slowing down a technology adoption.
00:14:03.075 --> 00:14:05.220
Sometimes when we, we hear.
00:14:06.044 --> 00:14:08.144
Proponents of of AI taxes talk.
00:14:08.144 --> 00:14:16.934
And this has parallels to the way, like other conversations go on, like the, the trade policy space and what do we do for the manufacturing sector.
00:14:16.934 --> 00:14:22.004
And there's kind of this like long thread of thinking that the tax code.
00:14:22.075 --> 00:14:24.235
Favors capital over labor.
00:14:24.235 --> 00:14:32.815
And so we need to, you know, change policy to write that, to make sure that we're not incentivizing investment here over investing in workers.
00:14:32.995 --> 00:14:36.144
Can you speak to that and, and what's actually the case there?
00:14:37.937 --> 00:14:52.278
Yeah, so you know, the biggest picture, uh, illustration of this is, uh, how companies deduct certain kinds of costs, um, and.
00:14:53.743 --> 00:15:04.062
And you know, just about, just about every case, uh, companies deduct the cost of like worker salaries immediately, our wages and salaries, uh, immediately.
00:15:04.393 --> 00:15:08.052
Uh, while often they have to spread the.
00:15:08.878 --> 00:15:14.008
Costs of, of capital investments, uh, out over time.
00:15:14.398 --> 00:15:27.087
Um, now there are some provisions that are, are currently in effect in, in, in us, uh, tax law, na, many of which are relatively new, that allow for the, uh, full deductions for investment.
00:15:27.477 --> 00:15:29.763
Um, the, uh, uh, introduction of.
00:15:30.493 --> 00:15:36.373
A hundred percent bonus depreciation for equipment, um, expensing for research and development.
00:15:36.373 --> 00:15:40.993
Immediate deductions for r and d investment that has been brought back into law.
00:15:41.322 --> 00:15:41.802
Um.
00:15:42.207 --> 00:15:49.707
But those just put those investments on a sort of level playing field with, um, you know, worker compensation.
00:15:50.067 --> 00:15:56.008
Um, and not all investment is eligible, uh, for those provisions.
00:15:56.337 --> 00:16:05.998
Uh, and so if anything on the whole, um, that tax system kind of dis favors a, a, a capital investment.
00:16:06.447 --> 00:16:10.528
Um, but I think the, the point about how.
00:16:12.822 --> 00:16:39.102
This links to other debates about like displacement, uh, of which, which is usually wrapped up in, in at least historically related to manufacturing jobs or, um, uh, you know, I know our, our, uh, our producer talked about, uh, our produ, our great, great producer Dan, uh, talked, uh, talked to us in our, our prep call about, um, history of the coal industry in West Virginia.
00:16:39.177 --> 00:16:39.437
You know.
00:16:40.092 --> 00:16:55.903
There are, uh, uh, precedent for or localized disruptions, uh, uh, to, uh, um, uh, employment and, and, you know, general economic prospects related to economic change and technological changes.
00:16:56.322 --> 00:16:56.773
Um.
00:16:57.373 --> 00:16:59.682
But those are, are sort of unprecedented.
00:16:59.682 --> 00:17:07.603
And the, the answer wouldn't be just to try and like slow down change it would be to figure out how best to, to ameliorate it.
00:17:07.633 --> 00:17:11.413
We have, you know, some experience with that and, uh, to, to think through.
00:17:11.413 --> 00:17:12.702
There are lessons to be learned.
00:17:13.031 --> 00:17:20.652
So rather than slow the adoption of new technology, find ways to directly help the workers who may be impacted.
00:17:20.682 --> 00:17:23.922
Rather than saying, no, we're gonna freeze technology in place.
00:17:23.922 --> 00:17:29.112
It, it reminds me of a line from a paper, um, that was looking at, you know, the.
00:17:29.846 --> 00:17:35.576
The decline in manufacturing jobs and looking at whether technology or trade was a bigger contributor.
00:17:35.576 --> 00:17:47.906
And this particular paper concluded it was technology and it made the argument and has the line, like if you wanna go back to higher levels of manufacturing employment, pretty much the only policy solution is to smash the machines.
00:17:48.146 --> 00:17:48.807
Um, so I don't think.
00:17:50.007 --> 00:17:55.227
We should be at a point where we're saying, yes, smash the machines or tax the machines to slow adoption.
00:17:55.586 --> 00:18:05.787
It should be more of an openness to innovation and to growth, also acknowledging that there are other policy tools maybe that, that we could look at.
00:18:06.067 --> 00:18:12.876
a lot of this feels like it's, it's largely based on, there's a lot of fear around how, what will happen to labor.
00:18:13.057 --> 00:18:16.057
Most of this, like most of this conversation that we're having is fear around labor.
00:18:16.057 --> 00:18:26.916
So how could reforming maybe, you know, unemployment insurance or raising some of those caps or worker training deductions or something, you know, how could that maybe, you know, help,
00:18:27.366 --> 00:18:44.136
I would say the sort of best way in, in principle to address, um, the disruption that introducing a new technology, uh, would have is to make it easier for people to.
00:18:45.606 --> 00:19:02.436
Transition to doing something different, uh, or in a different, and, and put in a different way, sort of lean into different skills, uh, or, or, or applications for their, their, uh, you know, alternative applications for their existing sets of, of skills.
00:19:02.856 --> 00:19:05.826
we're the tax foundation I'm focused, focused on, on tax angles.
00:19:05.826 --> 00:19:15.487
And I don't mean to say that this is a, a comprehensive policy menu, but I think these are the, the relevant policies in the tax code to think about.
00:19:15.846 --> 00:19:24.007
Um, the first is unemployment insurance, um, unemployment insurance in the United States.
00:19:24.037 --> 00:19:24.892
Uh, the tax rates.
00:19:25.717 --> 00:19:33.186
Are, uh, uh, you know, vary by employer and relate to their experience with the unemployment system.
00:19:33.186 --> 00:19:43.416
And the sort of justification here is that, and, and, and you know, your rates go up if you have more unexperience with the, the unemployment system, which is means that like, you know, you've.
00:19:43.957 --> 00:19:50.436
Fired people, uh, uh, before, and, you know, the, the thinking there feels very, very intuitive.
00:19:50.436 --> 00:19:57.517
It's like we don't want people to, um, be, be laid off and so will tax companies more that are, are laying people off.
00:19:57.517 --> 00:19:59.136
But I think the, the.
00:19:59.811 --> 00:20:17.872
Real effect here is that people are, are, are firms are like more worried about, uh, risky hires than, uh, uh, they would otherwise be because they know that they're, if, if somebody, you know, oh, this, this, this candidate has a lot of potential, but also.
00:20:18.487 --> 00:20:20.406
Might really not be a fit for this role.
00:20:20.676 --> 00:20:25.057
Um, they know that they may well face a tax penalty in the future if the, that just isn't right.
00:20:25.446 --> 00:20:30.846
Um, and so I think that's a, a particular concern for people who are looking to change careers, uh, or.
00:20:31.176 --> 00:20:39.457
Looking to move from, from something, um, uh, that, you know, might only be tangentially related to what they previously worked in.
00:20:39.457 --> 00:20:46.146
And they might, you know, think that, that this, this, you know, I, I really couldn't apply my skills, uh, in this other, other career path.
00:20:46.567 --> 00:20:54.906
Um, and so, you know, if you're worried that new technology that is coming down the pipe will, um.
00:20:55.446 --> 00:21:03.247
Make certain forms of, uh, certain skills or certain particular occupations sort of, uh, largely automated away then.
00:21:04.402 --> 00:21:16.672
This policy, the, the basing, the unemployment, uh, uh, tax rates on experience rating could sort of dissuade companies from taking a cha taking chances on people who are looking to, to shift careers.
00:21:17.092 --> 00:21:28.432
Um, the other, uh, uh, angle would be how, uh, of a tax code treats, uh, company investment in, uh, worker training, and particularly training for new roles.
00:21:28.761 --> 00:21:31.731
Um, there's a cap on the deduction for.
00:21:32.261 --> 00:21:34.481
Investment in training for existing roles.
00:21:34.541 --> 00:21:37.902
It's a non-deductible, uh, for training in new roles.
00:21:38.412 --> 00:21:41.896
part of the reason I guess, uh, to, to like justify why that policy would.
00:21:42.876 --> 00:21:43.507
Where it is.
00:21:43.507 --> 00:21:51.876
Like there are you, you do, you do sometimes worry in, in the tax code of having sort of non uh, uh, taxed, uh, worker compensation.
