Sorry, but I really do not understand the point you are making.
Consumers in the EU pay the same VAT on products from the EU as from products from the US. There is no unfairness. Businesses in the EU do not have to pay VAT, whether from the EU or from US.
The only "additional tax burden" that US companies face are US internal taxes.
You are technically correct however your comment overlooks the structural disadvantages that US businesses face due to lack of VAT.
VATs are pass thru to the consumer as other comments on here have indicated. There is no inherent discrimination between US and EU products in terms of VAT rate. It’s neutral within the country. A US company selling to the EU still has to factor in US corporate taxes in its pricing and does not get a VAT rebate. And due to structural disadvantages, US does not have a VAT to refund like EU countries when they export. EU countries can price their goods more competitively.
Given those facts, Trump views VAT as a hidden tariff that puts the US at a disadvantage.
EU countries cannot price their goods more competitively. EU businesses aren't charged VAT by the EU on goods they export because consumers pay VAT. But when they export goods to another country, tax is still paid. It is just paid by the consumer in that country, including the US.
For example, if an EU business exports to NZ then they have to remit 15% GST on the goods to the NZ tax authorities. The same is true of Australia or the US or any other destination with sales tax.
The US has relatively low sales taxes, so its consumers have an advantage when importing over consumers in other countries. That is a good thing for US consumers: they pay less for goods. It has nothing to do with US exporters.
The US doesn't refund VAT to exporters because there is nothing to refund: they don't pay VAT on their inputs. The refund is fiscally neutral for businesses. The EU exporters that get VAT refunds aren't netting anything. They are just getting back tax they paid but never owed. US exporters don't pay any sales taxes on their inputs so they don't need refunds: they kept their money in the first place.
EU countries also pay income tax. In no possible sense are EU companies advantaged by the EU being a generally higher-tax place than the US. The opposite is true: if VAT were lower, they would be more competitive domestically and would have be better resourced to invest in their processes and become more efficient and competitive.
Without going line by line, I will say that you make all valid points and are correct.
But you are ignoring key trade competitiveness issues due to the way VAT and US tax system is structured. EU system incentivizes trade exports by making sure they leave the country tax free. US does not.
The disadvantage is not the VAT itself but lack of equivalent tax relief. The real fix is to tax policy but Trump’s only way to fix it right now is via executive action using tariffs. It’s sort of a blunt instrument but it can be effective for what he is trying to achieve.
>EU countries can price their goods more competitively.
Assuming we talk about US consumers -- EU manufacturer is more (?) competitive by not paying VAT on their manufacturing, the same way US companies don't on their domestic production. Where is an advantage exactly?
>A US company selling to the EU still has to factor in US corporate taxes in its pricing and does not get a VAT rebate
Which VAT rebate they should get if they and their suppliers didn't pay VAT in the first place? They get back 0 USD of the 0% VAT rate in US. Sounds about right.
There is a structural disadvantage. You are correct that both the US and EU do not pay VAT when exporting. EU countries get full VAT rebate when exporting. But, the US primarily taxes businesses via corporate income tax (~21% federally, plus state taxes). EU follows a VAT system which is refunded on export. Corporations don’t get refunds on exports of their own countries prescribed tax. EU companies can structure their pricing differently since it is leaving the EU without a tax burden. EU exports tax free.
On your other questions, I think you are missing the bigger picture. The part you are missing is that EU exporters get a tax refund. The rebate the US is missing is an equivalent tax relief system. VAT-based economies remove tax burdens for exporters and keep them for importers.
To be clear, the EU is not “cheating”. Trump just sees the entire system resulting in a trade imbalance and the only mechanism he has right now is tariffs without overhauling the entire US tax code or some other similar relief system. There are other options that could be implemented like this and the one chosen was the tariff.
>On your other questions, I think you are missing the bigger picture. The part you are missing is that EU exporters get a tax refund.
I'm not missing it, you are missing it. Yes EU companies get a tax refund and US companies don't, because there is nothing to refund when nothing was paid.
