Tuesday, January 17, 2023

CHaTGPT and Whole Hospital Joint Ventures, Part I

Here are a few paragraphs from an interesting Axios article entitled,  How a Silicon Valley Nonprofit Became Worth Billions:

There are various different ways to make hundreds of millions of dollars, but historically "starting a nonprofit" has not been one of them. Silicon Valley, however, has managed to find a way, at ChatGPT creator OpenAI.

Why it matters: OpenAI pivoted from nonprofit to for-profit status in 2019, a mere four years after it was founded with $1 billion of donations from Elon Musk and others. It's now reportedly in talks to raise $10 billion from Microsoft, much of which is likely to go straight into shareholders' pockets.

The bottom line: OpenAI is a very valuable for-profit company, expertly using hype cycles to maximize the value of its shares.

It has lost the moral high ground, however — and everybody who who funded it while it was a nonprofit has the right to feel a bit aggrieved that they were unwittingly providing seed capital to make Sam Altman and his colleagues even richer than they already were.

In other words, a nonprofit dedicated to advancing the science of artificial intelligence had an initial valuation (as determined by donors) of at least $1 billion.  It is now projecting "net revenues" in the trillions (though that may be an exaggeration, who knows), thanks really to public funded (via tax exemption and deductible charitable contributions) and owned technology.  The nonprofit has converted to a for-profit -- which means net revenues have evolved into profits and donors into investors. The donors -- now investors -- stand to get rich.  Is this right or wrong? This is Part I of a two part post.  Part II comes tomorrow.  

Let's go back to the future to figure this out.  Remember, during the heyday of "managed care," how nonprofit hospitals felt negative economic impacts from insurance companies.  Insurance companies shifted from a fee for service model to a capitation fee model.  Under the former model, health care providers were compensated for each service or treatment ordered up for the patient.  Obviously, a fee for service model allowed for over-consumption of health services on demand.  The fact that health care provider compensation was based on the number of services provided, that they made more money if they ordered more tests or treatments, and the fear of malpractice liability for missing something meant that health care services would be over-prescribed.  There were no limits on health care demand because doctors and patients were essentially permitted to treat healthcare like an "all you can eat" buffet.  Meanwhile, unchecked demand hiked prices and a significant minority were unable to pay for healthcare.  A capitation fee model, where a healthcare providers get only a flat fee per patient (rather than per service), injects market discipline by putting health care providers at risk of loss for over prescribing.  Insurance companies capped what they would pay on a per patient basis, so health care providers might lose money by ordering all sorts of tests or treatments.  Capitation pays one fee per patient no matter how few or many treatments are provided. 

Nonprofit hospitals -- including medical school hospitals --  felt the pinch most of all because they depended on patient revenues (over and above donations) to "cross subsidize" very expensive research or experimental treatments.  Those hospitals might have even charged their non-indigent patients higher fees than those patients might have paid at a for-profit hospital.  Generally speaking, for profits don't have the imperative to get more money to fund expensive research or experimental treatment. Anyway, all of this lead to an increase in formal partnerships between nonprofit health care and for-profit physician practice groups, for example.  The idea boiled down to giving a for-profit partner (the physician practice group) a cut of the nonprofit hospital's revenue, thereby incentivizing that group to use the hospital services.  By doing so, nonprofit hospitals hoped to increase patient volume and revenue sufficient to fund research and expensive experimental procedures.  EO Heads exploded everywhere and eventually the Service issued rulings on so called "whole-hospital" and "ancillary" joint ventures, both of which imposed requirements intending to ensure that in a partnership between a nonprofit hospital and a for-profit entity (either a for-profit hospital or a physician practice group, typically), the nonprofit would continue operating for public rather than private benefit (by subordinating any profit incentive to the charitable purpose), and that literally allowing a for-profit partner to pocket a variable share of the revenue generated by the nonprofit the nonprofit was not engaging in private inurement or excess benefit.  Note the green lines in the picture below.  It shows that the nonprofit hospital is sharing profits with a physician group that has invested in the exempt hospital's operations.  It was this literal economic reality of private inurement -- the for-profit, in exchange for an investment just like a normal shareholder, was taking a share of subsidized profits, just like a normal shareholder -- that lead the Service to coin the phrase "private inurement, per se," discussed at footnote 178 of this article.  

The fretting and gnashing of teeth has pretty much died down since the Service issued its joint venture rulings, perhaps reflecting the changes in insurance reimbursement and hospital patient care practices since the enactment of Obamacare.  I think pretty much everyone felt like few other nonprofit endeavors generated the vast amount of revenue (and the temptation to line private persons' pockets) that nonprofit hospitals do, so there hasn't been much attention paid to the issue since the early 2000's.  

But the arc of the nonprofit universe is long and apparently bends towards . . . well, astronomical wealth.  OpenAI is not a health care organization.  It is a scientific research corporation whose purpose is to develop open source Artificial General Intelligence (AGI).  It was founded when Elon Musk and other rich people "contributed" over $1 billion to support OpenAI's research towards publicly available AGI.  OpenAI is getting a whole lot of attention these days because it has apparently developed AGI known as CHaTGPT, which is touted as artificial intelligence technology that can think (and learn) like a human who never forgets anything and, on top of that, can store and retrieve unlimited information.  Theoretically, a student, or even a professor for that matter, can order up a paper or dissertation on Rawlsian philosophy and whether tax exemption is consistent with that philosophy, for example.  CHaTGPT can spit out the product in short order and that is one of the relatively benevolent uses.  More heads are exploding now!  Open AI is now worth upwards of $30 billion.  And in an explicit move to harness investor capital, OpenAI has entered into a "capped profit" partnership with an entity known as OpenAI Ltd.   OpenAI is the general partner, and private investors are the limited partners.  Google is one of the limited partners, having put up another $1 Billion in investor capital.  In addition to a profit share, Google will be entitled to exclusive use of the technology developed by the partnership.  Open AI calls the partnership a "capped profit entity" because investors are explicitly limited to returns no greater than 100 times their capital contribution.  Net revenues -- or profits -- beyond investor return (the founders are claiming profits will be in the trillions) are devoted to the nonprofit general partner's charitable mission.  One billion times 100 equals $100 billion, by the way.  

So, let's summarize what happened here:

1.  A nonprofit forms to develop remarkable and potentially very valuable open source technology.  We all help pay for their work via tax exemption and charitable contribution deductions so it makes sense that the technology should be open source.

2.  The nonprofit lands a huge charitable donation, generating tax deductions for the donors.  Ok, so far so good, this is how its supposed to work.

3.  The nonprofit successfully develops incredible, tax subsidized technology generating tremendous investor interest.  Fine.

4.  The nonprofit drops the technology into a subsidiary entity (a limited partnership), and sells shares in that entity to private investors who, like any other private investor, expect and are contractually entitled to profits from the exploitation of the technology.  Wait, what?

5.  And by the way, in exchange for Google's investment, Open AI promises Google an exclusive license to the incredible technology.   Who said you could do this?!

In other words, Open AI has entered into a "whole hospital joint venture" and as part of the deal has promised a major investor exclusive access to its tax subsidized intellectual property.  This post is already pretty long, so in my next post I will provide brief analysis and opinion as to whether this arrangement ought to be carefully examined by the Service.

darryll jones


Current Affairs | Permalink


Post a comment