OpenAI: The 800 pound gorilla

OpenAI 3d logo

OpenAI, under Sam Altman’s leadership, is responsible for the creation and upgrading of:

a) ChatGPT, which is, to date, the most viral software ever, having achieved 1,000,000 (one million) registered users in the space of 5 days (and now, 100 million users in the space of 60 days), and has been the darling of the media (and pranksters) since its release in November 2022.

b) DALL-E, which pretty much issued the clarion call that was the official started of the massive “AI Art” (technically, “generative AI imagery”) movement.

c) Bing AI Chat, which actually challenged the utter monopoly of Google in search for the first time in 20 years.

And those are only the beginning.

$13 billion gets you…

a great deal, in fact. Microsoft, in perhaps the smartest investment bet ever, invested a cumulative $13 billion in the company …Does this seem huge only to me? I was raising funds in Silicon Valley in 1999, and $10 million was the standard ante, and a funding round (even a B or C round) north of $200 million was simply unheard of, even for Googles and Facebooks… now, 50x that… $13 billion (cash + cloud credits)okey dokey…

Yes, $13B. For 49% of the company. Not even, actually. (it’s complicated). but nonetheless, if OpenAI is the first to achieve AGI, or even just give Google a run for its money in search (which at this point, is a given), then… then YES, it’s worth it.

In fact, it’s a bargain.

Because, in the visionary words of Sam Harris:

“This… this race to build the first AGI…
this is a Winner Take All game they’re playing.”

A few more points on the unique ingredients that form OpenAI:

  • OpenAI has an insanely bold, audacious, and novel business plan
  • Along with that bizplan comes a never-before seen Corporation Structure
  • It is highly possible that his company, OpenAI — not any nation or government — will be the gifter of a genuine global UBI (Universal Basic Income) program.

Let’s drill down into those points:

OpenAI’s one-line Business Plan

Altman has been asked repeatedly, given the insane level of funding and cash reserves, how he plans to turn a profit, eventually. And to his credit, from the get (2015), his answer has not changed. We play back one of the many Q&As here:

Interviewer:

“So, Sam, you have
$10 billion in the bank.
Near zero revenue.
How do you actually plan
to make money?”

Altman:

Our plan is very simple:
At OpenAI, We are devoted to one singular goal: the creation of AGI: an Artificial General Intelligence that is smarter and faster than any human across every domain. And once we have created that — and we will create that — we will just tell it:

‘AGI, make a plan to maximize wealth.
..and execute that plan, now.’

And it will be so.

woah.

Update 2023/11: Since the launch of ChatGPT, GPT4, and the OpenAI Enterprise API, it appears that OpenAI is on track to record close to a billion dollars in revenue for FY 2023. NOT BAD! Probably still significantly less than operating / R&D costs, but nonetheless, that’s a respectable figure for a start-up / non-profit research lab coming from essentially zero in revenue 12 months prior.

OpenAI: A Novel Corporate Structure

OpenAI, technically, is a limited-capitalization (“profit-capped”) corporation that is joined at the hip (and governed by) a separate non-profit entity. That non-profit entity controls the board of directors (i.e. totally controls) the for-profit entity, and even has the right to cancel all debt & equity obligations in case of “emergency.” The foreseen emergency, of course, is a mis-aligned AGI that attempts to take over the world. No, this isn’t science fiction.

This is real people, real AI, a real company, real money, and real reality. Today.

So, in the case that a crazed AI goes megalomaniacal and attempts to commandeer human energy resources to further its goals, Sam Altman (or more realistically, the board majority) can just say “Sorry, Microsoft, we’re going into a hard shutdown. Your money just went poof!”

Not a joke. Excerpted directly from the OpenAI Investment Prospectus:

safety vs speed - warning to investors in OpenAI

the Path to UBI

Why? See the business plan, above. Once the AGI starts making money, its game over. With AGI at the helm, unhindered by human politics or speed limits, $10B can grow to $100B to $1T to $10 trillion relatively quickly. That would be, to put it in perspective, approximately 1/2 of the entire U.S. GDP. <gasp?>

But might be even more interesting is that, in the spirit of Gates, OpenAI has a plan for global wealth redistribution. And not just a plan, but a contract: In the heart of OpenAI’s charter is a plan for true global UBI, or Universal Basic Income. Essentially, after three conditions are met:

  1. Investors are made whole (principal + “capped” 100x return)
  2. Founders / Employees are compensated, and
  3. Company is generating profit and has substantial cash reserves in place

After these conditions are met, then all the remaining & future profit “will be returned to the Nonprofit for the benefit of humanity.” What does this mean, practically speaking? It means OpenAI will attempt to fund a global UBI program. In other words, the wealth that the AGI generates will be automatically redistributed across the poorest 33% of the world’s humans.