>But, the US primarily taxes businesses via corporate income tax (~21% federally, plus state taxes)
Believe it or not, but corporate income taxes in Europe also exist. I can see the argument that tax burden for companies in US is higher because government collects more in income tax and less in VAT tax. I'm not sure that is actually true and strongly suspect it's not, as EU is anything but a low-tax jurisdiction. If anything, that's a good argument for US to introduce VAT tax.
You are correct that US exporters don’t get a VAT refund because they don’t pay VAT and corporate tax exists in both systems. But US exporters still face a tax burden that EU exporters do not. Introducing the VAT would fix this particular issue but it doesn’t fix everything. Trump isn’t wrong that the EU system is giving an edge to their exporters. In the US, VAT has been seen as regressive and hurts lower income consumers more. Changing to VAT would be politically difficult and those major changes to the US tax code have been resisted by Congress. US could offer credits on exports to act sort of like the VAT system. I am not sure what kind of impact that would have. That tax would still have to be collected elsewhere. Or corporate taxes could be reduced on exports.
Any US goods exported to a VAT country (e.g Germany, France, Italy, Japan, etc) and the industries affected are automotive, manufacturing, agriculture, technology, pharmaceuticals, etc. All of them
Some examples
A Ford car exported from the US to Germany must be priced after paying US corporate tax (21% federal + state taxes)
A BMW exported from Germany to the US gets a full VAT refund (~19% in Germany), making it tax-free on export
US does not refund taxes on exports, but the EU does
My final points are that EU exporters sell their products abroad with no tax burden due to VAT refunds and US exporters must price in US corporate taxes
Final example
Boeing (US) exports planes priced after US corporate tax (~21%)
Airbus (EU) exports planes after receiving a VAT refund (~20%), making their planes cheaper in global markets
This is why Trump and others see VAT as a “hidden tariff”
US corporate tax is on net income (profit), not on revenue. Ford's export price still carries the cost of US corporate tax
You assume the argument is that BMW doesn’t pay corporate tax but that is not the point. You assume the US is taxing exports unfairly. It isn’t, but it also doesn’t provide tax relief like VAT systems do.
You are ignoring the fact that VAT refunds lower the export price for BMW, but Ford gets no equivalent tax break. Even though both BMW and Ford pay corporate taxes, BMW’s exports are tax-free because VAT is refunded.
The real issue is structural tax asymmetry.
The EU removes tax burdens from exports (via VAT refunds) and US does not because it has no such system. This makes US exports less competitive in VAT countries like Germany.
The entire system makes for a trade imbalance for US exporters to the EU but there is no fair equivalence in EU hence Trump’s actions and remarks.
>The EU removes tax burdens from exports (via VAT refunds) and US does not because it has no such system.
US didn't collect the tax in the first place, so there is no burden to relieve the company from. Refunding tax on exports makes it symmetric, otherwise EU would tax exports that US doesn't. I don't get why you bring up corporate taxes here at all, as they are present in both situation and are not refunded.
I doesn't sound like you are arguing in a good faith to me.
>This makes US exports less competitive in VAT countries like Germany.
No it doesn't. Both imported and locally produced goods will have the same tax applied to them. Both local manufacturers and US companies have to price in their corporate taxes. Nobody pays the VAT except the end consumer in the end.
US collects tax through the entire process and at the point of sale to the consumer. EU refunds the entire VAT so on that point you are wrong. EU exporters have a tax as advantaged export. This is mostly because the US does not have a VAT system.
Okay so the problem is actually collecting corporate taxes. A solution would be perhaps an international treaty for a global minimum corporate tax rate?
Trump left that treaty recently, oh well.
Of course the exported car is not "tax-free". There is a 19.325% corporate tax rate + municipal tax on corporations in Germany. This is already pretty close to the one you mentioned. Then labour is more heavily taxed in Germany. And you have a sales tax in the US as well, you haven't taken that into account.
The argument is not that BMW doesn't pay corporate tax. The issue is that BMW gets a VAT refund on exports, while Ford does not get equivalent tax relief.
Remember that the sales tax applies only at the point of sale, while VAT is applied at every stage of production but refunded for exports. This is what I have been saying all along.
Even though BMW still pays German corporate tax, the tax on exports is removed via VAT refunds.