00:21:51.876 --> 00:21:57.836
So I think that that's the sort of original like thought process of, not having ways to do like non wage compensation.
00:21:57.836 --> 00:21:59.007
That's, that's UNT taxable.
00:21:59.277 --> 00:22:00.297
Um, but.
00:22:01.811 --> 00:22:11.172
The effect of not having this, this deduction available is that there isn't, uh, much of an incentive for companies, uh, or the very least there.
00:22:11.561 --> 00:22:28.676
There's a disincentive, I guess, for companies to retrain workers who may be, are no longer, um, you know, productive or necessary in their original role, but may still be like, you know, you think you're, they're like solid workers who, who are like diligent and will.
00:22:29.336 --> 00:22:33.057
Um, might be fits for alternate employment in the company.
00:22:33.116 --> 00:22:41.967
And so, um, those are two sort of option, the employment insurance and tax treatment of worker training or retraining.
00:22:42.416 --> 00:22:43.916
Um, that.
00:22:44.906 --> 00:22:49.767
Those are like t like, like policy changes in the regular tax toolkit.
00:22:50.217 --> 00:23:08.696
Um, that would be sort of sound structural improvements sort of regardless of what, uh, uh, happens with, uh, uh, artificial intelligence, which again, there's a very wide spectrum of outcomes and like if this is a whole uh, uh, you know, nothing burger, then it would still be good that we made those changes.
00:23:09.476 --> 00:23:12.615
Real quick, just to kind of follow up to make sure I understood this correctly.
00:23:12.645 --> 00:23:13.830
It's kind of like based.
00:23:14.925 --> 00:23:28.336
The way you're setting this up is like, okay, a company X could invest in said software and deduct that, but they can't deduct the same thing for investing in people at that work for them.
00:23:28.383 --> 00:23:30.055
So they can take a deduction.
00:23:30.970 --> 00:23:37.210
It's up to, I think,$5,250 for, uh, like worker training.
00:23:37.660 --> 00:23:45.309
Um, but uh, that is supposed to be sort of related to their, uh, uh, current role.
00:23:45.609 --> 00:23:49.750
Uh, and like full sort of retraining, uh, is non-deductible.
00:23:50.076 --> 00:23:55.836
there is a justification here where you don't want to have, um, uh.
00:23:56.527 --> 00:24:06.876
Things that are effectively a form of employee compensate or might kind of be a form of employee compensation that's deductible at the firm level and also non-taxable at the individual level.
00:24:06.876 --> 00:24:10.386
That's like a common concern at the margins of tax policy.
00:24:10.747 --> 00:24:15.126
Um, but in this particular case, you know, you could have some sort of cap.
00:24:15.846 --> 00:24:26.571
Uh, uh, it might not be unlimited, but, uh, having, you know, more than, than zero, uh, uh, you know, uh, uh, limit, um, could, could be effective.
00:24:27.057 --> 00:24:31.438
this reminds me of a book that Glenn Hubbard wrote in like 2022.
00:24:31.498 --> 00:24:37.827
So before AI was really being discussed, but the book is called The Wall on the Bridge, and it's like policy ideas.
00:24:37.877 --> 00:24:41.958
To help build bridges for people to respond to things that are disruptive.
00:24:42.377 --> 00:24:49.907
So a lot of the ideas he talks about in there, they expand beyond tax policy, but like unemployment insurance reform, rethinking how social insurance works.
00:24:49.907 --> 00:25:00.917
So if it is a technological disruption that's causing hardship for people, don't block the disruption, like bridge people to deal with that in a, in a productive way.
00:25:00.917 --> 00:25:03.587
So I think like those ideas you, you mentioned are.
00:25:04.343 --> 00:25:06.532
A perfect fit with that kind of concept,
00:25:06.689 --> 00:25:13.798
was just realizing it's like I, I think our bookshelves are, are radically different, what you have on your shelf and what I have on my shelf.
00:25:13.798 --> 00:25:16.769
I was like, I don't think I have any economic books like.
00:25:17.527 --> 00:25:24.457
I was like my, one of the, my favorite books I've read, and I talk about that book all the time because it's so applicable, whether, whether people are upset about like.
00:25:24.906 --> 00:25:28.537
trade policy disrupting people or new technology disrupting people.
00:25:28.537 --> 00:25:34.326
Now AI disrupting the the workforce, there's always going to be some kind of disruption happening.
00:25:34.717 --> 00:25:36.487
The answer isn't stop disruption.
00:25:36.487 --> 00:25:40.416
It's let's be more agile in response to disruption.
00:25:40.645 --> 00:25:50.635
So bigger picture when we're thinking about like the federal budget situation and the sustainability of debt, we all know that it's unsustainable.
00:25:50.816 --> 00:25:54.865
Deficits are already really large, and one of the things people talk about with AI is what.
00:25:55.675 --> 00:25:58.256
What will the impact be on the federal budget?
00:25:59.066 --> 00:26:01.465
it going to make the deficit situation worse?
00:26:01.496 --> 00:26:13.346
Um, you know, if these labor doomsday scenarios come, come into play, do income tax revenues go in the tank or will it help growth and help reduce the budget deficit?
00:26:13.615 --> 00:26:16.435
Where, where do you think the, the truth of the matter is?
00:26:16.884 --> 00:26:17.275
Yeah.
00:26:17.275 --> 00:26:25.224
So beyond the like first order answer, which is, I don't know, uh, which I, I think is, is like.
00:26:26.529 --> 00:26:28.240
The most a hundred percent accurate answer.
00:26:28.240 --> 00:26:28.569
It's true.
00:26:28.630 --> 00:26:30.099
I like, I don't know the future.
00:26:30.430 --> 00:26:42.700
Um, and I think this area, like in in particular, um, like people, people don't know, um, exactly what the effect of this, this new technology will be on all kinds of things.
00:26:43.119 --> 00:26:49.509
Um, but I think the, like putting a whole lot of belief in the strong versions.
00:26:49.855 --> 00:26:55.434
Of either prediction, I, I, I think is, is, uh, um, a mistake.
00:26:55.855 --> 00:27:18.025
Um, there is this sort of Optimus story where the, you know, great that, that AI will unleash like great productivity growth and that will, uh, um, increase, uh, tax revenue and that will, you know, solve a lot of structural, uh, the long-term sort of structural problems, uh, in the us uh, fiscal picture.
00:27:19.255 --> 00:27:42.025
And then the sort of negative, uh, uh, um, sort of pessimistic story that, you know, the, the labor disruption means that the individual income tax, which is the sort of workhorse of the, uh, us uh, uh, federal revenue, um, you know, if revenues from that and then, you know, payroll taxes too, you know, go through the floor.
00:27:42.025 --> 00:27:45.025
Now we've gotten even worse, uh, a fiscal situation.
00:27:46.075 --> 00:27:47.184
I, uh.
00:27:48.444 --> 00:28:00.775
If anything, you know, generally you'd expect if AI like improves productivity growth, you'd expect that to directionally, uh, uh, positively contribute to the US fiscal situation.
00:28:01.045 --> 00:28:05.845
Just generally, um, we expect productivity growth to mean like more, more revenue.
00:28:06.265 --> 00:28:15.680
Um, on that, there are a bunch of different like little interactions where, based on how certain programs are calculated, but on net this is a, a deficit generally like.
00:28:16.654 --> 00:28:19.055
Productivity growth should mean like reduced deficits.
00:28:19.595 --> 00:28:33.569
Um, but the promise of dramatically improved productivity growth as a solution to uh, uh, long-term fiscal problems is just like.
00:28:34.444 --> 00:28:38.674
A classic like siren song of, classic wishful thinking.
00:28:39.035 --> 00:28:51.035
Um, and I certainly wouldn't, you know, count on that or, or have a lot of saying, you know, oh, well, well AI's on the way, we'll have this, this incredible productivity growth and you don't really have to worry about, but have a set.
00:28:51.394 --> 00:28:53.015
Um, I would be very wary of that.
00:28:53.535 --> 00:28:58.244
particularly because baseline budget protections already include productivity growth.
00:28:58.244 --> 00:29:05.595
Like there's some productivity growth already in forecasts and that's gotta be coming from somewhere.
00:29:05.684 --> 00:29:16.664
And like, maybe AI is like good for, for productivity growth, but that's just all the stuff that's in the budget, you know, forecast and economic outlooks already.
00:29:17.119 --> 00:29:28.700
the other thing with the pessimistic story that, that, you know, all of a sudden labor is going to be replaced, or the very least, you know, depleted, uh, substantially.
