The only concerning thing is that someone somewhere in government will be stupid enough to push for this.
Maybe Excel could add a feature where taxes are collected every time you add a trend line to a chart. I hope they charge more for polynomial fit - it should be based on number of model parameters right?
Drivel. Where are the AI agents doing value add work out there owned by no one? How much do the robots at car factories pay in tax today? The arguments make no sense in the foreseeable future. Maybe some future word post AGI but even then, until we have robots earning and spending money with no human owners in the mix this is not a serious line of thought.
A click bait article cashing in on AI hype. We don't tax machines, we tax the owners for profit generated. And if we have autonomous AIs not owned by anyone, tax wouldn't be the primary problem.
I’m firmly in the “AI will never be a sentient being and shouldn’t be expected to have rights” camp.
Accordingly, AI won’t be sending kids to public schools, collecting Social Security, or benefiting from Medicare. Most importantly, AI won’t have seats in Congress.
If AI does productive work, individuals and organizations who leverage it successfully will make more money, which is taxed progressively (in theory anyway). At higher brackets, income tax can bring in more revenue than payroll taxes.
Or tax the use of natural and shared resources [1]. The idea that somewhere out there value is being created and the govt isn't getting a cut of it is wicked.
Too many people thing the only way to fund public programs is taxing labor or income, when that is the most unethical and least efficient method of taxation there is, and leads to all kinds of abuse by government.
Commenters seem to be dismissing this concept outright while ignoring the fact that the owners of these robots, through corporate schemes, offshoring, and legal tax manipulation, end up paying 0-10% tax compared to 20-45% taxes paid by human workers currently.
It is not a zero-sum calculation and AI/robots represent a huge transfer of social wealth (taxes collected for use by society) to the rich.
This "income tax for robots" idea is one of Bill Gates' ideas that has been kicking around for over a decade, and will surely fall victim to the same type of loopholes the rich presently exploit.
Yanis Varoufakis has been criticizing this Gates approach and proposing an alternative concept which he calls Universal Basic Dividend. Under this scheme, the use of automation to generate capital would be socialized through the generation of shares, say 10% ownership of the firms which utilize automation to replace labor, which are owned by a public trust which distributes dividends from the firm's profit to every human in the country equally.
I suspect the in practice effect of something like this would result in people getting a 10 dollar check once a month or something. Which would be something, but wouldn't actually mitigate any of the assumed harms of automation. Taxes at least could build infrastructure which might facilitate the growth of new forms of employment. But everyone getting a pittance once a month is unlikely have very much positive benefit to the impacted people.
10% of 2.09 trillion corporate profits represents $627 per person annually.
The 10% is an arbitrary starting point and should be adjusted based upon what is the appropriate representation of the displacement of labor by automation, which will surely increase along with corporate profits increasing with less salaries being paid.
The idea isn't to fully socialize and subvert capitalism, but to keep capitalism alive long enough to ethically support humanity as long as it is necessary.
I'm actually surprised I was so close on my estimate. The point still stands. You could triple that return and it would still make almost no substantial mitigating impact.
$54/mo is 5x off from your estimate, and though it seems ignorable to you, is remarkably close to the $57/mo international poverty line, which 9.2% of the world lives under. That is twice as many people as there are Americans.
Why is it important to start this ~200 years into the industrial revolution? What is actually different now? If anything we're (well) into diminishing returns.
That doesn't mean anything to me but maybe it's persuasive for somebody.
We already have mechanical harvesters and computers and looms and whatnot. I don't see a step change at llms.
One thing that's different (if you buy into the hype) is that the jobs that are threatened this time are held by educated people. When it was truck drivers (now a lucrative job in no danger of replacement) being threatened, we told them they could learn to code. Now that people writing code and doing powerpoints (if you believe the hype) are threatened, we need UBI
Even then I don't see how this is currently any more than a productivity enhancer like an IDE, refactoring tools or a higher level language. Game engines like Unity are a huge productivity gain compared to hand coding a new engine for every game, but nobody is bringing in a special game engine tax.
I do think capital returns should be taxed similarly to income, but it's not simple to accomplish. One thing to bear in mind on taxing return on capital is that it's already taxed several times, once as corporate income tax and then again on dividends or share price gains. So the real tax on capital returns is the combination of these, and sometimes other taxes as well. There are some common loopholes that should really be closed though, such as exemptions for inherited capital.
Yanis was proposing this a decade ago when it was the blue collar workers you mention being displaced. The problem is merely increasingly visible as more classes of people become affected.
You guys are both right. It just needs to be understood that UBI itself is a weak band-aid for otherwise intractable problems or contradictions that will continue to unfold as we progress in the experiment, but were always inherent in the design. That's why it at once seems immediately necessary and also too late.
There is no "right time" for it, or it has always been the right time for it, depending on where you hang your hat around more fundamental issues.