The real solution is that the US needs export tax relief to offset corporate tax on exported goods. Trump’s tariffs do not fix this - it is just a temporary fix to the trade imbalance. A global corporate tax agreement does not solve this issue either as that is not the root of the issue.
Ireland is a great example because it is corporate friendly and has lower tax rates plus a VAT system. Dell (Ireland) can export tax free while Apple cannot. So Apple has to absorb the full corporate tax burden. Granted both are PC makers but sort of a different class or consumer target. But you get what I am saying.
You are misunderstanding. Both VAT and sales tax are only paid by the end consumer. The main difference is that VAT is progressively collected at each production step whereas sales tax is only collected during the final transaction to the end consumer.
If a good is exported, no VAT nor sales tax is due in the export country. But because some VAT has already been collected in the earlier production stages, that VAT needs to be returned to the exporter. No sales tax will have been collected at the time a good is exported, hence there is none to return.
In both cases, the end result is that no VAT or sales tax is paid in the export country. Hence the situation is completely the same.
Corporate taxes are orthogonal to this discussion. Both US and EU have them and they apply to corporate profits, independently on if they derive from exports or domestic sales. So no reason to complicate the discussion by bringing them up.
VAT countries can export tax-free because VAT is refunded
Even though VAT itself is not a tariff, its refund system effectively subsidizes exports by removing domestic tax costs
You are missing the trade impact of VAT refunds though
It is a fact that US goods tend to be more expensive in VAT countries like the EU because they face an extra tax burden upon import that EU goods do not face when exported to the US. This extra burden is the VAT that the consumer now has to pay. Whereas when it’s exported to the US, the US consumer does not have to pay the VAT so the trade playing field is not equal.
That’s why Trump sees VAT as a “hidden tariff.” Even though VAT is not a tariff, it functions like one by favoring VAT-based exports
Yes, you are indeed repeating yourself. Unfortunately, that does not make what you are saying more right.
VAT countries export VAT free. Sales tax countries export sales tax free.
When you import something to a VAT country, the end consumer has to pay VAT. When you import something to a sales tax country, the end consumer has to pay sales tax.
Maybe we can make some progress if you could explain how these situations are not completely analogous, and what precisely is it that gives VAT countries an unfair advantage?
You mean the problem that sales tax is applied on inputs in the US sometimes? How do you know Ford is affected by this and hasn't structured their supply chains in a way to avoid this problem?
As mentioned in the article this is a domestic problem as well in inter-state commerce. I would also be worried that it distorts supply chains.
Tariffs are not a temporary fix for this either, since the USD will just appreciate and perhaps we'll put counter-tariffs on your tariffs. You'll need to fix this problem on your side.
I agree it’s a structural problem. A few different comments I have said this. Tariffs are one option and it’s the one Trump has chosen. Apparently it has triggered the EU. Trump can’t make Congress overhaul the tax system to include a VAT.
You might as well argue that it’s the US tax system that puts the US at a disadvantage regarding exports. Similarly, the relative lack of worker rights in the US could be argued to put US producers at an advantage. This is economies competing against each other, but the VAT is applied equally to EU and foreign producers, so it seems really odd to blame the very factor that is actually neutral here.
I am arguing that the US tax system puts them at a disadvantage and the option that Trump has available to him without Congress fixing the perceived trade imbalance is executive action through tariff.
> A US company selling to the EU still has to factor in US corporate taxes in its pricing
A EU company selling to the EU also has to factor in their local corporate taxes. There's no structural difference there. (There are EU countries where corporate taxes are higher than the US and ones where they're lower).
> and does not get a VAT rebate.
They only get no VAT rebate because they paid no VAT. There is no structural advantage or disadvantage here: neither the EU or US producer paid any net VAT.
> Given those facts, Trump views VAT as a hidden tariff that puts the US at a disadvantage.
Given that those "facts" are demonstrably untrue, would you agree that the "hidden tariff" narrative is actually untrue?
I believe you are misunderstanding the core trade issue by focusing too narrowly on VAT as a tax rather than the overall tax structure difference between the US and VAT-based economies.
US companies must factor in corporate tax and foreign VAT when selling to VAT countries.