00:29:30.095 --> 00:29:32.134
Thanks to AI technology.
00:29:32.404 --> 00:29:45.545
Now, this is, uh, uh, where it makes sense to think about things in a, in a historical context and not sort of, assume AI is dramatically different from historical examples of technological change or, or disruption.
00:29:45.545 --> 00:29:58.384
And when you look at the, the labor share of income, you know, measured properly, uh, over the course of, you know, the 20th century into the 21st, so far, labor share of income is quite stable.
00:29:58.789 --> 00:30:12.365
Um, and so assuming that things will be dramatically different and that will sort of warp, uh, the, the structure of, uh, uh, tax revenue, uh, I, I, I don't think that's a, a sound bet.
00:30:12.803 --> 00:30:19.647
That measured properly point that you just mentioned is like the key because like today there are some charts going around with.
00:30:20.548 --> 00:30:23.488
share of gross income rather than labor share of net income.
00:30:23.728 --> 00:30:27.208
And if you look at the gross income picture, it does look like labor share is declining.
00:30:27.387 --> 00:30:32.607
But the real big story there is like depreciation and, and taxes have have gone up.
00:30:32.938 --> 00:30:37.708
Um, and of course, depreciation and taxes aren't income that actually goes into anyone's pockets.
00:30:37.857 --> 00:30:42.988
So you wanna net out those things that income to anyone, and then look at how it's split between capital and labor.
00:30:43.228 --> 00:30:44.968
And there you do see this really.
00:30:45.718 --> 00:30:47.248
Really stable share.
00:30:47.248 --> 00:30:51.208
Um, you know, some, some slight changes since the eighties, but nothing dramatic.
00:30:51.208 --> 00:30:55.768
It's really close to that average, and it has been since, like you said, the, the 1930s.
00:30:56.097 --> 00:31:00.373
Um, so assuming a big deviation from that would, would be a big jump.
00:31:02.250 --> 00:31:12.630
Yeah, and we already we're taxing corporate profits right now, and we are the only developed country actually that doesn't have a value added tax.
00:31:12.630 --> 00:31:12.690
Um.
00:31:13.890 --> 00:31:21.180
So in instead of, you know, inventing new AI taxes, should we just be fixing the tax system we already have?
00:31:21.436 --> 00:31:25.017
Anthropic, which is one of the big AI research labs.
00:31:25.076 --> 00:31:33.176
Um, did like a symposium recently of economists coming up, sharing, you know, ideas and, and for different policy scenarios.
00:31:33.176 --> 00:31:38.606
And some of them I, I think makes have some of the ideas I think are good and some of them i, I, I think are not so good.
00:31:38.997 --> 00:31:42.207
Um, but I think the.
00:31:43.811 --> 00:31:59.112
Idea that some people who buy into the pessimist story, um, have proposed a, so as a solution, uh, would be to introduce a, a, a consumption, taxes or, or, uh, instead of income taxes.
00:31:59.382 --> 00:32:07.662
Um, because consumption is, is more stable and, um, it's easier to tax than, than income, uh, generally.
00:32:07.961 --> 00:32:09.787
Um, and.
00:32:11.352 --> 00:32:20.261
While I might not buy the, the pessimist sort of story here, um, consumption taxes are more efficient than income taxes.
00:32:20.592 --> 00:32:35.142
And so if this is a reason to switch from income to consumption taxes or somewhat replace, uh, uh, the, uh, some of the income tax with consumption tax, um, I think that would be, that would be a good outcome.
00:32:35.517 --> 00:32:41.426
Um, and it would be, again, a sort of sound reform that I think would be good regardless of scenario.
00:32:41.846 --> 00:32:48.717
Um, but if you're particularly worried about the pessimist scenario, I, I can see how that is an appealing, uh, idea as well.
00:32:48.890 --> 00:32:50.029
So, so let's wrap it up.
00:32:50.029 --> 00:33:02.390
Let's say like in 60 seconds or less, if you were talking to a policymaker and they were concerned about AI had ideas about taxes, what would you say that they need to be thinking about going forward?
00:33:02.680 --> 00:33:12.609
I would say the biggest point is to not assume a radical scenario.
00:33:14.440 --> 00:33:27.880
As your, you know, as the building block for your, your policy solution, I think you should draw on the lessons of different waves of technological change.
00:33:28.299 --> 00:33:34.599
Uh, and both the, the fact that, you know, investment in new technology is important in.
00:33:35.410 --> 00:33:40.779
Long-term economic growth, uh, and that you shouldn't, you know, turn away from new technology.
00:33:41.019 --> 00:33:56.940
Um, but at the same time, you know, draw, uh, while avoiding, you know, penalizing new technology, just coming up with sort of positive ways to help people, uh, adjust to the, the changes in disruptions that it, it, it will bring areas of, of technological change, you know, you know, brought as well.
00:33:57.359 --> 00:33:57.839
Um.
00:33:58.170 --> 00:34:11.250
I guess the, the other sort of big picture point is that I wouldn't look to tax, uh, uh, for any social problem that you are worried about.
00:34:11.650 --> 00:34:14.110
we've been focused, as I said at the beginning on.
00:34:15.010 --> 00:34:23.530
Economics and, and you know, how AI will change the production of goods and services at a relatively broad level, maybe at a sector level.
00:34:23.889 --> 00:34:29.469
Um, but there are a lot of sort of social concerns that, that people might bring up with new technologies.
00:34:29.800 --> 00:34:34.719
Uh, like, you know, uh, people have a lot of concerns about the effect social media has had.
00:34:35.199 --> 00:34:35.650
Um.
00:34:36.219 --> 00:34:51.849
But I would be skeptical of designing, uh, uh, tax instruments for, uh, a lot of these, these concerns you should keep in mind that there's a, a wide array of like policy tools available for a whole variety of social ills.
00:34:51.909 --> 00:34:56.590
Uh, uh, and tax just is probably not the right.
00:34:56.994 --> 00:34:57.625
Tool to pick.
00:34:57.715 --> 00:35:10.795
And, and most of them, um, you know, I was thinking about with social media, if you're worried about, uh, if you're worried about social media, would you put a tax on like digital technology and like, the 1990s?
00:35:10.795 --> 00:35:20.005
And that ends up also hitting like Google Sheets because you can like, share you, you're, you're, you're sharing information with somebody else, you know?
00:35:20.065 --> 00:35:25.135
Um, and so I'd be skeptical of these sort of specific concerns that are valid.
00:35:25.675 --> 00:35:25.945
Um.
00:35:27.760 --> 00:35:32.485
Just getting sort of lumped in, uh uh, with a sort of broad policy on AI as well.
00:35:33.143 --> 00:35:37.634
There is at least one super niche AI tax that I would like, which is, I don't know.
00:35:37.753 --> 00:35:39.619
I'm sure you guys have seen this, is that there are.
00:35:40.498 --> 00:35:44.938
podcasts and as a podcast host, I find myself vaguely concerned about this.
00:35:44.938 --> 00:35:51.239
We created an outline for this podcast and I was like, oh man, I wonder what would that podcast would be like if we fed it to them?
00:35:51.659 --> 00:35:54.268
And I would like, maybe we just need to tax that.
00:35:54.268 --> 00:36:01.978
Like all AI podcasts get heavily taxed and we maybe just push those down for a little while so I can at least retire.
00:36:02.039 --> 00:36:02.128
Yes.
00:36:03.059 --> 00:36:03.809
But that's actually a real.
00:36:04.952 --> 00:36:06.572
but I, I think we can compete.
00:36:06.572 --> 00:36:09.992
I, I think people will like us better than an AI SLAP podcast.
00:36:11.603 --> 00:36:12.264
I'd like to think so.
00:36:12.940 --> 00:36:17.534
I believe, Kyle, you have many, many valuable skills that would, would make you, um.
00:36:17.889 --> 00:36:25.614
Valuable, uh, uh, uh, contributor in other forms if such, if such a, a technology, um, was spread.
00:36:26.739 --> 00:36:27.148
Thank you.
00:36:27.268 --> 00:36:28.168
I appreciate that.
00:36:28.318 --> 00:36:30.568
And Alex, thank you for being on this show today.
00:36:30.748 --> 00:36:31.889
Erica, thank You for being here.
00:36:31.889 --> 00:36:32.278
today.
00:36:32.639 --> 00:36:33.373
And before we.
00:36:33.724 --> 00:36:34.713
Completely sign off.