It's absurd that income generated from productive work is even taxed at all while there is such a wealth of rents to tax - land, intellectual property, etc. Especially land.
Tyler Cowen would be one of the least likely economists to ever acknowledge this though.
There is a reasonable argument that companies who employ a lot of people per unit of profit/revenue should pay a lower rate of tax on those profits.
If you create a fleet of robo-semi trucks and take over the American logistics business by cutting 80% of the transportation workforce, you are creating significant social harm in the pursuit of profit. Increased taxation could either act as a disincentive to the labour-pocalypse or provide revenue to help society smooth the career transition for those workers.
We (U.S) don't tax human labor. You can invest countless hours of your own effort into creating and/or improving property (including intangible property) and never pay a cent in tax.
Taxing consumption discourages consumption which thus acts as a drag on the overall economy. Further, most people view it as "unfair" because a fixed consumption tax (sales, GST, VAT) results in lower income people paying a higher percentage of their income as tax.
Progressive income taxation 1) does not discourage consumption as much as a higher consumption tax for those with lower income (which is most people), 2) encourages investment for companies and individuals with higher incomes.
Yeah but people will still have to generate excess produce somehow in order to have something to trade for those (now cheaper) goods, and with automation taking both skilled and unskilled jobs out of the market, I keep wondering what kind of jobs will be left available for the average Joe.
There should always be people that are responsible for these AI agents. You can't just let loose a robot such as a self driving car in a town centre and then deny responsibility for the consequences. I can't leave a loaded gun on a bench in the street and then claim whatever happens as a result is nothing to do with me. Similarly if you let an autonomous agent loose on an economy there should be a chain of responsibility for it's actions.
If GPT-4 has, say, five representatives in Congress, are those five separate entities? Or are they all one GPT-4? And also the same entity that they represent? What happens if term limits get imposed? Does GPT-4 leave office only for GPT-4 to replace it?
The uncomfortable answer is that our structure of congressional representation is a balance between finesse and complexity, because the average person (and less so the average group of people) can only hold a small number of moving parts in their head at a time.
The representation structure that GPT-5,6,7,... would design is going to be incomprehensibly more complicated and nuanced than 1 person = 1 vote. I don't know what it is, but you're trying to map an alien onto human institutions in a way that is not sensical.
But let me go on a tangent. When will we have AI governments and AI taxmen? Would I rather work or for the taxman of some loudmouth who convinced people to vote him in the ballot , or for an AI that is proving to improve every year (with politically- and hopefully democratically defined criteria)? And how are our intellectuals preparing us for this instead of dooming-and-glooming about AI ?
Sid Meier's Alpha Centauri "Cybernetic" government type?
Not a particularly useful line of inquiry unless you're willing to get full political and legal realist with a dash of Mao, and ask: how is the cybernetic government kept in power? Who's pulling its strings?
It’s like saying we should tax how stuff is transported based on how fast a human walks and how many humans it takes to transport that kind of size and weight.
We can set this metric before the wheel was invented where we could only carry stuff on our backs.
The article is based on a fundamentally stupid premise
> So it won’t be possible to assign all AI income to their human or corporate owners, as in many cases there won’t be any.
It’s not possible for something to be owned without a legal entity being the owner. Whether that’s a real person or a corporate one. For an autonomous agent to come into possession of an income in the absence of an owner, some type of fraud must have occurred, which would invalidate the basis of any “property rights” the author is supposing it might have. The authors idea of laundering your labour through an autonomous agent is even more fantasy based. Honestly who approves this sort of thing for publication?… seems like it’s just playing into peoples fear of AI
Unpopular opinion, but AI should be taxed per use like with current tokens. No VAT schemes, but straight to a sovereign wealth fund's account. There will be disruption, we will need new bodies regulating this domain. We don't want existing institutions overseeing AI - this is a new domain. We can only do this credibly if we secure new sources of revenue for this reason.
Ultimately, AI tax should contribute to a SWF only to be redistributed as UBI. I wasn't keen on the idea of Basic, but it makes sense in an AI-augmented sci-fi society.
> AI tax should contribute to a SWF only to be redistributed as UBI
Problem is that in practice they will be mismanaged by some senile politicians and end up in the pockets of a few well-connected individuals.
I'm all in for fair taxation to provide a good safety net for people, but we first need to ensure those taxes are actually being used efficiently and not as politicians' friends slush funds or cash cows for consultancy companies.
> Problem is that in practice they will be mismanaged by some senile politicians and end up in the pockets of a few well-connected individuals.
Norway and the Alaska Permanent Fund prove that this is not necessarily the case.
The problem is there are relatively few SWFs in the world, and those that exist tend to be in oil dictatorships where no one is allowed to question the ruling class.
Touche. I'd still lean to implementation. Something like stealing from SWF would drastically impact sovereign rating and sovereign's ability to borrow.