A example is a German car manufacturer exporting to the US gets a full VAT refund and only has to price based on cost + profit.
For VAT countries, the export price is lower because it does not need to factor in corporate taxes like the US companies do.
Finally, to your last point. Trump is not completely wrong just people are oversimplifying. His use of calling a hidden tariff is not technically correct but it is a trade barrier. Trump’s current use of tariffs is not the best fix but he is still right on the trade imbalance.
VAT is not a tariff, but its effect on trade functions like one when compared to the Us tax system.
> A example is a German car manufacturer exporting to the US gets a full VAT refund and only has to price based on cost + profit.
> For VAT countries, the export price is lower because it does not need to factor in corporate taxes like the US companies do.
This is just totally untrue. A German car company will pay German corporate taxes on their profits, no matter whether the cars that produced those profits were sold domestically, exported to other EU countries, or exported to the US.
You are correct that in both cases corporate taxes are factored in and simply put there is no difference at the profit level.
But a US car manufacturer does not get an equivalent tax relief when exporting, meaning the export price still carries corporate tax costs.
When a company sells goods domestically in a VAT country (eg Germany), the final price includes VAT (eg 19%). But when they export, the VAT is refunded, lowering the effective price of the exported goods.
Happy to see some progress here, I trust you will stop bringing up corporate taxes as a distorting element.
Next up, what you need to understand is that there's no tax relief happening here either.
When the end consumer is in the US, nobody pays any VAT on anything in the value chain. Not the US company, not the EU company.
And conversely when selling to the EU, exactly the same amount of VAT is paid in the end, no matter how the value chain is distributed among different jurisdictions.
Where's the unfair asymmetry here?
VAT is a consumption tax. It's conceptually paid by the end consumer. It's not paid for goods exported to a non-VAT country because it's up to the destination country to decide how they want to tax consumption, not the source country.
I have explained this across multiple threads for multiple people.
To put it simply, VAT-based countries have a built-in trade advantage because their exported goods leave the country tax-free, while US goods do not get equivalent tax relief when exported.
You are right that VAT is a consumption tax and is paid by the consumer in the destination country.
However, given the choice, a consumer will typically choose the cheaper option which is often the option that does not carry embedded tax costs from the country of origin. Since VAT refunds remove tax from exports, goods from VAT-based economies often have a competitive price advantage in global markets.
VAT returns ensure that exports leave the country VAT free, just like exports from the US leave the country free of sales tax. It is entirely analogous. How do you mean this gives VAT countries an advantage?
This is an answer for your above comment because HN has a timer delay on responding.
You have a common argument
Because VAT was already collected during production, then VAT refunds remove that embedded tax cost. U.S. sales tax is only applied at final sale. There is no sales tax in production for EU
Even though VAT refunds seem analogous to US sales tax exemptions, the key difference is that VAT refunds remove tax burdens from exports in a way that US sales tax exemptions do not
You seem to be implicitly assuming a closed system in which both economies have the same tax burden, economic productivity and efficiency. So, when EU exporters don't pay VAT on exports there must be a cost for the US producer somewhere that he cant write off.
Which you haven't shown.
You started with corporate tax rate but from my understanding those are pretty similar if being paid at all:
"• Over the past three years, Tesla has reported $10.8 billion in U.S. income but paid only $48 million in federal taxes, bringing its effective tax rate to 0.4 percent—far below the 21 percent corporate tax rate."
https://www.nationofchange.org/2025/01/31/tesla-paid-zero-fe...
You are correct that the overall tax burden, economic productivity, and efficiency are not identical between the US and VAT-based economies like the EU. However, the VAT refund system is what I have been saying is what causes the imbalance along with differing tax structures. This is not a dig against the EU as some people might be thinking I am saying. I am merely explaining the thought process. Devil’s advocate.
US is still disadvantaged despite subsidies for EV for like Tesla that were created to encourage clean energy. There is still a system wide US problem. This is despite Tesla receiving subsidies that are far below the 21% corporate tax rate. The tax is still there and then you have VAT that can’t be refunded because the US does not have a VAT system.
