Looking at the actual article, the people suggesting taxes on AI are American Nobel laureate Edmund Phelps, and Bill Gates, founder of MSFT. The Europeans suggest more general taxes on capital instead.
Taxing wealth is much harder on a practical and algorithmic level than taxing income.
But either way, taxing the tool is micromanaging the problem, and some powerful people cynically promote that because they can aim the details away from themselves.
Switzerland taxes wealth instead of capital gains (except for professional investors).
In some ways taxing wealth is quite simple, because wealth is already meticulously recorded via contracts and owners go out of their way to estimate the magnitude of their wealth for example to borrow money or for other financial and economic obligations.
Another approach would be to tax capital gains at the same rate as income and introduce additional top brackets. I have a hard time to find a good faith reason for capital gains to be taxed less than labor.
But there are enough other ways where it's really hard and unsolved. Imagine someone bought an $X irreplaceable ancient urn to hold the ashes of their parents.
How do you calculate the $Y "wealth" inside that non-fungible urn on their mantelpiece today? How can one determine which "I would buy that for X" statements are falsely low or falsely high?
> owners go out of their way to estimate the magnitude of their wealth for example to borrow money
I have no inherent problem linking one voluntary claim of wealth to another conclusion of wealth... But what happens when someone wealthy who doesn't actually need any loans applies for them while presenting themselves as a pauper?
Or cases where someone seeks a loan and their rationale is "I may have negative net worth but you'll be made whole because you're first in line", as opposed to "I have high net worth"?
> I have a hard time to find a good faith reason for capital gains to be taxed less than labor.
Consider a small company of AcmeCo with 1-10 workers all dedicated to the art of Acme'ing, each taking tiny wages (but accepting shares) because they believe in the mission and want to launch the company.
On a technical level, anything they (might) get would be capital gains, but clearly it's not the same as passive rents with no labor behind it. It's closer to deferred wages.
Capital gains are risky to generate. Many investments completely fail and when that happens, investors get very little tax relief.
If you increase capital gains tax, the more risky ideas will no longer be viable investment vehicles even though some of them would have been successful. Across the entire economy, the net effect will be less innovation, stagnation, and loss of power relative to foreign countries.
Tax rates are carefully tuned to maximize tax revenue without unduly disincentivizing production. To change them purely based on vibes would be catastrophically stupid.
> Capital gains are risky to generate. Many investments completely fail and when that happens, investors get very little tax relief.
You seem to conflate and misunderstand a few things here. For one, capital gains are paid out from profits, so they leave the company's budget and are not part of investment.
Secondly, holding stock and stock trading has been immensely profitable, relatively low risk and require little if done reasonably and over long periods.
More and more studies are affirming that simply buying index funds and holding them over long periods is _less_ risky than buying bonds.
> Tax rates are carefully tuned to maximize tax revenue without unduly disincentivizing production.
They reflect power relations and tax competition more than anything.
> To change them purely based on vibes would be catastrophically stupid.
Not based on "vibes". The people who actually generate wealth are workers and consumers, not stock holders.
> Please don't vote.
We should mutually respect our rights to vote even if we disagree.
> Many investments completely fail and when that happens, investors get very little tax relief
I don't really believe it. Investment is always incentivized by tax breaks and other political gifts. But once things turn bad it's the citizen's turn to pay for it. Fire all staff? We pay for unemployment. Pollute the soil? We pay for cleanup. Empty the water table? Guess who's gonna depend on the state for clean water...
> To change them purely based on vibes would be catastrophically stupid.
Please tell that to every neo liberal in my country. Reducing taxes on the rich seems to be their passtime, while every time some kind of capital gain is mentioned, everyone and their dogs become experts in economics and can tell you it's folly.
> Taxing wealth is much harder on a practical and algorithmic level than taxing income.
I find this argument somewhat unconvincing.
Where is most of the wealth? In hard assets, such as real estate and financial assets, such as stocks and bonds. The former are very difficult to hide, for obvious reasons. As for the latter, the ownership of every single share is recorded in large databases (e.g. DTCC, Clearstream and Euroclear). In that sense, the "physical location" of most of the wealth is well known, so in theory it should really not be difficult to tax it.