00:36:34.833 --> 00:36:38.313
As always, if you have burning questions on taxes, you can send them our way.
00:36:38.313 --> 00:36:40.083
You can drop a comment on YouTube.
00:36:40.293 --> 00:36:42.994
You can email us at podcast@taxfoundation.org.
00:36:43.054 --> 00:36:47.463
You can slide into our dms at Deduction Pod on Twitter or in many other ways.
00:36:47.463 --> 00:36:48.094
We're on Instagram.
00:36:48.094 --> 00:36:48.784
We're on TikTok.
00:36:48.934 --> 00:36:49.594
You'll find us.
00:36:49.858 --> 00:36:50.768
Thank you for listening.
00:00:00.299 --> 00:00:04.790
There is at least one super niche AI tax that I would like, which is, I don't know.
00:00:04.910 --> 00:00:06.775
I'm sure you guys have seen this, is that there are.
00:00:07.655 --> 00:00:12.095
podcasts and as a podcast host, I find myself vaguely concerned about this.
00:00:12.095 --> 00:00:18.394
We created an outline for this podcast and I was like, oh man, I wonder what would that podcast would be like if we fed it to them?
00:00:18.815 --> 00:00:21.425
And I would like, maybe we just need to tax that.
00:00:21.425 --> 00:00:29.135
Like all AI podcasts get heavily taxed and we maybe just push those down for a little while so I can at least retire.
00:00:29.195 --> 00:00:29.285
Yes.
00:00:30.214 --> 00:00:30.964
But that's actually a real.
00:00:32.107 --> 00:00:33.728
but I, I think we can compete.
00:00:33.728 --> 00:00:37.148
I, I think people will like us better than an AI SLAP podcast.
00:00:38.759 --> 00:00:39.420
I'd like to think so.
00:00:40.094 --> 00:00:44.689
I believe, Kyle, you have many, many valuable skills that would, would make you, um.
00:00:45.045 --> 00:00:52.770
Valuable, uh, uh, uh, contributor in other forms if such, if such a, a technology, um, was spread.
00:00:53.895 --> 00:00:54.304
Thank you.
00:00:54.424 --> 00:00:55.325
I appreciate that.
00:00:56.374 --> 00:00:59.643
hello and welcome to the Deduction of Tax Foundation Podcast.
00:00:59.643 --> 00:01:02.884
I'm your host, Kyle Houlahan, and we are back today with another episode.
00:01:02.884 --> 00:01:09.754
And I'm joined by my co-host, Erica York, and we're joined by Alex Muresianu, senior Policy Analyst at the Tax Foundation.
00:01:10.013 --> 00:01:10.763
all right, so here's the deal.
00:01:12.548 --> 00:01:14.257
AI is everywhere right now.
00:01:14.528 --> 00:01:17.378
It's like, I think you guys saw the Super Bowl, right?
00:01:17.617 --> 00:01:20.707
Like the a, it was all ai, which is kind of crazy.
00:01:20.707 --> 00:01:23.347
It's in the headlines every day, all the time, 24 7.
00:01:24.367 --> 00:01:26.617
but now it's part of the tax policy conversation.
00:01:26.947 --> 00:01:36.578
And so, you know, Alex, I'm wondering, know, do people need to know about this and, and where the debate stands on taxing ai.
00:01:36.968 --> 00:01:42.643
Yeah, so I think there are a lot of different angles that people, uh, um.
00:01:43.367 --> 00:01:46.158
Come to the AI policy conversation.
00:01:46.608 --> 00:01:46.908
Um.
00:01:47.522 --> 00:01:52.203
And there are a lot of different proposals out there that we can, we can get into.
00:01:52.203 --> 00:02:00.783
But generally the angle that at least I'm sort of most interested in is a sort of purely economic one.
00:02:01.263 --> 00:02:12.302
Uh, about how AI will, uh, or will not, uh, change the future of work and how it affects sort of basic, uh, uh.
00:02:12.992 --> 00:02:16.473
How we produce goods and services, sorts of questions.
00:02:16.473 --> 00:02:20.913
And also what does that mean for, for, uh, employment in the way, uh, work works.
00:02:21.362 --> 00:02:38.883
Um, and so I think the, the driving concern, uh, for a lot of these proposals relates to how, um, artificial intelligence will change, uh, or potentially replace, uh, human workers in various jobs.
00:02:39.242 --> 00:02:40.832
Uh, and, uh.
00:02:42.198 --> 00:02:48.407
Usually that's the focus of policy proposals, uh, to specifically target ai.
00:02:48.798 --> 00:02:57.798
The point that I, I think I'll, I'll probably make a, a couple times in various forms, is that like a lot of things that are AI might already be taxed.
00:02:57.798 --> 00:03:00.467
In other words, if you're thinking about profits from ai.
00:03:02.237 --> 00:03:04.367
Those are covered by the corporate income tax.
00:03:04.367 --> 00:03:22.758
you know, the, the particular concern here is, is sort of bespoke, uh, policy making that's sort of targeting, uh, whether it's, uh, investment in, you know, data centers or just, or revenue perhaps from AI products, no, I'll use the word again.
00:03:22.788 --> 00:03:26.657
Bespoke, bespoke policy specifically related to, um.
00:03:27.527 --> 00:03:28.758
Yes, it's a good word.
00:03:28.758 --> 00:03:29.268
It's a good word.
00:03:29.268 --> 00:03:39.828
Although I felt like saying it again, reminded me that there's a Veep episode where everybody uses the word, um, robust, um, in like three different meetings and then it becomes like a news story.
00:03:39.828 --> 00:03:42.437
So what are people in particular most worried about?
00:03:43.276 --> 00:03:49.967
A lot of science fiction stories, of course, uh, uh, talk about how machines will, will, like replace humans.
00:03:49.967 --> 00:04:10.516
And in the most extreme case, that has been covered in a few sort of, um, theoretical economics papers, uh, recently, and it is an interesting sort of theoretical exercise, at the very least, um, is that artificial intelligence will replace like all human labor as like an input in the economy.
00:04:10.997 --> 00:04:22.697
That's a, a sort of pretty extreme example, and I think people have predicted that technology would end, you know, uh, uh, human labor for a very long time like that.
00:04:22.697 --> 00:04:28.187
That's a, a, a prediction that has kind of failed, um, many, many times before.
00:04:29.026 --> 00:04:38.656
Um, to be fair, you only have to be right once, uh, for that one, but at a slightly less extreme level.
00:04:38.836 --> 00:04:58.276
The concern is more that that large portions of, uh, what people might call, like knowledge work or, or white collar employment, um, will be, will be replaced because all this technology, um, makes the work sort of much, much easier to, to, uh, a finish.
00:04:58.276 --> 00:05:04.733
And that this rapid productivity increase will replace, uh, a lot of human workers and the.
00:05:05.033 --> 00:05:11.963
The thing is that like, well, at the very high level, we don't know what the future holds.
00:05:12.204 --> 00:05:24.504
Um, but at a more specific level, we don't know how technology that makes a sector or, uh, a certain type of work, um, more productive.
00:05:25.043 --> 00:05:25.704
More productive, or.
00:05:26.338 --> 00:05:27.269
More efficient.
00:05:27.478 --> 00:05:33.059
We don't actually know what that means for employment, uh, in that particular sector.
00:05:33.389 --> 00:05:40.798
Um, there are story that there, there are sort of two, uh, terms to think about here, uh, that people sometimes use.
00:05:40.798 --> 00:05:45.629
One's bowel, mals cost, uh, cost disease, and the other one's Jevons Paradox.
00:05:45.959 --> 00:05:48.899
Um, those are just, just term good, good terms to know.
00:05:48.899 --> 00:05:53.699
But like the, the bowel mo cost disease story, is that like as.
00:05:54.298 --> 00:06:00.091
Some sectors get more productive, they kind of shrink, as a share of employment.
00:06:00.451 --> 00:06:06.752
fewer people work in them because new technology means you can get like a lot more, uh, uh, done with fewer people.
00:06:06.752 --> 00:06:09.182
So like a good example here might be like agriculture.
00:06:09.512 --> 00:06:17.072
Um, like we've gotten a lot more productive in producing, uh, uh, food and.
00:06:18.107 --> 00:06:25.966
The share of people who work in agriculture has declined dramatically, uh, over the past, you know, 200 years.
00:06:26.206 --> 00:06:29.447
That's like one end where we're getting better means fewer workers.
00:06:30.226 --> 00:06:34.487
The other end of the spectrum, I think would be like software or data analysis.