This has its own complications, to be sure, and merits a much longer conversation about pros and cons. So we’ll just say for now: For a for-profit capitalist company to have this type of grand philanthropic vision baked into its very charter, is, like so much else about OpenAI… simply without precedent.

ChatGPT… true viral growth

It’s been pretty well covered in the mainstream media about ChatGPT’s (OpenAI’s first chatbot AI with a friendly interface) rise to web app supremacy. The fact that, within a mere 5 days of the launch (November 22, 2022), they had acquired 1 million registered human users.

We present the graph, because even though its known, it is worthy of repeat visualization (shorter bars are “better”):

Sam Altman's CHatGPT growth to 1 million users

 

What’s slightly less known
is that a mere 90 days later,
that one million
had transformed into
100 million users:

OpenAI ChatGPT speeds to 30 million DAU daily average users

Again, no ads, no sponsorships, clearly no marketing (let’s name it ChatGPT?!?!? really?!), no mega-stadium naming rights — just 100% pure, organic word of mouth.

It’s pretty important to note that during this whole time: OpenAI’s consumer marketing & advertising budget is essentially zero. ChatGPT was released more as a research test than as an actual product. And even today, OpenAI has less than 500 approximately 770 people on its payroll (11/2023).

OpenAI: Raiding Valley Talent

While we’re on the topic, just curious: where did the first hundred or so members of the OpenAI All-Star AI Dev Dream Team come from?

Your answer:

Sam Altman's hires: from whence did they come?


OpenAI: Founding Team / Origin Story

OpenAI — a non-profit from the get — claims no less than 11 individuals with the title of co-founder. The mythology goes that it was a conversation between Altman and Musk in mid-2015 that got the ball rolling, because they wanted a global, democratized AI that wasn’t controlled by Google or any other of the BigTech Mafia (Apple, Amazon, Meta, Microsoft, etc). What Sweet Irony.

With that, the team was assembled in Dec 2015, and the rest is history.

Original OpenAI co-founders (2015) are:

In 2015, a superteam of AI researchers, under the guidance of Altman, Brockman, and Musk, came together in a nonprofit in order to unabashedly “build an AGI, untroubled by capitalist greed, for the benefit of all humanity.” Those 11 members of the co-founding team were:

  1. Sam Altman,
  2. Elon Musk,
  3. Ilya Sutskever,
  4. Greg Brockman,
  5. Trevor Blackwell,
  6. Vicki Cheung,
  7. Andrej Karpathy,
  8. Durk Kingma,
  9. Jessica Livingston,
  10. John Schulman,
  11. Pamela Vagata, and
  12. Wojciech Zaremba

OpenAI Funding Timeline

OpenAI raised roughly $120 million in charitable donations at around the time of its formation in 2015. As the need for scaling and compute expanded dramatically, it became clear that the company would need huge capital reserves in order to pay for servers and electrical needs to train the large LLMs.

Thus Altman formed a for-profit subsidiary, still 100% controlled by the non-profit Board of Directors (this came to bite him, hard), which in short order received $13 billion (13,000 million) in investment from Microsoft. Traditional VCs also joined into the equity structure, mostly by purchasing shares directly from employees at ever increasing valuations.

As of Nov 2023, OpenAI had raised a total of close to $12 billion in funding, and had achieved a book valuation of $33B, putting it in the realm of the most highly valued private companies on planet earth (behind #1. TikTok @ $255B, and #2. Space-X @ $175B)

DATE INVESTORS AMOUNT VALUATION
2015
Dec
Company Formation :
a donation-based non-profit
Elon Musk $100 million donation $0.5 B
Sam Altman $20 million donation
   
2018 Musk exits due to conflict of interest with Tesla, who is pivoting into SotA AI / dojo
2019 Altman comes in as CEO, immediately forms a for-profit subsidiary to facilitate “traditional” capitalist / strategic / VC  investment
2019
Jul
Microsoft $1B (1,000 million) investment $10 B
2021 Microsoft $2B $20 B
2021 Sequoia
Andreesen-Horowitz
Tiger Global
Bedrock
cash purchase from employees *
     
