It's highly likely in my uneducated opinion that OpenAI will be told to adopt a standard corporate structure in the near term. They will likely have to pay out a number of stakeholders as part of a "make right" setup.
It's not unusual for nonprofits to have spinoffs, but it is unusual for the nonprofit to be so consumed by its for-profit spinoffs.
They are still able to actually make a profit (and quite often will, because careful balancing of perfect profit and loss is almost impossible and loss is bad), and I thought those profits were still taxed because otherwise that's too obvious as a tax dodge, it's just that profit isn't their main goal?
For a good point of comparison, until 2015, when public scrutiny led them to decide to change it, the NFL operated as a nonprofit, with the teams operating as for-profits. Other sports leagues continue to have that structure.
Socialism seems to create a lot of markets for the Capitalist private sector.
The hype and the credulity of the general public play right into this scam. People will more or less believe anything Sam the Money Gushing Messiah says because the neat demos keep flowing. The question is what's we've lost in all this, which no-one really thinks about.
That would be a Nuke in the AI world.
https://www.publicsource.org/why-is-the-nfl-a-nonprofit/
The total revenue of the NFL has been steadily increasing over the years, with a significant drop in 2020 due to the impact of the COVID-19 pandemic12. Here are some figures:
2001: $4 billion
2010: $8.35 billion
2019: $15 billion
2020: $12.2 billion
2021: $17.19 billion
2022: $18 billionThe reality: we don't even get public LLM models, let alone source code, while their coffers overfloweth.
Awesome for OpenAI and their employees! Every else goes without. Public benefit my arse.
Want to open a bakery in your small town? Start it as a 501(3)(c) and promise it’s a charitable endeavor for the local community. Then invest your $500k into the bakery maybe even from your local community (it’s a tax deductible donation!) to get the bakery up and running.
Then once it’s turning a profit, ditch the original 501(3)c and replace it with a LLC, S-Corp or C-corp and start paying taxes. (And hope you don’t get sued or audited)
His point is mom and pop bakeries aren’t typically sophisticated enough to pull of schemes like this, even if it would save tens of thousands on taxes.
What activities couldn’t they do with their charity arm that required this for-profit arm?
That they make money incidentally to that is really no problem and a positive because it provides reasonable funding.
What if Firefox made a world beating browser by accident. Would they be justified in closing the source, restricting access and making people pay for it?
That's what OpenAI did.
They don't have moat. Their main advantage have been people and aleady we see entire Anthropic spinoff, Sutskever absent, Karpathy leave, who is next?
I wish they hadn't because they are thinking too commercial (extremely high paid CEO) for instance but they have a foundation to answer to which doesn't manage them like shareholders would (eg not rewarding the CEO for dropping marketshare!). This model is the worst of both worlds imo.
Every dollar of income generated through television rights fees, licensing agreements, sponsorships, ticket sales, and other means is earned by the 32 clubs and is taxable there. This will remain the case even when the league office and Management Council file returns as taxable entities, and the change in filing status will make no material difference to our business.
The parent comment is making a common mistake that non-profits can not make profits, that is false. Non-profits can't distribute their profits to their owners and they lack a profit motive, but they absolutely can and do make a profit.
This site points out common misconceptions about non-profits, and in fact the biggest misconception that it lists at the top is that non-profits can't make a profit:
https://www.councilofnonprofits.org/about-americas-nonprofit...
no longer a non-profit but no less hypocritical
What's next? Can the OpenAI nonprofit shell divest itself of the for-profit OpenAI and spend the remainder of its cash on "awareness" or other nonsense?
If OpenAI Co. is gatekeeping access to the fruits of OpenAI's labors, what good is OpenAI providing?
They could release the source with a licence that restricted commercial use, anything they wanted, that still allowed them to profit.
Instead we get "AI is too dangerous for anyone else to have." The whole thing doesn't inspire confidence.
There is a difference between positive cash flow and profit as profit has differences in accounting rules. If you invest in some asset (let's say a taxi car) today, all of that cash flow will happen today. But there will be no effect on the profit today, as your wealth is considered to have just changed form, from cash into an asset. For the purposes of profit/loss, the cost instead happens over the years as that asset depreciates. This is so that the depreciation of the asset can be compared to the income it is generating (wear and tear on car vs ride fare - gas).
That would mean that any publicly traded company that didn't issue a dividend didn't make a profit which no one believes.
Do you really want to claim that Google has never made any profit?
OpenAI had to start as a non profit because there was no clear path forward. It was research. Kind of like doing research with the goal of curing cancer.
The unexpected breakthroughs came a bit quicker than anticipated and everybody was seeing the dollar signs.
I believe OpenAIs intial intention at the beginning was benign. But they just couldn't let go of the dollars.