00:06:34.487 --> 00:06:54.863
I think where the technology for doing data analysis is so much better than it was in say, 1970 We have way more people who like do sort of data analysis or software, uh, now than did back in the 1970s as like a very direct example.
00:06:55.314 --> 00:07:04.084
Um, Tax Foundation has like software engineers, like I do not think that we had that in the eighties.
00:07:04.384 --> 00:07:06.694
Uh, for instance, like baseball teams.
00:07:07.144 --> 00:07:07.894
Baseball teams.
00:07:07.894 --> 00:07:22.954
Now, every baseball front office has like an analytics department where they have a bunch of, uh, uh, uh, you know, uh, stack geeks pouring over these sort of advan new advanced metrics that didn't even exist, um, you know, 30 or 40 years ago.
00:07:23.314 --> 00:07:29.853
Uh, and so what it boils down to is like how much more of something people want.
00:07:29.853 --> 00:07:34.564
Sometimes new technology opens up new markets and new applications.
00:07:34.983 --> 00:07:44.374
For certain types of work and that that means more employment in that sector and sometimes it replaces work because people don't actually want that much more of whatever this stuff is.
00:07:45.064 --> 00:07:57.634
So ultimately that's to say that one things are ambiguous, uh, and, and in terms of how productivity growth affects employment in a particular sector.
00:07:57.934 --> 00:08:02.014
And two, the big picture claim that.
00:08:02.853 --> 00:08:13.113
All of human labor would be replaced is quite an extreme one, and it would make sense to think about AI in more of the terms as other periods of technological change.
00:08:13.492 --> 00:08:23.781
so we're at this like early stage where we don't know how this new technology is going to affect the labor market, but there are a lot of fears that it could be extremely disruptive.
00:08:24.201 --> 00:08:29.122
We've got senators, we've got some economists, we've got even AI company CEOs.
00:08:30.307 --> 00:08:35.017
Proposing these really targeted taxes that could try to slow adoption.
00:08:35.917 --> 00:08:38.797
is that kind of getting ahead of where we need to be right now?
00:08:38.797 --> 00:08:49.236
What's the evidence for like labor market disruption today and does that justify like new tax provisions targeted AI right now?
00:08:49.236 --> 00:08:49.897
Or what?
00:08:49.927 --> 00:08:52.356
What should we be thinking of and what, what do we actually know?
00:08:52.626 --> 00:08:54.187
Like today in 2026?
00:08:54.513 --> 00:09:08.852
there are a whole host of, um, uh, labor market sort of indicators that one could, could comb through and, and probably put together a, a compelling narrative once you've gone through enough of them to say that something, something, you know, big is happening.
00:09:08.852 --> 00:09:09.003
But the.
00:09:10.352 --> 00:09:16.393
Big picture for most of the headline numbers is, not indicating a mass wave of, of automation.
00:09:16.783 --> 00:09:24.043
Um, the, you know, unemployment rate is lower than it was for the vast majority of the 2000 tens.
00:09:24.432 --> 00:09:31.692
Uh, labor force participation is at, you know, close to a, a, you know, 20 year high.
00:09:32.113 --> 00:09:32.653
Um.
00:09:33.087 --> 00:09:41.038
You, you know, so those are like the most basic high level, um, uh, employment indicators.
00:09:41.337 --> 00:09:59.158
Um, the one that people sort of most often, um, you know, rattle, the rattle the SARE about is the low rate of like entry level white collar work or, or like low, um, job hirings particularly, and, and software.
00:09:59.488 --> 00:10:01.168
Um, but the.
00:10:02.158 --> 00:10:11.998
Sort of confounding factor here is that there was this sort of massive hiring boom in those sectors in, in like 2021 into 2022.
00:10:12.717 --> 00:10:13.857
Um, and.
00:10:14.337 --> 00:10:38.217
The sort of low hiring rate is mostly I, I think a reflection of those, um, decisions that, that, you know, you had, um, you pulled a bunch of hiring to 20 21, 20 22, and now you're hiring last because you walked all those people in and it's, um, not confirm morale to just lay everybody, uh, uh, off if you, you were hired.
00:10:38.217 --> 00:10:40.648
So you just avoid hiring new people.
00:10:40.738 --> 00:10:42.613
Um, and so I.
00:10:43.182 --> 00:10:46.783
Like, like I sort of said, about like uncertainty.
00:10:46.783 --> 00:11:05.273
Like this is not like proof that, that some massive wave of automation could come in the future, but to say like, it's here and we need to like design some crazy new policy because, uh, uh, you know, the, the writing is on the wall for all this stuff.
00:11:05.423 --> 00:11:09.442
I think that is a, a pretty extreme reaction to existing evidence.
00:11:10.351 --> 00:11:11.422
Yeah, and, and obviously.
00:11:12.581 --> 00:11:14.591
We don't wanna completely invalidate the fear around this.
00:11:14.591 --> 00:11:19.211
This is obviously, is a very new thing and it's a very real massive change in all of our lives.
00:11:19.572 --> 00:11:21.522
Um, and so that, that makes sense.
00:11:21.522 --> 00:11:30.761
I understand some of the fear, but, but what do you see as like the, what are the main tax proposals out there right now and, and who or what is driving those?
00:11:31.278 --> 00:11:43.668
there's a, a category that I think of as explicitly targeted at the employment issue and or the employment issue, I think is the, the driving focus of a lot of them.
00:11:43.878 --> 00:11:48.707
The intent, uh, of the policies might vary.
00:11:48.918 --> 00:11:51.768
So, uh, the two, uh, um.
00:11:52.472 --> 00:12:09.302
Proposals that I highlighted in the, uh, piece, uh, for tax foundation that I wrote are on the one hand, uh, Senator Bernie Sanders has proposal to, um, both tax like AI profits and also, uh, to.
00:12:09.947 --> 00:12:14.118
Uh, get rid of deductions for, uh, investment in equipment.
00:12:14.118 --> 00:12:18.587
So that's covering physical automation as well as sort of digital automation.
00:12:18.798 --> 00:12:24.197
And, you know, the Sanders perspective is explicitly like, this is a problem.
00:12:24.197 --> 00:12:29.597
This is going to, you know, replace workers and like, we need to stop.
00:12:29.972 --> 00:12:39.452
The, uh, uh, replace, you know, you know, we, we just need to, to penalize these investments to, to prevent them from happening and prevent the employment disruption from happening.
00:12:40.592 --> 00:12:46.442
Then the sort of other perspective is not as, uh, sort of ideologically like motivated.
00:12:46.682 --> 00:12:47.852
It's more oh.
00:12:48.677 --> 00:12:55.817
It's good that we have this new technology, but it will come with downsides or, or, or some people will be disrupted.
00:12:55.822 --> 00:13:04.067
Or, or, or, or, you know, you, you might have some sector level unemployment and the proposed answer is to sort of fund programs, uh.
00:13:04.977 --> 00:13:09.837
Four people who are, are disrupted through some sort of tax targeted at the industry.
00:13:10.227 --> 00:13:14.337
Um, this is the, uh, Senator Mark Kelly proposal.
00:13:14.668 --> 00:13:21.227
Um, and the details of, of that proposal, at least when I, I wrote the piece.
00:13:21.227 --> 00:13:24.587
Maybe there's something that has come out that's more fleshed out that I, that I missed.
00:13:24.888 --> 00:13:31.368
Um, but his proposal, basically some sort of tax on, on industry revenue.
00:13:31.727 --> 00:13:32.148
Um.
00:13:33.038 --> 00:13:39.217
And so I think Kelly's thought process is a little bit different and a little bit more reasonable.
00:13:39.398 --> 00:13:50.258
However, I wouldn't be confident that the tax designed to just ameliorate some of the issues wouldn't itself disincentivize a new investment.
00:13:50.768 --> 00:13:59.557
Um, you know, again, the details aren't out there, but just because your intentions are a little bit different doesn't mean that you wouldn't have the same effect of, of slowing down a technology adoption.
00:14:03.075 --> 00:14:05.220
Sometimes when we, we hear.
00:14:06.044 --> 00:14:08.144
Proponents of of AI taxes talk.
00:14:08.144 --> 00:14:16.934
And this has parallels to the way, like other conversations go on, like the, the trade policy space and what do we do for the manufacturing sector.
00:14:16.934 --> 00:14:22.004
And there's kind of this like long thread of thinking that the tax code.
00:14:22.075 --> 00:14:24.235
Favors capital over labor.
00:14:24.235 --> 00:14:32.815
And so we need to, you know, change policy to write that, to make sure that we're not incentivizing investment here over investing in workers.