The unit of account for tax is the currency of the relevant sovereign. Most contracts for income are denominated in that unit of account, even if it is not there is often a highly liquid market (FX) between units of account.
Most wealth is not stored in assets where the unit of account is that of the sovereign. This counts double for assets with a physical location.
This isn't something that can be easily hand-waived away.
My understanding is that you say that taxing things denominated in a foreign currency is difficult? But why? I already pay taxes on my capital gains denominated in a foreign currency (for example dollars). There are official government exchange rates for tax reasons, published daily. I don't see anything to hand wave here, because there's no problem.
Not parent-poster, but I imagine the most difficult cases involve non-public stocks or non-fungible physical assets. Consider the problem of: "Someone purchased an irreplaceable ancient urn for $1m and put their parents ashes in it, what's that in taxable wealth today?"
It's too easy for people to offer hypothetical prices they'll never have to execute on. You could establish a price by forcing people to sell anything to the highest bidder, but that kinda explodes any conventional idea of property, and now the government is spending all its time running a trillion sketchy auctions while no human has time to do productive work anymore because your neighbor is trying to buy your car for $1 and you need to arrange a more-plausible offer before you lose it.
Apologies, in an attempt to be precise I have used convoluted language.
The point I'm trying to make is that assets such as land are not denominated in any currency and typically end up being held for such large amounts of time with such substantial transaction costs that's there would be a large cost involved in knowing what the value of the thing being taxed is.
If I pay you $100k, £100k or ¥100k we can use spot rates to work out how many € that is within much less than 1%.
If I own a piece of land how would you answer the question, "what should the value for taxation be?"
If you go with the last transaction price then this will have a substantial impact on properties that haven't been sold for a long time and encourage people to enter into transactions that look like sales but aren't (such a 999-year lease).
Leave it up to a government agency to decide and this agency will come under huge pressure to favour one type of activity over another. How do you value land owned by the government? What if that land is privatised? The UK's attempts to deal with this when it privatised BT completely destroyed the fibre to the premises industry in the UK for years.
Financial assets are extremely easy to hide. Set up an international chain of shell companies, foundations, and trusts, install a fake beneficial owner or trustee or two at various points, carve out deductibles for IP and "services", and the ownership becomes completely opaque.
And that's just the legal version.
I know someone who used to work as a business lawyer. She spent years trying to track down the true owners in various cases. At the very least it's an expensive business. And sometimes it just couldn't be done.
Of course governments can cut the knot with physical assets, walk into a building with troops and/or police, and say "This is ours now." Or they can order banks to hand over the money in accounts.
But before they can do that, there has to be some certainty about the owner. And even getting part way there can take a while and cost a lot.
A flat tax on wealth would be extremely easy to enforce. Basically if the bill for an asset doesn't get paid it goes to the government, and the bill is trivial to calculate because it doesn't need the rest of the entity.
If you apply automatic tax on bonds people will just not buy them unless you also increase their returns. It's a pointless exercise.
Same goes for stocks, it's just a bit bigger circle in this case.
Capital gain tax is just a bad tax that distorts decisions and make things less efficient for no reason. It's much better to tax resources (mainly land but also infrastructure usage) and charge for enforcement of IP/patents.
But financial assets do not need physical space, so they can be tied to smaller countries which will be very happy to tax them at a lower rate so they can "steal them" from the original country where they were generated.
You can take your financial assets with you but they're ultimately worthless, they're just a construct that represents something which has real value, like shares of a company: its real estate, inventory, employees, institutional knowledge, and future productive output have real value, your piece of paper doesn't.
Distribution of wealth is about the distribution of real resources, especially control over human labor. And that underlying thing can always be taxed, optimized, or even repurposed to better serve the needs of society.
You can still tax based on the persons residence or citizenship. In the end someone can be attributed to wealth, and if they want to stay where they are physically, they should also tax like it.
I need to think about this more, but the first thing that comes to my mind is not that this looks like “taxing the tool”, but that this can (ought to?) be similar to an alcohol or a fuel duty.
Nobody calls alcohol duty “micromanagement”.
For products like petrol, it’s widely known that from money paid for a liter when it’s sold, say, in the UK, more money stays in the UK’s government pocket via a complex web of taxes and duties, than profits the oil production company that supplied crude oil for that petrol.