2023
Jan
Microsoft $10B (10,000 million),
bringing total invest to $13B and equity stake to 49%
$29 B
2023
Apr
Sequoia
Andreesen-Horowitz
Tiger Global
SVA
Founders Fund
+3 additional
$300 million $33 B
2023
Aug
Arrowshare Ventures ?? ??
2023
Nov
Thrive Capital
(pending)
$1 billion
cash purchase from employees *
$88 B

* “CASH PURCHASE FROM EMPLOYEES” / i.e. “SECONDARY MARKET SALE”. These “investments” (2021 and 2023) do not require the issuance of new stock (Series A, B, C etc) by the company, nor alterations to the basic cap table. Rather, they allow traditional VCs to purchase shares directly from employees. This produces premature “liquidity events” for employees and is a major part of the (already extremely generous) compensation packages that OpenAI lavishes on its staff.

In other words, employees take what, in any other company would be “stock options waiting for an IPO or acquisition” and allows them to transform those into cold hard cash well prior to any corporate liquidity event. This can also be seen as a major reason why 95% of the staff mutinied upon the firing of Sam Altman: by removing Altman, the board put Thrive’s planned $1B cash infusion (a billion dollars divided amongst less than 1,000 employees… you do the math) in severe jeopardy. So while they might very well have “loved Sam,” they were almost certainly also operating from a place of self-interest, whether consciously or not.

Which brings us to the question: who exactly was this semi-suicidal board who elected to decapitate the company leadership at such a promising moment in its history?

OpenAI Board of Directors

The Board, initially, was populated by a number of high-power Silicon Valley stars, as one might expect. Over time, for various reasons, it decayed to a 3/3 composition: 3 co-founders (Sam, Greg & Ilya), and 3 outsiders (Adam, Helen, and Tasha). It was this weakened position that made it possible for the tragedy of 11/17 to actually take place:

Tasha McCauley Adam d'Angelo Helen Toner OpenAI Board of Directors

OpenAI executive team / key players (2023) were:

On November 16, 2023, just prior to the implosion of the company, these were some (not all!) of the more important & publicly visible execs that drove the strategy, research & product vision at OpenAI:

  1. Sam Altman, CEO
  2. Greg Brockman, President / Chairman
  3. Ilya Sutskever, Chief Scientist (Superintelligence / Safety)
  4. Mira Murati, CTOthe first 3 mutineers N17:
  5. Jakub Pachocki, [CMU] Director of Research — LinkedIn
  6. Szymon Sidor, [MIT] Research Scientist — LinkedIn
  7. Aleksander Madry, [MIT] Head of Preparedness — LinkedIn
    .
  8. Brad Lightcap, COO@bradlightcap
  9. Jason Kwon, CSO@jasonkwon
  10. Jan Leike, Lead, Superalignment@janleike
  11. Wojciech Zaremba, co-founder — @woj_zaremba
  12. Boris Power, Head of Applied Engineering — @BorisMPower

How a non-profit raises $12 billion

In 2019, Altman formed a for-profit subsidiary, because the non-profit was struggling to get past an initial $130 million in charitable donations, and Altman and his team realised that in order to scale as hard as they needed to with GPUs and training runs, they would need substantially greater capital reserves.

Thus the “capped profit” concept was born, with the non-profit board still fully in control (this came to bite Altman in 2023, to be sure). But now Altman was able to raise funds in the traditional Valley / VC manner, and raise funds he did, spectacularly. Almost immediately he had closed a clean billion dollar round led by Microsoft.

After the launch of ChatGPT, Microsoft doubled down, and added another $12 billion to the kitty. A large portion of that would be paid back to Microsoft as OpenAI’s exclusive datacenter / supercomputing partner, but no matter; OpenAI had to spend that anyways, may as well be with a partner.

An inglorious ending to a high-flying star

As of noon on Monday November 20 2023, more than 90% of OpenAI’s 770 employees had signed this letter, effecting a total mutiny on the great ship OpenAI:

openAI mutiny Nov 19 2023

OpenAI R.I.P. : EOTWAWKI

openAI RIP 2015-2023

It was nice knowing you. Thanks for all the Fish.