00:14:32.995 --> 00:14:36.144
Can you speak to that and, and what's actually the case there?
00:14:37.937 --> 00:14:52.278
Yeah, so you know, the biggest picture, uh, illustration of this is, uh, how companies deduct certain kinds of costs, um, and.
00:14:53.743 --> 00:15:04.062
And you know, just about, just about every case, uh, companies deduct the cost of like worker salaries immediately, our wages and salaries, uh, immediately.
00:15:04.393 --> 00:15:08.052
Uh, while often they have to spread the.
00:15:08.878 --> 00:15:14.008
Costs of, of capital investments, uh, out over time.
00:15:14.398 --> 00:15:27.087
Um, now there are some provisions that are, are currently in effect in, in, in us, uh, tax law, na, many of which are relatively new, that allow for the, uh, full deductions for investment.
00:15:27.477 --> 00:15:29.763
Um, the, uh, uh, introduction of.
00:15:30.493 --> 00:15:36.373
A hundred percent bonus depreciation for equipment, um, expensing for research and development.
00:15:36.373 --> 00:15:40.993
Immediate deductions for r and d investment that has been brought back into law.
00:15:41.322 --> 00:15:41.802
Um.
00:15:42.207 --> 00:15:49.707
But those just put those investments on a sort of level playing field with, um, you know, worker compensation.
00:15:50.067 --> 00:15:56.008
Um, and not all investment is eligible, uh, for those provisions.
00:15:56.337 --> 00:16:05.998
Uh, and so if anything on the whole, um, that tax system kind of dis favors a, a, a capital investment.
00:16:06.447 --> 00:16:10.528
Um, but I think the, the point about how.
00:16:12.822 --> 00:16:39.102
This links to other debates about like displacement, uh, of which, which is usually wrapped up in, in at least historically related to manufacturing jobs or, um, uh, you know, I know our, our, uh, our producer talked about, uh, our produ, our great, great producer Dan, uh, talked, uh, talked to us in our, our prep call about, um, history of the coal industry in West Virginia.
00:16:39.177 --> 00:16:39.437
You know.
00:16:40.092 --> 00:16:55.903
There are, uh, uh, precedent for or localized disruptions, uh, uh, to, uh, um, uh, employment and, and, you know, general economic prospects related to economic change and technological changes.
00:16:56.322 --> 00:16:56.773
Um.
00:16:57.373 --> 00:16:59.682
But those are, are sort of unprecedented.
00:16:59.682 --> 00:17:07.603
And the, the answer wouldn't be just to try and like slow down change it would be to figure out how best to, to ameliorate it.
00:17:07.633 --> 00:17:11.413
We have, you know, some experience with that and, uh, to, to think through.
00:17:11.413 --> 00:17:12.702
There are lessons to be learned.
00:17:13.031 --> 00:17:20.652
So rather than slow the adoption of new technology, find ways to directly help the workers who may be impacted.
00:17:20.682 --> 00:17:23.922
Rather than saying, no, we're gonna freeze technology in place.
00:17:23.922 --> 00:17:29.112
It, it reminds me of a line from a paper, um, that was looking at, you know, the.
00:17:29.846 --> 00:17:35.576
The decline in manufacturing jobs and looking at whether technology or trade was a bigger contributor.
00:17:35.576 --> 00:17:47.906
And this particular paper concluded it was technology and it made the argument and has the line, like if you wanna go back to higher levels of manufacturing employment, pretty much the only policy solution is to smash the machines.
00:17:48.146 --> 00:17:48.807
Um, so I don't think.
00:17:50.007 --> 00:17:55.227
We should be at a point where we're saying, yes, smash the machines or tax the machines to slow adoption.
00:17:55.586 --> 00:18:05.787
It should be more of an openness to innovation and to growth, also acknowledging that there are other policy tools maybe that, that we could look at.
00:18:06.067 --> 00:18:12.876
a lot of this feels like it's, it's largely based on, there's a lot of fear around how, what will happen to labor.
00:18:13.057 --> 00:18:16.057
Most of this, like most of this conversation that we're having is fear around labor.
00:18:16.057 --> 00:18:26.916
So how could reforming maybe, you know, unemployment insurance or raising some of those caps or worker training deductions or something, you know, how could that maybe, you know, help,
00:18:27.366 --> 00:18:44.136
I would say the sort of best way in, in principle to address, um, the disruption that introducing a new technology, uh, would have is to make it easier for people to.
00:18:45.606 --> 00:19:02.436
Transition to doing something different, uh, or in a different, and, and put in a different way, sort of lean into different skills, uh, or, or, or applications for their, their, uh, you know, alternative applications for their existing sets of, of skills.
00:19:02.856 --> 00:19:05.826
we're the tax foundation I'm focused, focused on, on tax angles.
00:19:05.826 --> 00:19:15.487
And I don't mean to say that this is a, a comprehensive policy menu, but I think these are the, the relevant policies in the tax code to think about.
00:19:15.846 --> 00:19:24.007
Um, the first is unemployment insurance, um, unemployment insurance in the United States.
00:19:24.037 --> 00:19:24.892
Uh, the tax rates.
00:19:25.717 --> 00:19:33.186
Are, uh, uh, you know, vary by employer and relate to their experience with the unemployment system.
00:19:33.186 --> 00:19:43.416
And the sort of justification here is that, and, and, and you know, your rates go up if you have more unexperience with the, the unemployment system, which is means that like, you know, you've.
00:19:43.957 --> 00:19:50.436
Fired people, uh, uh, before, and, you know, the, the thinking there feels very, very intuitive.
00:19:50.436 --> 00:19:57.517
It's like we don't want people to, um, be, be laid off and so will tax companies more that are, are laying people off.
00:19:57.517 --> 00:19:59.136
But I think the, the.
00:19:59.811 --> 00:20:17.872
Real effect here is that people are, are, are firms are like more worried about, uh, risky hires than, uh, uh, they would otherwise be because they know that they're, if, if somebody, you know, oh, this, this, this candidate has a lot of potential, but also.
00:20:18.487 --> 00:20:20.406
Might really not be a fit for this role.
00:20:20.676 --> 00:20:25.057
Um, they know that they may well face a tax penalty in the future if the, that just isn't right.
00:20:25.446 --> 00:20:30.846
Um, and so I think that's a, a particular concern for people who are looking to change careers, uh, or.
00:20:31.176 --> 00:20:39.457
Looking to move from, from something, um, uh, that, you know, might only be tangentially related to what they previously worked in.
00:20:39.457 --> 00:20:46.146
And they might, you know, think that, that this, this, you know, I, I really couldn't apply my skills, uh, in this other, other career path.
00:20:46.567 --> 00:20:54.906
Um, and so, you know, if you're worried that new technology that is coming down the pipe will, um.
00:20:55.446 --> 00:21:03.247
Make certain forms of, uh, certain skills or certain particular occupations sort of, uh, largely automated away then.
00:21:04.402 --> 00:21:16.672
This policy, the, the basing, the unemployment, uh, uh, tax rates on experience rating could sort of dissuade companies from taking a cha taking chances on people who are looking to, to shift careers.
00:21:17.092 --> 00:21:28.432
Um, the other, uh, uh, angle would be how, uh, of a tax code treats, uh, company investment in, uh, worker training, and particularly training for new roles.
00:21:28.761 --> 00:21:31.731
Um, there's a cap on the deduction for.
00:21:32.261 --> 00:21:34.481
Investment in training for existing roles.
00:21:34.541 --> 00:21:37.902
It's a non-deductible, uh, for training in new roles.
00:21:38.412 --> 00:21:41.896
part of the reason I guess, uh, to, to like justify why that policy would.
00:21:42.876 --> 00:21:43.507
Where it is.
00:21:43.507 --> 00:21:51.876
Like there are you, you do, you do sometimes worry in, in the tax code of having sort of non uh, uh, taxed, uh, worker compensation.
00:21:51.876 --> 00:21:57.836
So I think that that's the sort of original like thought process of, not having ways to do like non wage compensation.
00:21:57.836 --> 00:21:59.007
That's, that's UNT taxable.
00:21:59.277 --> 00:22:00.297
Um, but.
00:22:01.811 --> 00:22:11.172
The effect of not having this, this deduction available is that there isn't, uh, much of an incentive for companies, uh, or the very least there.
00:22:11.561 --> 00:22:28.676
There's a disincentive, I guess, for companies to retrain workers who may be, are no longer, um, you know, productive or necessary in their original role, but may still be like, you know, you think you're, they're like solid workers who, who are like diligent and will.