> Even though VAT refunds seem analogous to US sales tax exemptions, the key difference is that VAT refunds remove tax burdens from exports in a way that US sales tax exemptions do not.
I’m sorry, I fail to see the difference. In one case the tax is added during production and then returned when the good is exported, and in the other case it is never added at all. In both cases you end up with zero net VAT/sales tax paid. Where exactly does the difference lie?
The key difference lies not in whether VAT or sales tax is applied to exports, but in how VAT-based economies structure taxation throughout production versus how the US does.
The imbalance exists not because VAT is unfair. It exists because the US system does not have a VAT that can be “discounted” and therefore US exporters end up paying more and baking the cost into their products. Consumer ends up paying more on the other EU end.
In VAT systems, businesses prepay taxes at every stage, then receive a full refund on export, removing embedded costs. I think I said that? There is no equivalent mechanism on the US side.
Bottom line is US businesses cannot discount or deduct state and local sales taxes on purchases made from their suppliers in the same way that VAT operates. Correct that sales tax is not paid on export.
Yes, you keep making this assertion about "tax relief", but you have yet to provide a concrete example of it. (Other than the ludicrous claims about EU exports being exempt from corporate taxation.)
I don’t know if you can’t read or not but I was responding for hours and included a Ford vs BMW example multiple times. Maybe English is your second language. I will go a little slower for you
Here is another one
Caterpillar (US) manufactures construction equipment and exports to Germany
Liebherr (Germany) manufactures similar construction equipment and exports to the US
When producing a bulldozer in Germany, Liebherr pays 19% VAT on all components and services (steel, electronics, labor, etc.). When Liebherr exports the bulldozer to the U.S., Germany fully refunds the 19% VAT. This lowers Liebherr’s export price by 19%. Liebherr sells the bulldozer tax-free in the US. The only tax applied is the US. state/local sales tax (6-10%), which is much lower than 19%. Liebherr bulldozer is now more price-competitive in the US. than Caterpillar’s
You spent hours equating US corporate taxes to VAT, not examples that had any connection to reality. When it was pointed out that you were confused, you did not actually engage with those rebuttals, but just copy-pasted the same nonsense again. So I think it's pretty obvious who here can't read.
In your new example, Caterpillar pays state/local taxes for their sales to the US, corpirate taxes to the US, and no VAT to the EU. Liebherr pays state/local taxes for their sales to the US, corporate taxes to the EU, and no VAT to the EU. Where is the imbalance?
There's no export subsidy here, hidden or otherwise. Everyone is on a level playing field.
You could do the same exercise on US exports to EU vs. domestic production in the EU, and it'd likewise show that there are no artifical barriers to imports to the EU
> This lowers Liebherr’s export price by 19%.
No, it really doesn't. The export price is exactly at the level it would be if there were no VAT at all.
> Liebherr sells the bulldozer tax-free in the US.
I get it - you don’t understand how this works. One more time
You are correct that that all companies pay corporate taxes. We have established that.
VAT is already baked into the cost of production at every step in a VAT-based economy. If VAT were not refunded, that 19% tax cost would be included in the price, making the bulldozer more expensive abroad. Technically, VAT refunds are not an "export subsidy" in the traditional sense (where the government directly pays companies to export). BUT, the effect is the same as a subsidy and it removes tax costs from exports, giving VAT-based economies a competitive advantage. The US does not provide tax relief for its exporters and does not impose VAT on imports.
"Everyone is on a level playing field"
Wrong. The US does not refund taxes on exports, while VAT countries do
When you are ready to understand that part let me know.
> The US does not provide tax relief for its exporters and does not impose VAT on imports.
It does. Exports from the US are exempt from sales tax, the same way others exempt their exports from VAT. Neither side pays domestic consumption taxes on exported products, and both sides charge consumption taxes equally on all domestic consumption, no matter where the products were produced.
Both American and European products cost a little more in Germany due to 19% VAT, and both cost a little less in the US due to 7% sales tax. Nobody gets a free ride.
US is not exempt because it accumulates throughout the entire process whereas VAT is refunded. US bakes the sales tax on acquired parts from their suppliers into the cost of their exported products whereas the EU exporters have an advantage. It’s a structural problem with the US tax system in this case.