Maybe taxing a kWh of the AI data center energy consumption should be a thing? I don’t know.
They don't, but it really is! There's different rates for different specific gravity and different processes.
Re: petrol, I note that the UK government is trying to replace this as part of the EV transition with a milage tax, which is proving controversial and fiddly.
Energy tax is a hugely fraught political issue. The "poster child" for cheap energy is a little old lady huddled over a 1kW one bar electric heater. Energy bills are a big "fixed" cost for households. Many small businesses have been affected by energy price rises - e.g. restaurants. And yet at the other end AI represents such a huge deployment of capital expenditure that it's distorting prices for everything else - energy, RAM, and so on.
I think I'd favor a "personal allowance" model similar to income tax, where you get the first X units of energy tax free and then have to pay VAT, carbon taxes etc. on the rest of it.
> I think I'd favor a "personal allowance" model similar to income tax, where you get the first X units of energy tax free and then have to pay VAT, carbon taxes etc. on the rest of it.
I can see why this is tempting, but I think there's a better way to legislate with this, especially with that poster child.
I'm a landlord of a flat. I used to live in it before I left the UK. The EPC rating is D, so despite the double glazing it's still pretty cold in winter. I am now living in a fancy new-build in Berlin which, despite being 3 times the size of that flat, can be kept warm for 10 months of the year just by body heat and waste energy from the white goods — even with higher electricity costs in Germany, it costs less to be comfortable in this building in a T-shirt all year round (even while snow is falling outside), than to be wearing fleeces and sleeping with hot water bottles and still not be completely comfortable in that flat in the UK.
A few years back there was a proposal for legislation that would increase the requirements for all rental property to be at minimum C-rated by 2030, as I understand it this was dropped and the current minimum is F or something ridiculous like that. My agent's advice is to not do anything until the legislation is actually sorted, even though I'm happy to spend whatever to upgrade the place, because until you know what the legislation demands there's always a risk of doing the wrong work beforehand, having to rip it out and put something else in.
IMO, government should push for this kind of boost, as it has with other energy-saving and insulation-boosting measures.
My first rental after graduation was a Welsh solid stone wall construction; like the example you gave, I couldn't keep warm there even with the electric bar heater a meter from me.
The current minimum EPC rating is D. The legislation to raise it to C hasn't been dropped, they just haven't decided exactly what date it will take place. And it's stupid legislation because many old properties cannot be sensibly raised from D to C, and these are the properties (e.g. terraced housing) which are typically rented out. So, we have a housing crisis with too few properties available to rent and the legislation will force landlords to take rental property off the market. Madness.
There's a lot of not-joined-up thinking in the UK government. Has been for ages.
Like, the housing crisis in the UK, there's a lot of empty houses, they're just in places with no jobs, could encourage employers to go there, but HS2 mumble mumble. Could build more houses, but greenbelt, and existing homeowners like the house prices going up, and lots of builders were Polish and oh look Brexit.
Right now, winter fuel allowance is literally burning money because the houses are not good enough. This is also not sensible.
I remember there was a pressure group "insulate britain". Their aggressive tactics got them banned and arrested, and the idea was never heard from again. I sometimes wonder if that wasn't the intended outcome, a low-temperature conspiracy theory.
Thinking about who might benefit from it being a conspiracy, the only finger I can point at would be Russia? (Well, unless it's a long-term generational anger at the British Empire, which I have discovered is more of a thing than most Brits realise).
That would be a highly bureaucratic solution with significant overheads.
Would everyone pay extra tax per kWh or just AI computers? Tax it on the producer or consumer side?
How would you verify that a particular data center is "bad computation" and needs a different tax rate on its energy usage.
Should an AI data center from pharmaceuticals or biotech startup be taxed extra per kWh, even if the AI is purely used for medical research?
Just big AI datacenters. If this encourages people to run local AI, all the better.
> Should an AI data center from pharmaceuticals or biotech startup be taxed extra per kWh, even if the AI is purely used for medical research?
That's not a gotcha.. those are all policy choices. My personal preference is, yes, of course - medical research today is taxed just fine. If there's lobbying to specifically grant tax benefits to medical research, I can see an exception being carved.