00:22:29.336 --> 00:22:33.057
Um, might be fits for alternate employment in the company.
00:22:33.116 --> 00:22:41.967
And so, um, those are two sort of option, the employment insurance and tax treatment of worker training or retraining.
00:22:42.416 --> 00:22:43.916
Um, that.
00:22:44.906 --> 00:22:49.767
Those are like t like, like policy changes in the regular tax toolkit.
00:22:50.217 --> 00:23:08.696
Um, that would be sort of sound structural improvements sort of regardless of what, uh, uh, happens with, uh, uh, artificial intelligence, which again, there's a very wide spectrum of outcomes and like if this is a whole uh, uh, you know, nothing burger, then it would still be good that we made those changes.
00:23:09.476 --> 00:23:12.615
Real quick, just to kind of follow up to make sure I understood this correctly.
00:23:12.645 --> 00:23:13.830
It's kind of like based.
00:23:14.925 --> 00:23:28.336
The way you're setting this up is like, okay, a company X could invest in said software and deduct that, but they can't deduct the same thing for investing in people at that work for them.
00:23:28.383 --> 00:23:30.055
So they can take a deduction.
00:23:30.970 --> 00:23:37.210
It's up to, I think,$5,250 for, uh, like worker training.
00:23:37.660 --> 00:23:45.309
Um, but uh, that is supposed to be sort of related to their, uh, uh, current role.
00:23:45.609 --> 00:23:49.750
Uh, and like full sort of retraining, uh, is non-deductible.
00:23:50.076 --> 00:23:55.836
there is a justification here where you don't want to have, um, uh.
00:23:56.527 --> 00:24:06.876
Things that are effectively a form of employee compensate or might kind of be a form of employee compensation that's deductible at the firm level and also non-taxable at the individual level.
00:24:06.876 --> 00:24:10.386
That's like a common concern at the margins of tax policy.
00:24:10.747 --> 00:24:15.126
Um, but in this particular case, you know, you could have some sort of cap.
00:24:15.846 --> 00:24:26.571
Uh, uh, it might not be unlimited, but, uh, having, you know, more than, than zero, uh, uh, you know, uh, uh, limit, um, could, could be effective.
00:24:27.057 --> 00:24:31.438
this reminds me of a book that Glenn Hubbard wrote in like 2022.
00:24:31.498 --> 00:24:37.827
So before AI was really being discussed, but the book is called The Wall on the Bridge, and it's like policy ideas.
00:24:37.877 --> 00:24:41.958
To help build bridges for people to respond to things that are disruptive.
00:24:42.377 --> 00:24:49.907
So a lot of the ideas he talks about in there, they expand beyond tax policy, but like unemployment insurance reform, rethinking how social insurance works.
00:24:49.907 --> 00:25:00.917
So if it is a technological disruption that's causing hardship for people, don't block the disruption, like bridge people to deal with that in a, in a productive way.
00:25:00.917 --> 00:25:03.587
So I think like those ideas you, you mentioned are.
00:25:04.343 --> 00:25:06.532
A perfect fit with that kind of concept,
00:25:06.689 --> 00:25:13.798
was just realizing it's like I, I think our bookshelves are, are radically different, what you have on your shelf and what I have on my shelf.
00:25:13.798 --> 00:25:16.769
I was like, I don't think I have any economic books like.
00:25:17.527 --> 00:25:24.457
I was like my, one of the, my favorite books I've read, and I talk about that book all the time because it's so applicable, whether, whether people are upset about like.
00:25:24.906 --> 00:25:28.537
trade policy disrupting people or new technology disrupting people.
00:25:28.537 --> 00:25:34.326
Now AI disrupting the the workforce, there's always going to be some kind of disruption happening.
00:25:34.717 --> 00:25:36.487
The answer isn't stop disruption.
00:25:36.487 --> 00:25:40.416
It's let's be more agile in response to disruption.
00:25:40.645 --> 00:25:50.635
So bigger picture when we're thinking about like the federal budget situation and the sustainability of debt, we all know that it's unsustainable.
00:25:50.816 --> 00:25:54.865
Deficits are already really large, and one of the things people talk about with AI is what.
00:25:55.675 --> 00:25:58.256
What will the impact be on the federal budget?
00:25:59.066 --> 00:26:01.465
it going to make the deficit situation worse?
00:26:01.496 --> 00:26:13.346
Um, you know, if these labor doomsday scenarios come, come into play, do income tax revenues go in the tank or will it help growth and help reduce the budget deficit?
00:26:13.615 --> 00:26:16.435
Where, where do you think the, the truth of the matter is?
00:26:16.884 --> 00:26:17.275
Yeah.
00:26:17.275 --> 00:26:25.224
So beyond the like first order answer, which is, I don't know, uh, which I, I think is, is like.
00:26:26.529 --> 00:26:28.240
The most a hundred percent accurate answer.
00:26:28.240 --> 00:26:28.569
It's true.
00:26:28.630 --> 00:26:30.099
I like, I don't know the future.
00:26:30.430 --> 00:26:42.700
Um, and I think this area, like in in particular, um, like people, people don't know, um, exactly what the effect of this, this new technology will be on all kinds of things.
00:26:43.119 --> 00:26:49.509
Um, but I think the, like putting a whole lot of belief in the strong versions.
00:26:49.855 --> 00:26:55.434
Of either prediction, I, I, I think is, is, uh, um, a mistake.
00:26:55.855 --> 00:27:18.025
Um, there is this sort of Optimus story where the, you know, great that, that AI will unleash like great productivity growth and that will, uh, um, increase, uh, tax revenue and that will, you know, solve a lot of structural, uh, the long-term sort of structural problems, uh, in the us uh, fiscal picture.
00:27:19.255 --> 00:27:42.025
And then the sort of negative, uh, uh, um, sort of pessimistic story that, you know, the, the labor disruption means that the individual income tax, which is the sort of workhorse of the, uh, us uh, uh, federal revenue, um, you know, if revenues from that and then, you know, payroll taxes too, you know, go through the floor.
00:27:42.025 --> 00:27:45.025
Now we've gotten even worse, uh, a fiscal situation.
00:27:46.075 --> 00:27:47.184
I, uh.
00:27:48.444 --> 00:28:00.775
If anything, you know, generally you'd expect if AI like improves productivity growth, you'd expect that to directionally, uh, uh, positively contribute to the US fiscal situation.
00:28:01.045 --> 00:28:05.845
Just generally, um, we expect productivity growth to mean like more, more revenue.
00:28:06.265 --> 00:28:15.680
Um, on that, there are a bunch of different like little interactions where, based on how certain programs are calculated, but on net this is a, a deficit generally like.
00:28:16.654 --> 00:28:19.055
Productivity growth should mean like reduced deficits.
00:28:19.595 --> 00:28:33.569
Um, but the promise of dramatically improved productivity growth as a solution to uh, uh, long-term fiscal problems is just like.
00:28:34.444 --> 00:28:38.674
A classic like siren song of, classic wishful thinking.
00:28:39.035 --> 00:28:51.035
Um, and I certainly wouldn't, you know, count on that or, or have a lot of saying, you know, oh, well, well AI's on the way, we'll have this, this incredible productivity growth and you don't really have to worry about, but have a set.
00:28:51.394 --> 00:28:53.015
Um, I would be very wary of that.
00:28:53.535 --> 00:28:58.244
particularly because baseline budget protections already include productivity growth.
00:28:58.244 --> 00:29:05.595
Like there's some productivity growth already in forecasts and that's gotta be coming from somewhere.
00:29:05.684 --> 00:29:16.664
And like, maybe AI is like good for, for productivity growth, but that's just all the stuff that's in the budget, you know, forecast and economic outlooks already.
00:29:17.119 --> 00:29:28.700
the other thing with the pessimistic story that, that, you know, all of a sudden labor is going to be replaced, or the very least, you know, depleted, uh, substantially.
00:29:30.095 --> 00:29:32.134
Thanks to AI technology.
00:29:32.404 --> 00:29:45.545
Now, this is, uh, uh, where it makes sense to think about things in a, in a historical context and not sort of, assume AI is dramatically different from historical examples of technological change or, or disruption.
00:29:45.545 --> 00:29:58.384
And when you look at the, the labor share of income, you know, measured properly, uh, over the course of, you know, the 20th century into the 21st, so far, labor share of income is quite stable.