Raw materials like steel used in bulldozers, as well as machinery and supplies, are often exempt from sales tax, especially when the finished products are exported.
Nevertheless, in some scenarios, sales tax can indeed cascade and be charged multiple times during the manufacturing process, so even if it's not charged at the final step, there's still a component of it in the final price to the buyer.
This is not a hidden tariff by other countries against US products. Other countries have no control over this. The US government simply sucks at taxation because it has not transitioned to VAT, and products cost more to manufacture than they would have under a better tax system.
It’s your opinion that it’s better. Not everyone agrees.
Tariffs are legal under US law to apply. EU also needs to step up their military contributions before Russia bulldozes them.
I'm far from the only on holding this opinion. 175 out of 193 countries have transitioned to VAT because of the obvious benefits.
Trump shaking more money out of US consumers with new taxes on imported products does nothing to fix the issue. As the article points out, domestically produced and consumed products suffer from the same problem.
> When producing a bulldozer in Germany, Liebherr pays 19% VAT on all components and services (steel, electronics, labor, etc.). When Liebherr exports the bulldozer to the U.S., Germany fully refunds the 19% VAT. This lowers Liebherr’s export price by 19%. Liebherr sells the bulldozer tax-free in the US. The only tax applied is the US. state/local sales tax (6-10%), which is much lower than 19%. Liebherr bulldozer is now more price-competitive in the US. than Caterpillar’s
Let's follow your example more closely.
Caterpillar manufactures a bulldozer in the US. The price tag for the bulldozer is 200k. When sold to domestic customers, a 7% sales tax is charged by domestic authorities, and the bulldozer ends up costing 200+14=214k for the American buyer. When exported, no sales tax is charged due to the export exemption. The bulldozer arrives in Germany, where German tax authorities charge 19% VAT, as they do on all sales. The bulldozer ends up costing 200+38=238k for the German buyer.
Liebherr in Germany manufactures a bulldozer too. The price tag for the bulldozer is 200k. When sold to domestic customers, 19% VAT is charged by domestic authorities, and the bulldozer ends up costing 200+38=238k for the German buyer. When exported, no VAT is charged due to export exemption. The bulldozer arrives in the US, where US tax authorities charge 7% sales tax, and the bulldozer ends up costing 200+14=214k for the American buyer.
In either case, the American buyer gets a 200k bulldozer for 214k, and the German buyer gets a 200k bulldozer for 238k, regardless of whether it was domestically produced or imported. The difference in the total cost to the buyer is due to differences in local tax rates and not unfair trade rules.
Caterpillar's bulldozer is $55,000 more expensive in Germany than Liebherr’s. Liebherr benefits from tax neutrality (VAT applies equally), while Caterpillar faces a 19% VAT on import. This confirms that VAT-based countries create a "home-field advantage" by making foreign imports more expensive through VAT, while domestic goods are not taxed differently.
Please break down how a 200k Caterpillar becomes 55k more expensive for a German buyer compared to a 200k Liebherr manufactured in Germany.
The US government does not charge consumption taxes on a Caterpillar exported to Germany, so both are priced at 200k before taxation in Germany. The German government charges its 19% consumption tax equally, regardless of where a bulldozer was made, so both end up costing 238k for the German buyer.
Liebherr’s bulldozer is priced at €500,000 (~$540,000 USD equivalent) in Germany. Total price paid by the U.S. consumer is $485,545 due to VAT refund. Caterpillar’s bulldozer is 22.5% more expensive in Germany than in the US.
There is approximately a $109,455 price advantage for the German company in global competition
> Caterpillar’s bulldozer is 22.5% more expensive in Germany than in the US.
Liebherr's bulldozer is also more expensive in Germany, because Germany taxes both sales at a higher rate than the US.
Nobody gets an advantage from this. American buyer pays 7% tax and German buyer pays 19% when they purchase a bulldozer, regardless of where it came from.