You think multiple localised heat centres are more efficient than centralised managed heat centres. Why don't we all just have a coal-fired power station in our back garden?
The issue I have with your proposal is that it discloses too much metadata to tax authorities in order to enforce compliance. They'll have an almost perfect map of the legal compute in their jurisdiction. Access to compute should be free to all and not gated by taxes.
Tax on electricity is already a thing. That can be adjusted and even be made progressive. Extra for fossils and so on.
> Taxing wealth is much harder on a practical and algorithmic level than taxing income.
Depends on the tax. It is a lot easier to move move profits to a low tax jurisdiction than it is to move land or machinery.
> But either way, taxing the tool is micromanaging the problem, and some powerful people cynically promote that because they can aim the details away from themselves.
I definitely agree with that.
There are all sorts of problems. Do you tax this notional "income" where the work is done or where the AI runs or where the company that owns it is incorporated?
I think the simplest explanation is also the best one. Like you said, it is very difficult to tax wealth.
I think we have no option other than taxing loans and other money movements like that in sufficiently large scale as ordinary income. If I get a loan for USD 200k for a house once a year, I think it isn't income but if ElMo gets loans worth USD 20M a year, every year, he should pay income tax on all of that as if it was ordinary income. How he pays it? I don't care. Sell some assets. Oh and that sale is also taxable.
The problem is that rich people and large companies usually go to great lengths to avoid taxes, use loopholes or get special deals (and with great success). The missing tax income has to come from the middle class, who can't avoid it.
With increased automation, this only gets more extreme.
> The missing tax income has to come from the middle class, who can't avoid it.
Taxes on labor are actually a method of extracting money form the rich capital owners.
As you mentioned it's easy for the rich people to hide their wealth and avoid taxes on its growth.
The one thing that was very hard for them to avoid or hide was purchasing labor which they had to do to enlarge their wealth. So governments taxed that.
If governments lowered the taxes on labor it wouldn't mean middle class would earn more. It would only result in capital owners paying less for work. They always pay as little as possible and how little a person is willing to work for is the same, tax or no tax. Because money in hand is what counts.
Of course since as labor is being replaced with automation this way of collecting tax on capital growth becomes less and less feasible, so things are bound to change.
Surely if we can recognize this, an AI worthy of the name would be able to recognize this at scale, and what can be recognized can be remediated…
Or perhaps this could serve as a kind of test: a technology that cannot be reliably used in tax evasion enforcement simply isn’t worthy of the name AI.
Or perhaps it reveals that we have structural problems, and certain concentrations of wealth with or without automation are a threat to the just and effective operation of society and should therefore be as vigorously opposed as crime or foreign attacks.
Directly taxing AI is very hard. Imagine if a company had to pay taxes for every AI agent operating in the U.S. or the E.U. As if they were regular employees. Big corporations would simply move the AI agents to countries without taxes.
It's actually trivial. AI apis are pretty streamlined by now. Just slap a tax on processed tokens and you're guaranteed to reach every AI agent out there. It already happens everywhere with sales tax for normal products. Just treat tokens as the product and create an extra tax for it.
Let's say EU and US taxes AI tokens. India doesn't, so almost all prompting done by international companies now is outsourced to India, and still not taxed.
Or do you tax AI companies and tax tokens "at source"? Then, obviously, they either lose competition with foreign (let's say Chinese) companies that do the same but are not taxed, or more likely all AI companies move out of EU and US.
You could tax the energy and subsidize it for individuals. It's the ultimate resource that all business uses. But that would mean unscrupulous countries could tax their energy less and attract AI farms. So probably you need to tax imported tokens (and other goods) as well. There could be many benefits of taxing grid energy instead of labor.
This is how sales tax already works. If you sell something to another country that has sales tax, you need to pay it irrespective of where you produced it.
Audits? Like it happens with licensed software. The issue is that if any country won't play ball with either not adding the taxes or by closing an eye, everyone is gonna put their datacentes there and become un-auditable.
I guess the other countries can slap sanctions on them, but the people benefitting won't care really.
Not only that, do you tax AI that doesn’t replace humans? How can you tell? Do you tax differently depending on how many workers it replaces? How do you measure that? Do you create exemptions for non-profit or humanitarian use? How do you measure that?