00:29:58.789 --> 00:30:12.365
Um, and so assuming that things will be dramatically different and that will sort of warp, uh, the, the structure of, uh, uh, tax revenue, uh, I, I, I don't think that's a, a sound bet.
00:30:12.803 --> 00:30:19.647
That measured properly point that you just mentioned is like the key because like today there are some charts going around with.
00:30:20.548 --> 00:30:23.488
share of gross income rather than labor share of net income.
00:30:23.728 --> 00:30:27.208
And if you look at the gross income picture, it does look like labor share is declining.
00:30:27.387 --> 00:30:32.607
But the real big story there is like depreciation and, and taxes have have gone up.
00:30:32.938 --> 00:30:37.708
Um, and of course, depreciation and taxes aren't income that actually goes into anyone's pockets.
00:30:37.857 --> 00:30:42.988
So you wanna net out those things that income to anyone, and then look at how it's split between capital and labor.
00:30:43.228 --> 00:30:44.968
And there you do see this really.
00:30:45.718 --> 00:30:47.248
Really stable share.
00:30:47.248 --> 00:30:51.208
Um, you know, some, some slight changes since the eighties, but nothing dramatic.
00:30:51.208 --> 00:30:55.768
It's really close to that average, and it has been since, like you said, the, the 1930s.
00:30:56.097 --> 00:31:00.373
Um, so assuming a big deviation from that would, would be a big jump.
00:31:02.250 --> 00:31:12.630
Yeah, and we already we're taxing corporate profits right now, and we are the only developed country actually that doesn't have a value added tax.
00:31:12.630 --> 00:31:12.690
Um.
00:31:13.890 --> 00:31:21.180
So in instead of, you know, inventing new AI taxes, should we just be fixing the tax system we already have?
00:31:21.436 --> 00:31:25.017
Anthropic, which is one of the big AI research labs.
00:31:25.076 --> 00:31:33.176
Um, did like a symposium recently of economists coming up, sharing, you know, ideas and, and for different policy scenarios.
00:31:33.176 --> 00:31:38.606
And some of them I, I think makes have some of the ideas I think are good and some of them i, I, I think are not so good.
00:31:38.997 --> 00:31:42.207
Um, but I think the.
00:31:43.811 --> 00:31:59.112
Idea that some people who buy into the pessimist story, um, have proposed a, so as a solution, uh, would be to introduce a, a, a consumption, taxes or, or, uh, instead of income taxes.
00:31:59.382 --> 00:32:07.662
Um, because consumption is, is more stable and, um, it's easier to tax than, than income, uh, generally.
00:32:07.961 --> 00:32:09.787
Um, and.
00:32:11.352 --> 00:32:20.261
While I might not buy the, the pessimist sort of story here, um, consumption taxes are more efficient than income taxes.
00:32:20.592 --> 00:32:35.142
And so if this is a reason to switch from income to consumption taxes or somewhat replace, uh, uh, the, uh, some of the income tax with consumption tax, um, I think that would be, that would be a good outcome.
00:32:35.517 --> 00:32:41.426
Um, and it would be, again, a sort of sound reform that I think would be good regardless of scenario.
00:32:41.846 --> 00:32:48.717
Um, but if you're particularly worried about the pessimist scenario, I, I can see how that is an appealing, uh, idea as well.
00:32:48.890 --> 00:32:50.029
So, so let's wrap it up.
00:32:50.029 --> 00:33:02.390
Let's say like in 60 seconds or less, if you were talking to a policymaker and they were concerned about AI had ideas about taxes, what would you say that they need to be thinking about going forward?
00:33:02.680 --> 00:33:12.609
I would say the biggest point is to not assume a radical scenario.
00:33:14.440 --> 00:33:27.880
As your, you know, as the building block for your, your policy solution, I think you should draw on the lessons of different waves of technological change.
00:33:28.299 --> 00:33:34.599
Uh, and both the, the fact that, you know, investment in new technology is important in.
00:33:35.410 --> 00:33:40.779
Long-term economic growth, uh, and that you shouldn't, you know, turn away from new technology.
00:33:41.019 --> 00:33:56.940
Um, but at the same time, you know, draw, uh, while avoiding, you know, penalizing new technology, just coming up with sort of positive ways to help people, uh, adjust to the, the changes in disruptions that it, it, it will bring areas of, of technological change, you know, you know, brought as well.
00:33:57.359 --> 00:33:57.839
Um.
00:33:58.170 --> 00:34:11.250
I guess the, the other sort of big picture point is that I wouldn't look to tax, uh, uh, for any social problem that you are worried about.
00:34:11.650 --> 00:34:14.110
we've been focused, as I said at the beginning on.
00:34:15.010 --> 00:34:23.530
Economics and, and you know, how AI will change the production of goods and services at a relatively broad level, maybe at a sector level.
00:34:23.889 --> 00:34:29.469
Um, but there are a lot of sort of social concerns that, that people might bring up with new technologies.
00:34:29.800 --> 00:34:34.719
Uh, like, you know, uh, people have a lot of concerns about the effect social media has had.
00:34:35.199 --> 00:34:35.650
Um.
00:34:36.219 --> 00:34:51.849
But I would be skeptical of designing, uh, uh, tax instruments for, uh, a lot of these, these concerns you should keep in mind that there's a, a wide array of like policy tools available for a whole variety of social ills.
00:34:51.909 --> 00:34:56.590
Uh, uh, and tax just is probably not the right.
00:34:56.994 --> 00:34:57.625
Tool to pick.
00:34:57.715 --> 00:35:10.795
And, and most of them, um, you know, I was thinking about with social media, if you're worried about, uh, if you're worried about social media, would you put a tax on like digital technology and like, the 1990s?
00:35:10.795 --> 00:35:20.005
And that ends up also hitting like Google Sheets because you can like, share you, you're, you're, you're sharing information with somebody else, you know?
00:35:20.065 --> 00:35:25.135
Um, and so I'd be skeptical of these sort of specific concerns that are valid.
00:35:25.675 --> 00:35:25.945
Um.
00:35:27.760 --> 00:35:32.485
Just getting sort of lumped in, uh uh, with a sort of broad policy on AI as well.
00:35:33.143 --> 00:35:37.634
There is at least one super niche AI tax that I would like, which is, I don't know.
00:35:37.753 --> 00:35:39.619
I'm sure you guys have seen this, is that there are.
00:35:40.498 --> 00:35:44.938
podcasts and as a podcast host, I find myself vaguely concerned about this.
00:35:44.938 --> 00:35:51.239
We created an outline for this podcast and I was like, oh man, I wonder what would that podcast would be like if we fed it to them?
00:35:51.659 --> 00:35:54.268
And I would like, maybe we just need to tax that.
00:35:54.268 --> 00:36:01.978
Like all AI podcasts get heavily taxed and we maybe just push those down for a little while so I can at least retire.
00:36:02.039 --> 00:36:02.128
Yes.
00:36:03.059 --> 00:36:03.809
But that's actually a real.
00:36:04.952 --> 00:36:06.572
but I, I think we can compete.
00:36:06.572 --> 00:36:09.992
I, I think people will like us better than an AI SLAP podcast.
00:36:11.603 --> 00:36:12.264
I'd like to think so.
00:36:12.940 --> 00:36:17.534
I believe, Kyle, you have many, many valuable skills that would, would make you, um.
00:36:17.889 --> 00:36:25.614
Valuable, uh, uh, uh, contributor in other forms if such, if such a, a technology, um, was spread.
00:36:26.739 --> 00:36:27.148
Thank you.
00:36:27.268 --> 00:36:28.168
I appreciate that.
00:36:28.318 --> 00:36:30.568
And Alex, thank you for being on this show today.
00:36:30.748 --> 00:36:31.889
Erica, thank You for being here.
00:36:31.889 --> 00:36:32.278
today.
00:36:32.639 --> 00:36:33.373
And before we.
00:36:33.724 --> 00:36:34.713
Completely sign off.
00:36:34.833 --> 00:36:38.313
As always, if you have burning questions on taxes, you can send them our way.
00:36:38.313 --> 00:36:40.083
You can drop a comment on YouTube.
00:36:40.293 --> 00:36:42.994
You can email us at podcast@taxfoundation.org.
00:36:43.054 --> 00:36:47.463
You can slide into our dms at Deduction Pod on Twitter or in many other ways.
00:36:47.463 --> 00:36:48.094
We're on Instagram.
00:36:48.094 --> 00:36:48.784
We're on TikTok.
00:36:48.934 --> 00:36:49.594
You'll find us.
00:36:49.858 --> 00:36:50.768
Thank you for listening.