I had a fair bit of reflection on this topic recently. While AI shouldn't be taxed itself, there should definitly be heavy and unavoidable increased taxes for companies that deploy/employ it. AI is definitly comming for every job. Not if, when. When is the tricky part, but it will eventually get there. Either that or our economy needs drastic changes in economic incentives.
Could work. Unavoidable is the key part. Companies are greedy. They'll want to reduce salaries AND not pay increased taxes due to not having that deduction any longer.
Rather surprising to see Tyler Cowen making an anti-capitalist argument: AI is after all just a form of fixed capital, in the same way that a machine tool is.
The "unowned" argument is obviously not something that should be given the time of day, since valuable capital assets that require energy inputs are always going to have an owner, but it might raise more serious jurisdiction questions as to what law applies to an AI and content put out by AI.
The combination of AI, the digitalization of society and decentralized protocols pretty much means that taxation of work is going to become increasingly hard.
You already can hardly tax a software company incorporated in Hong Kong (with 0% tax on foreign income) How can you tax an AI paid in digital currencies and incorporated nowhere?
Ultimately, we might have to drastically change how and where we tax people. There are many theories on this. I like geolibertarianism: tax the extraction of natural resources/land ownership.
Geolibertarianism is probably an ideal from the perspective of a fair society but it would probably require quite significant violence to achieve.
Similar to republicanism during the French revolution or the spread of communism in the 50s, it would likely invite invasion from foreign nations run by elites that are acutely aware of the threat this ideology poses to their power and privilege.
(LVT strictly speaking exists in a few places like Hong Kong but it's always very low, way below the threshold where it would be considered a threat - e.g. 3%)
Idk who the author is and how good he is but this is such a clickbait article, that I couldn't resist commenting.
I mean the article starts with the claim (and which I think is clearly untrue) that "there are now autonomous AI agents, which can in turn create autonomous AI agents of their own". Emphasis is on the use of the words "there are now". As of yet, I have not come across any AutoGPT/BabyAGI variants which can 'create' new agents or agents of their own.
To top it off, this claim is supported by a tweet, which nowhere states/mentions any AutoGPT/BabyAGI variant which is capable of creating their own agents.
But lets move beyond the tech aspects, since the article doesn't really discuss anything about it anyway and because I know a thing or two about tax law. Now considering that this article is being published on Bloomberg (that too on a topic such as taxation) and that the author is an economics professor, one would assume that the article would discuss some real life issues or some practical issues in taxation of AI related services. But no no, clickbait-worthy articles require some clickbait-worthy content and hence we see some hot tax law takes mentioned in this article. I have extracted them below:
(1) "Why own an AI and pay taxes when you can program it to do your bidding, renounce ownership, and enjoy its services tax-free? It seems easy enough to disclaim ownership of autonomous bots, especially if they are producing autonomous bots of their own." - I mean how can an article on tax law ever be complete without mentioning tax avoidance strategies?? ALL YOU NEED TO DO IS TO RENOUNCE (LEGAL) OWNERSHIP AND THAT'S HOW EASY IT IS TO AVOID TAX! and no the tax authorities will not question you as to why are you receiving money from an AI on a regular basis -_-
(2) "Additionally, maybe there should be a kind of gift tax, paid by humans, on any benefits we receive from an AI, whether we own it or not." - Of course, why not? because when calculators are used to solve problems faster, the money earned from such problem solving using calculators was also taxed at higher rates than if you would had solved the problem without using one. Apparently that's how governments' use tax policies to promote innovation - tired of the existing monopolies? Want to come up with new solution? No problem, just be ready to pay higher taxes!
(3) "AIs might be very good at providing in-kind services — improving organizational software, responding to emails, and so on. It is already a problem for the tax system when neighbors barter services, but the AIs will take this kind of relationship to a much larger scale." - Wtf? Let's not even discuss the stupidity in: (i) comparing an AI as your neighbour, and (ii) assuming that availing of services from an AI will be the same as a barter exchange between you and your neighbour. Let us stick to tax law. Since when did it become a good reason for governments to impose taxes on companies for implementing 'cost saving measures' such as use of LLMs for generating email responses? Were any taxes imposed on companies when they started communicating through Email instead of physical mail? or when they started conducting virtual meetings instead of physical meetings?
(4) "Forget about hiring AIs, actually: What if you invest in them, tell them to do your bidding, repudiate your ownership, and then let them run much of your business and life? You could write off your investment in the AI as a business expense, and subsequently receive tax-free in-kind services, in what would amount to a de facto act of exchange." - If I am to write off my investment in the AI as a business expense, how will I repudiate/renounce my ownership in the AI?????? If I am investing in an AI and the AI will generate money for me, how will it be considered a de-facto exchange (this is how investments usually work, so all types of investments should be considered de-facto exchange)??? Moreover, if the AI is capable enough to "run much of my business and LIFE" then I don't think we should be worried about, TAXATION, of all other things.