The Sherman Fairchild Foundation (which manages the post-humous funds of the guy who made Fairchild Semiconductor) pays its president $500k+ and chairman about the same. https://beta.candid.org/profile/6906786?keyword=Sherman+fair... (Click Form 990 and select a form)
I do love IRS Form 990 in this way. It sheds a lot of light into this.
OpenAI has always argued that the for-profit is furthering the aims of the non-profit.
Also employees can't get shares of the non-profit so of course they would from the for-profit arm.
IMO you could cut the CEOs salary from 6 million to 300k and get a new CEO - and we probably wouldnt see any difference in Firefox results. Perhaps improvement even. Since the poorly paid CEO would try to demonstrate value - and this best is done by bringing back firefox market share.
This should be changed to
“Then the US government fails to fund the billions of dollars required for medicinal trials needed to get FDA approval”
No one is stopping the US government from doing all the necessary work to verify the medicines work and put them in the public domain.
IANAL but I think the tax issue would likely hinge on how well that $500k was isolated from the for-profit side. If the non-profit has no substantial operations and is just a shell for the for-profit, I could see getting in trouble for trying to deduct that as a donation. But if there's an audit trail showing that the money is staying on the non-profit side, it would likely be fine.
Unprofitable businesses of every sort don't pay income taxes. Startups like OpenAI don't pay income taxes because they don't have income. And investors don't get a writeoff merely for investing in a nonprofit; it's not like a donation to a nonprofit (which would be deductable).
Academia produces tens of thousands of papers per year; many of these are garbage, p-hacking or low value - the rest are often contradictory, misleading, hard to interpret or just report a giant body of raw-ish data. It is a very valuable process - despite all the waste - but the result of this is too raw to be actionable.
This body of raw 'science' is the necessary substrate for biotechnology and drug development - it needs to be understood, processed, and conceptualised into a hypothesis (which most likely fail) strong enough to invest billions of dollars into.
Pharmaceutical industry is the market-based approach to prioritising investment into drug development (what is it, 100B$ p/y?) - and even a leftist who might want to debate in favour of a different economic model would have to agree that this job is hard, important, and needs to be done.
In the case of many/most (honest) non-profits, the operating costs are paid out of a combination of the dividends of an invested principal (endowment, having been previously donated by donors) and grants/current donations. Any operating profit could then be returned to the endowment, allowing the organization to maintain higher operating costs indefinitely, thus giving the organization more capacity to further their mission.
The median annual wage in 2021 in the US was $45,760,
https://usafacts.org/data/topics/economy/jobs-and-income/job...
Just to put bit of perspective...
Someone took inspiration from this.
There's this weird thing where charities are judged by how much they cost to run and pay their employees to even a greater degree than other organizations, and even by people who would resist that strategy for businesses. It's easy to imagine a good leader executing the mission way more than 500k better than a meh one, and even more dramatically so for 'overhead' in general (as though a nonprofit would consistently be doing their job better by cutting down staffing for vetting grants or improving shipping logistics or whatever).
Open source staying behind commercial products even if they are technically really close … ? I think I have already seen this.
"In conversations with recruiters we’ve heard from some candidates that OpenAI is communicating that they don’t expect to turn a profit until they reach their mission of Artificial General Intelligence" https://www.levels.fyi/blog/openai-compensation.html
Where is my $20/month for GPT-4 going then?
Put another way, a $1bn hedge fund is considered a small boutique that typically only employs a handful of people.
It feels like there should be a way to tax these startups that exist as vehicles for cash grabs, but are not profitable.
Most non-profit employees receive their compensation in the form of a salary. If you need to pay "market rate" competing with organizations that offer equity, you pay a bigger salary. When non-profits spin for-profits off (eg research spinoffs), they do it with a pretty strict wall between the non-profit and the for-profit. That is not the case for OpenAI.
If I make $100 in a year and spend $1000 that year, my income is ($900). How can I spend $1000? Generally through loans and bonds. How do I secure said loans? Generally simply by showing how much VC and income comes in with a business plan that banks accept.
But that's the secret to the money flow. That's also partially why the collapse of SVB was such a blow to the tech industry. A LOT of loans were issued by them.
"We" got a free-as-in-beer general knowledge chat system leagues better than anything at the time, suitable for most low-impact general knowledge and creative work (easily operable by non-technical users), a ridiculously cheap api for it, and the papers detailing how to replicate it.
The same SOTA with image generation, just hosted by Microsoft/Bing.
Like, not to defend OpenAI, but if the goal was improving the state of general AI, they've done a hell of a lot - much of which your average tech-literate person would not have believed was even possible. Not single-handedly, obviously, but they were major contributors to almost all of the current SOTA. The only thing they haven't done is release the weights, and I feel like everything else they've done has been lost in the discussion, here.
