✨ dateline: 21 NOV 2023. At OpenAI, an opaque decision by a rump board-of-directors to fire the CEO has led to chaos within the company flooding the news cycle. In this type of situation, it is advised to wait at least six weeks before commenting as an outsider on the specifics of the situation.
💡 in a situation where there are so many different things that have gone wrong, discussing a few of the more tangential causes does not fall under the EMBARGO requirements.
⚙️ so EMBARGO is a word where I would like to use 汉字 rather than Latin characters. but I have so consistently gotten negative feedback about that usage pattern I am not doing so for now. also, I don't actually know how to say EMBARGO in 汉字; the word means more 神奇的沉默 than 不能交易 here.
Read More ...
The first rule of corporate governance is that the concept of "governance" is a myth.
⚙️ as always, the word "myth" does not imply that it is "false".
Companies are run by people. And no amount of words on paper can change that. Modern history is rife with examples of people deliberately misinterpreting those words for their own benefit.
In that line, OpenAI's many attempts to make sure that "nobody controls the technology" have only resulted in other people controlling the technology.
💡 or, more accurately, other people thinking they control the technology. Authority does not always reside where people think it does. If 90% of the staff can go to a different company, along with all their intellectual property, the OpenAI board didn't actually control anything.
OpenAI has been very insistent about not allowing anyone true equity. Instead, they are offering "profit participation units". As a matter of law, these must be viewed as "employee profit sharing" and not equity. Both because of the cap on possible returns, and because of the lack of shareholder voting rights.
💭 I tried an unusual corporate structure myself, in a failed "Big Solar" venture last year. It failed for a dozen different reasons, but one of them was the cap table structure. I allocated myself a sizable "founder's equity" block, with a restriction that the shares would void completely if the rate-of-increase of the Mauna Loa CO2 concentration graph (currently at about 2.25 ppm/year) did not decrease substantially over the next five years.There were two problems with this solution. First, it was needlessly complicated; every conversation where it arose turned out badly. But this can be attributed to the 新 nature of the solution; people always dislike innovations when they are new and unknown.More important was that the "tether me to the mast" aspect didn't actually work. Other investors would have an direct financial benefit to sabotaging the company's mission. And a pure profit motive would only minimally impact my decisions.
But my approach did have one advantage over OpenAI's: it was abundantly clear who controlled the company. Control of OpenAI is extremely opaque.
💡 and, more bluntly than the above hints: OpenAI is controlled by a cult. A different cult than the one that controls 99% of American businesses.
🌎 yes, "single-minded focus on generating profits for shareholders (and executives)" is a cult. But when a cult becomes prevalent enough, it becomes impolite to describe it that way.