I made a small table to illustrate it:
+-----------+------------+-----------+----------------+------------+
| Direction | Base Price | VAT (19%) | Sales Tax (7%) | Total |
+-----------+------------+-----------+----------------+------------+
| US -> US | 200 000 | - | 14 000 | 214 000 |
| G -> US | 200 000 | - | 14 000 | 214 000 |
| G -> G | 200 000 | 38 000 | - | 238 000 |
| US -> G | 200 000 | 38 000 | - | 238 000 |
+-----------+------------+-----------+----------------+------------+
Tax rate differences lead to price variations across markets, but VAT does not create advantages within a single market.
For a model to be ‘fully’ open source you need more than the model itself and a way to run it. You also need the data and the program that can be used to train it.
Is it reasonable to expect companies to redistribute 100TB of copyrighted content they used for their LLM, just on the off-chance someone has a few million laying around and wants to reproduce the model from scratch?
With LLMs, the list doesn't even have to be kept up to date, nor the links alive (though publishing content hashes would go a long way here). It's not like you can get an identical copy of a model built anyway, there's too much randomness at every stage in the process. But, as long as the details of cleanup and training are also open, a list of training material used would suffice - people would fetch parts of it, substitute other parts with equivalents that are open/unlicensed/available, add new sources of their own, and the resulting model should have similar characteristics to the OG one who we could, now, call "open source".
That is incorrect -- you do not have to provide the full training data to meet the requirements.
I recommend reading the actual Open Source AI Definition[1] and the FAQ[2]. There's also the whitepaper[3] that goes into much more detail about the state of affairs.
FYI their open source ai definition released with a lot of controversy, unsurprisingly because it had heavy contribution from corporations with their own interests. It's best to ignore it for now until the wider community has decided on an appropriate open source definition.
Yes. I cannot comprehend this to this day. A model weights data + runner is how different from a closed source executable? Why do everyone call these open source?
Because typically adapting or improving traditional code to your needs is very difficult without access to the source code and build files.
For an LLM you can finetune and enhance, distill and embed given just the model weights, the runtime, and a permissive license. Having more is better. Well written detailed model release papers help a lot. Training code and training data are a great bonus.
However, I find the purity contest a bit too dismissive of the great contributions to the AI dev ecosystem that Meta and Deepseek have brought us. Without these, there wouldn't be the open ecosystem we have today.
It's not a purity contest, it's a clarity contest. If Meta and Deepseek want to operate the way they have been, where they release baked models and whitepapers, that's fine - and you're right, it's certainly more than they're obligated to release. They just shouldn't be calling it "open source" when the source is literally not open.
Eh, I can kinda see it. It depends on your definitions of words. People have been muddying the waters with what "open source" means anyway. I have known it to mean code released under an open source license. Other people use it to mean programs where the source is available regardless of license. I would use "source available" to describe that, but some people strongly disagree with my definitions.
If I write a program, then obfuscate it and then release the obfuscated code under an open source license, would you consider it open source(I would)? That's kind of the case here, they are releasing the model weights under an open source license.
Personally, I think it's fine to shorten it to "open source model" instead of "a model with the weights released under an open source license". What I would object to is releasing model weights under a restrictive license and calling that open source.
> If I write a program, then obfuscate it and then release the obfuscated code under an open source license, would you consider it open source(I would)?
I wouldn't. Most definitions of open source say something like "in the form used for editing". You can release a built binary under an unrestrictive license, but that does not mean that you've opened the source. It's literally the plain meaning of the words: the source, as in where the thing comes from, needs to be open for it be meaningful.
Because practically speaking, you can fine tune them I suppose?
But that's also true for binaries, games are a good example of where people pushed this quite far. Based on what little experience I have in ML, I'd say it's about the same thing. Whereas an API is more akin to a piece of software you can't tinker with in any way.
By what appears to be the logic for "open source AI", a locally executable proprietary program would be "open source" (because you can meddle with the executable). To me, that's mostly just "not SaaS". But somehow, a different definition appears to have stuck for LLMs than for other types of software.
Open source becomes really complicated once assets with unclear license are involved in any way. Lots of people for example would say that Jedi Knight 2 is open source because Raven Software released the source code and tools needed to build the game. But that alone doesn't mean you can run it, because you still need to get a hold of all the assets (models, textures, sounds) which may or may not still be property of LucasArts or its successors. Even if you have them, it's actually unclear if it is legal to use them this way. So while there are tons of people working on mods and conversions, noone in their right mind would distribute all the source assets.