For a business, revenue minus expenses in a given accounting period is considered profit. The only question is whether it gets treated as corporate profit or personal income.
Dangerous and powerful things like weapons and chemicals are restricted in both physical and informational form for safety reasons. AI needs to be treated similarly.
Not at all. With GPT-3 they only released a paper roughly describing it but in no way it allowed replication (and obviously no source code, nor the actual NN model, with or without weights).
GPT-4 was even worse since they didn't even release a paper, just a "system card" that amounted to describing that its outputs were good.
Even I as a software engineer have a minimum salary I expect because I’m good at my job.
Just because it’s a non-profit doesn’t mean I’m going to demand a smaller salary.
And if the non-profit can’t afford me and gets a more junior dev and they’re not very good and their shit breaks… well, they should have paid full price.
That said, there ARE a lot of dirty non-profits that exist just to pay their executives.
Where can I go get or drink from my free as in beer chat system from them then?
I remember one org had so many money pipes going in/out of it that I had to modify my code to make a special case for them.
Even if it was for profit company and it paid out all the surplus earnings to shareholders (owning clubs), it would be taxed zero on zero earnings (they'd just have to ensure all payouts happen within the calendar year).
They have the same moat that Google search has. Including as it pertains to usage and data.
You also can't train a new competitor like OpenAI was able to jumpstart GPT, the gates have already been raised on some of the best data.
Very few companies will be able to afford to keep up with the hyper scale models that are in our future, due to the extreme cost involved. You won't be able to get enough high-end GPUs, you won't be able to get enough funding, and you won't have a global brand that end users recognize and or trust.
The moat expands as the requirements get ever larger to compete with them. Eventually the VC money dries up because nobody dares to risk vaporizing $5+ billion just to get in the ring with them. That happened in search (only Microsoft could afford to fund the red ink competition with Google), the exact same thing will happen here.
Google search produces $100+ billion in operating income per year. Venture capital to go after them all but dried up 15+ years ago. There have been very few serious attempts at it despite the profit, because of the cost vs risk (of failure) factor. A lot of people know how Google search works, there's a huge amount of VC money in the tech ecosystem, Google mints a huge amount of profit - and yet nobody will dare. The winner/s in GPT's field will enjoy the same benefit.
And no, the open source at home consumer models will not come even remotely close to keeping up. That'll be the latest Linux consumer desktop fantasy.
Their sell-out path was hundreds of millions of dollars from GOOG to make their search engine the default, and, unspoken: allow FF to become an ugly, insecure, red-headed stepchild when compared to Chrome.
Likely part of what took priority away from Thunderbird, at the time, too.
Non-profits weren't really as much of a thing until the neoliberal era of privatizing everything.
Of course, there are "real" non-profits, those kinds of activities are a real thing, such as organizing solely member funded organizations to serve the people, but in America, this is a marginal amount of the money in the system.
> In 2003 the Internal Revenue Service revoked VSP's tax exempt status citing exclusionary, members-only practices, and high compensation to executives.[3]
Or later in the article https://en.wikipedia.org/wiki/VSP_Vision_Care#Non-profit_sta...
> In 2005, a federal district judge in Sacramento, California found that VSP failed to prove that it was not organized for profit nor for the promotion of the greater social welfare, as is required of a 501(c)(4). Instead, the district court found, VSP operates much like a for-profit (with, for example, its executives getting bonuses tied to net income) and primarily for the benefit of its own member/subscribers, not for some greater social good and, thereafter, concluded it was not entitled to tax-exempt status under 501(c)(4).[16]
Concisely, in any human matteres: Do what you say you'll do, or, add qualifiers/don't say it.
Take funds from a subset of users who need support services or patch guarantees of some kind, use that to pay people to continue to maintain and improve the product.
Today we all benefit from OpenAI, but its the for-profit Open AI that made it possible. How else would they spend billions on compute and take those large risks, on whose money?
Nobody in the hedge fund world works for salary.
They work for bonuses. Which for $1bn fund should be another $20m or so (20% profit share of 10% returns), otherwise you suck.
If bonuses aren’t available in non-profits, the base salaries should be much higher.
If you literally mean people (as in employees, executives, ect), they already are being taxed on income.
Unprofitable businesses always have expenses for labor, materials, ect. The distinction is that the company and owners arent making money, so they dont pay taxes. Those that do make money naturally do pay taxes.
What is the hard part to square?
Training LLMs requires a lot of text, and, as a practical matter, essentially all LLMs have committed copyright infringement on an industrial scale to collect training data.