I can only image the Kafkaesque tax code the government would come up with. Then it would create all sorts of weird incentives as companies attempt to minimize tax paid.
The OP thinks that candidates spending a lot of time on applications is OK, as long as the company shows respect by spending a lot of time themselves. I think this is mistaken - I care about how much time I have to spend, and am a lot less concerned about how much time the company takes.
There's a trade-off: if a company spends more time / requires more effort on an interview process, they can get a better signal on the candidate's abilities, but then they'll lose out on candidates who are unwilling / unable to commit this time. This might just be a hard trade-off in recruiting.
Excellent point. And for anyone who’s been a hiring manager / recruiter, you know how many candidates you will have to sort through. And you want to waste as little of your engineers’ time making them do interviews if possible.
Internet applications have made it so easy to apply to a position, companies have to find (usually arbitrary) ways of filtering the pipeline.
It’s a very difficult problem to solve - Coinbase had 500k applications for 500 positions.
Edit: I’m very concerned about AI tools flooding the pipelines even more by sending out tons of automated applications. This is going to cause an arms race where the companies have to use more arbitrary methods to sort through candidates, and it will only make it harder to find good ones.
But, yeah, it's not that, back in the day, I didn't post a ton of application resumes and form cover letters to HR departments out of school--and even got non-form responses from a number (and an offer from one sight unseen though I ended up going with someone else even after insisting on an in-person visit). But my sense is that, as you say, there's more of an arms race as you put it going on today where--if you don't have some way of cutting trough the noise, such as through your network, it's a tough slog. Which is one reason the anecdotal evidence at least suggests it's tougher for people who have't developed a network yet.
Historically, I'm not sure that isn't fairly normal.
But compared to maybe the decade plus prior to a couple years ago for (especially junior) software developers, it seems like a tough market based on a lot of conversations irrespective of overall unemployment rates.
Last time I was looking, a year or so ago, I sent out dozens of applications and got zero interviews. Last time I switched jobs, in 2022, I sent out the same amount of applications and >50% led to one or more interviews (and eventually two offers).
In 2023, I was Amazoned after 3.5 years and found myself looking for a job. Even worse, by then I had moved to an area that was tourist heavy, but not really tech heavy and was looking for remote only jobs. The remote role at AWS just kind of fell into my lap.
Plan A: Leveraging my network. This led to two offers both architect level positions - one over the architecture and migration to AWS at f500 company and the other over the architecture of PE owned company that was doing rollup acquisitions (been there done that).
Plan B: targeted outreach to companies looking for expertise in a niche of AWS
where at the time, I was one of the industry experts and was a major contributor to a popular official “AWS Solution” in its niche.
This led to two interviews and one offer.
I basically had three offers and a side contract within ten days of looking.
I’m not bragging, I’m old. I should have a network to lean on and a reputation. The point is that even with my experience, if that well above had run dry, I might have been screwed.
I also applied for hundreds of jobs during those two weeks while going through the interview process with my first hits. I heard crickets and for every job I applied to, it had hundreds of applications and my application wasn’t even viewed by most and my resume was only downloaded once (LinkedIn Easy Apply shows both).
I was not some old guy (well I am) with outdated skills. At the time I had over ten years of software development experience (on my resume) and 7 years of AWS experience including 3.5 working at AWS (ProServe) leading relatively complex projects.
>The OP thinks that candidates spending a lot of time on applications is OK, as long as the company shows respect by spending a lot of time themselves
if I spend 6 hours and the company has 1000 employees does that mean they spend 6000 hours? If so I might consider it a reasonable line of argument, but I guess they don't spend anywhere near that.
I think it's reasonable to make a donation on behalf or provide an honorarium if someone makes it far enough into the process. Something like a $250 gift card or equivalent.
Not enough to make it worth farming interviews for compensation, but enough to show that the company appreciates that you spent 2-4 hours working on their take home.
This is an interesting use of AI. If we just want to know what Charles III looks like, photography has solved that problem. So a portrait is interesting because it's saying something about him, rather than being a photo-faithful likeness. But what is the robot (or rather, the robot's programmer) trying to say here? Any ideas?
Trivially, the measure of how much it costs in dollars to drive into Manhattan along the affected routes has gone up. So there are likely some people who are worse off. It's rare to have a completely free lunch, but this one looks pretty cheap.