Much in the same way, no sane company will touch the legal nightmare of releasing LLM training data scraped from public websites. Even releasing the LLM alone might be infringement, there are literally court cases being fought over this right now.
Games like that, or the open-source clones of commercial games that require original assets to play (e.g. OpenXCOM), actually give a very clear analogy here: open source does not mean open assets. The software code is under a separate license from the data it processes. Emulators like Dolphin are kind of in this situation too - the program is open, the data it processes is not.
And that's fine! It's still valuable to have access to the source code, even if the "batteries" aren't included. Of course, if you really want to call it an open source model you should include the source for the data scraping/cleaning stages too; then the only thing missing would be the compute time and risk of acquiring dubiously-legal inputs.
I personally prefer a taxonomy like:
* Open weights: you can download the artifact and run it locally, not just use it through an application like chatgpt or an API.
* Open source: the code that created the artifact is provided in the same format that the authors used to work on it.
* Open data: the dataset that the source code was used on is available for download.
All three of those could be individually licensed or released, for 8 possible combinations. In the analogy to games, they would correspond to the licenses on the retail binary, the source code of the game, and the original uncompressed art assets or Blender projects, respectively.
If it has already been established that open source doesn't mean open assets, why would we change that now? After all, training data is literally nothing but assets - except that you don't need them to run the application. So in that sense open LLMs are more open than these games.
I agree that open source doesn't mean open assets, but neither does open assets mean open source. You could make a linguistic argument that the training data is part of the "source" of the model (as in, from whence it came), but in any case the point is moot because neither the training data nor the code is open.
Because marketing (open source is a buzzword after all), and the media just repeats what they read in press releases verbatim. But most people working with the models themselves call them open-weight, except for some occasional exception like OLMo that publishes the dataset and training scripts and is actually open source.
Allowing a parrot to iterate on given examples and generate a similar one with the information baked in their weights does not invalidate "Stochastic Parrot" take. On the contrary, it proves it.
LLMs are statistical machines. The catch is you feed it hundreds of terabytes of valid information, so it asymptotically generates valid information as a result of this statistical bias.
Even yet, they can hallucinate so badly. I mean, the same OpenAI model claimed that I'm a footballer, a goal keeper in fact.
It's clearly true that the LLMS are 'stochastic parrots', but for all we know that might be the key to intelligence. It is in itself not a deep observation any more than calling your fellow humans 'microbial meatbags'.
Saying that LLMs are stochastic machines does not establish an upper bound for success.
The thing is, this assumption of LLMs might be intelligent lies in the assumption is intelligence is enabled solely by the brain.
However, as the science improves, we understand more and more that brain is just part of a much bigger network, and its size or surface roughness might not be the only thing determines the level of intelligence.
Also, all living things have processes which allows constant input from their surroundings and they also have closed feedback loops which constantly change and tweak things. Call these hormones, emotions or self-reflection or whatnot.
We the scientists love to play god with the information we have at hand, yet we constantly humbled by the nature by experiencing the shallowness of what we know. Because of that I, as a CS Ph.D., am not so keen on to jump to that bandwagon which claims that we invented silicon brains.
They are arguably useful automatons built on dubious data obtained in ethical gray areas. We're just starting to see what we did, and we have a long way to go.
So, a living parrot might be more intelligent than these stochastic parrots. I'll stay on the cautious critics wagon for now.
We are not stochastic parrots. Old components of our brain help “ground” our thoughts and allow things like doubt or a gut feeling to develop which means we can question ourselves in ways an LLM cannot.
That's a shame. I suppose you'll need 4 of them with RDMA to run a 671B, but somehow that seems better to me than trying to run it on DDR4 RAM like the OP is saying. I have a system with 230G of DDR4 RAM, and running even small models on it is atrociously slow.
This is awesome. Pebble is still the best so many years later! I also have the Garmin, but Garmin thinks UX design is a waste of money or something, its horrendous. Pebble is the absolute undisputed King here.
reply