The US has a fair-use exception with a four-part test:
The second and third parts (nature of the work (creative) and how much of the work is used (all of it)) strongly favor copyright owners. The fourth part (which SCOTUS previous said is the most important part, but has since walked back) is neutral to slightly favoring the copiers: Most LLMs are trained to not simply regurgitate the input, so a colorable argument exists that an LLM has no impact on the market for, say, NY Times articles.
Taken together, parts 2 through 4 are leaning towards impermissible use. That leaves us with the first part: Could it make the difference? The first part really has two subparts: How and what are you using it for?
"How" they are using it is clearly transformational (it defeats the purpose of an LLM if it just regurgitates the input), so that argues in favor of copiers like OpenAI.
But where I think Altman had a brilliant/evil flash of genius is that the "what" test: OpenAI is officially a non-profit, dedicated to helping humanity: That means the usage is non-commercial. Being non-commercial doesn't automatically make the use fair use, but it might make the difference when considering parts 2 through 4, plus the transformativity of the usage.
Guess, what - you missed the loophole.
Take a look at Sarah Palin's Daughter's' charity foundation Against Teen Pregnacy - founded after she, herself, was impregnated as a teen and it was a scandal on Sarah Palin's political shenanigans.... (much like boabert - his Drug/Thievery ~~guild~~ Addiction Foundation, soon to follow)....
Sarah Palins daughter got pregnant as a team, caused shame on the campaign - and started a foundation to help "stop teen pregnancy"
Then when the 503 filed, it was revealed that the Daughter was being paid ~$450,000 a year plus expenses from "managing the foundation" for the donations they solicited.
---
If you dont know how "foundation" is the Secret Financial Handshake For "Yep, Ill launder money for you, and you launder money for me!... donate to my TAX DEDUCTABLE FOUNDATION/CHARITY... and Ill do the SAME to yours with the Money you "donated" to me! (excluding my fee of course)
This is literally what Foundations do.
(if you have never looked into the SEC filings for the Salvation Army (I have read some of their filings cover to cover.... biggest financial scam charity in the country, whos finances are available...)
money laundering is a game. Like Polo.
---
>>>The company remains governed by the nonprofit and its original charter today. "
https://i.imgur.com/I2K4XF5.png
-
NFL can achieve the same taxation level as a for-profit if it's more careful about distributing all surplus earnings before the end of the year.
Someone could certainly abuse the non-profit status there too, but nobody brought those cases up.
Current CEO earns 20 times more -> 6 million per year
tone is the one thing AI has yet to solve.
(plus intoning and atoning... AI has yet on these little Jungians)
A bold claim given their track record
Transitioning from “nice idea” to “consumer product” is a vast chasm. Most people that do not actually have experience taking things from research to production grossly under-estimate the amount of effort involved. From a purely economic perspective, the “research” part of the total bill is dwarfed by the activity required to turn it into a salable product.
Some of the most productive areas of US government biomedical research have not come from NIH but from DoD. Most people do not realize that virtually all modern trauma medicine was invented by US military research as a very active ongoing program. If you get in a bad automobile accident, most things that happen will be the product of US military research. But interestingly, these programs have very sparse research budgets, literally just single millions of dollars in many cases. But that is trauma medicine, not drug development.
Drug trials in particular are extremely expensive. People like to pretend these don’t exist. A few million in research costs doesn’t write off billions of dollars in development costs. There is no mathematical way to argue otherwise.
You are right. NSF backs this (https://ncses.nsf.gov/pubs/nsf23320). Businesses now fund ~80% of R&D, USG funds ~20%.
According to CBO pharma spends ~$90B on R&D (https://www.cbo.gov/publication/57126) so $30B I would not call trivial or a rounding area, but your points still stands that it is the minor share.
> A few million in research costs doesn’t write off billions of dollars in development costs. There is no mathematical way to argue otherwise.
There could be an important distinction between infra R&D and last mile R&D. The cost of developing a drug in our current system might be $3B today on average, but if you also had to replace all the infra R&D USG invested in over decades (GenBank, PubMed, and all the other databases from NCBI and the like) that these efforts depend on, it might be much higher. So I could still see an argument that the government pays for the research needed by all the drugs, then private sectors builds on that and pay for the last mile for each one.
However, I think you've put forward strong points against the argument "the research is done using public funds, and then privatized and commercialized later".
> Drug trials in particular are extremely expensive. People like to pretend these don’t exist.
I think in general people are frustrated because for all the money going into pharma people have not been getting healthier in the USA, in fact, in the median case, the opposite. So some big things are going wrong. I think you've shown that the problem is not that government is paying for high drug development costs and industry is coasting.
How do I form my own Church of Sam. Why would I pay taxes when people believe in me?