Number of people saying that you should just make sure you have backups. That's true, but there's still a role for government to play to prevent this sort of thing. We don't let companies sell poisonous food - why do we let them offer digital services that can be arbitrarily frozen?
It only appears rare for two reasons: the proposal consists of a literal fence, and we’ve no idea how many such failed or denied proposals have come before it.
Surely it must be well-known that it is exceedingly rare, but valuable, for an organization to document why they didn’t do certain things.
There are a number of industries where you may need experience to become more productive than AI, but nobody wants to hire you when AI is more productive in the first couple of years. Is there a good equilibrium for this, or does it end up with each company saying "We won't hire the juniors, just the experienced people" and then finding there aren't enough experienced people around?
There are many industries where you need lots of experience before you're a net contributor to productivity. This is true for everything from hairdressers to doctors. We have ways of dealing with this (eg. taking out loans to undergo years of training).
The problem comes if the number of years of experience you need to outperform the frontier AI models advances at more than 1 per year, which is not out of the question.
I think the solution is the same as it was in previous cases. Extend the education and make it more accessible so you can reach useful skill level before you dirty your hands with commercial work.
In the old times 12 year old could have economic utility. Now 26 year old often has none. It might be that with AI you might need to keep learning till 35 before you can usefully contribute to the economy.
Which leads to the obvious question of: who's footing the bill for this?
Is the taxpayer going to pay for another five or ten years of education for people? Are the young people expected to borrow hundreds of thousands more for training? Are their parents expected to house and feed them for another decade?
Not who, what. Most of the value in the economy is not produced by humans for a long time already. Most wealth comes from machines. So they need to be footing the bill instead of just lining pockets of people who bought the machines.
This used to be addressed by the fact that people were loyal to companies - so it was in the company's interest to spend years training them up investing in them, with the knowledge that they might get decades of productive work out of them afterwards. One of my grandparents joined a company as an apprentice as 16, got trained by them, and then worked there for 40 years until retirement.
But nowadys with the culture being much more to to repeatedly jump between companies looking for salary increases, there's a lot less incentive to train juniors - because odds are they're just going to get poached or jump ship before that investment has really paid off.
The big companies or startups with VC funding and deep pockets will always be able to hire experienced people - but it's going to become increasingly hard for other people (and particularly public sector and nonprofits) to do so, as the pipeline of juniors -> seniors is being eroded.
I'm sure you can find plenty of example of both. But TBH, I don't think it really matters which side you try and point the finger at after decades of decline - the point is that the employee/employer relationship has fundamentally changed, and it's hard to see it ever changing back.
Agreed. It's not about blame, just that employers are almost always working with a better perspective and more information.
Business circumstances made it advantageous, and then necessary, to break the social contract.
It might be possible to go back, but I can't imagine any series of events that leads in that direction which doesn't break the global economy in the process.
That's such a valid concern, and I think it applies across many industries — including home decor and interior design. While AI tools can suggest layouts or color palettes instantly, what really matters is human intuition, creativity, and an understanding of client emotions something that only comes with real-world experience.
At smithinteriors, we often balance AI-driven tools with hands-on creative insights. But we also believe in giving junior designers a chance to learn and grow, because without mentorship and opportunity, there’ll never be enough experienced professionals in the future. It’s all about creating an environment where AI supports human creativity, not replaces it.
Perhaps by juniors banding together and making new startups which outcompete the dinosaurs - a tale as old as time (or tech startup capitalism, anyway, so since the 1990s).
Peter already replied to this elsewhere in the thread. To a question
"Maybe a dumb question, but... I'm a Canadian who would qualify for a TN visa if I worked in the States, but don't currently have a visa or green card. On the online submission form for job openings, it always asks 'Are you legally entitled to work in the US?'. Am I meant to answer yes or no to that?"
Peter replies:
"Unfortunately, the correct answer is No because until you have the TN, you are not legally entitled to work in the U.S. Of course, this means that you will be excluded automatically for consideration of certain jobs."
From 0511 local, originally posted on X: "Due to a computer networking problem BART service is suspended system wide until further notice. Seek alternate means of transportation. Find more info at http://bart.gov/alternatives"
