Ilya has always seemed like he was idealistic and I’m guessing that he was the reason for OpenAI’s very strange structure. Ilya is the man when it comes to AI so people put up with his foolishness. Adam D'Angelo is, like Ilya, an amazing computer science talent who may have shared Ilya’s idealistic notions (in particular OpenAI is non-profit, unless forced to be capped profit and is categorically not in the business of making money or selling itself to MSFT or any entity). “Helen” and “Tasha” are comically out of their depth and are loony toons, and simply decided at some time ago to follow Ilya.
Sam got the call from MSFT to sell, MSFT really ponied up (300B ?). The inference costs for OpenAI are/were staggering and they needed to sell (or get a large influx of capital which was in the works). This ran counter to Ilya’s idealistic notions. Sam attempted to negotiate with Ilya and the loony toons, a vote was called and they lost, hard.
I think this tracks with all the data we have.
There are a couple of other scenarios that track given OpenAI’s comically poor board composition, but I think the one above is the most plausible.
If this did happen then OpenAI is in for a hard future. Imagine you worked at OpenAI and you just found out that your shares could have been worth a tremendous amount and now their future is, at best, uncertain. There will be some true believers who won;t care but many (most?) will be appalled.
Let this be a lesson, don’t have a wacky ownership structure and wacky board when you have the (perhaps) the most valuable product in the world.
It's not "wacky" to have goals other than the accumulation of capital. In fact, given the purpose of OpenAI, I think it's meritorious.
I'd personally prefer we just not work on AGI at all, but I'd rather a non-profit dedicated to safe AI do it than a for-profit company dedicated to returns for shareholders.
> Let this be a lesson, don’t have a wacky ownership structure and wacky board when you have the (perhaps) the most valuable product in the world.
I think the lesson is just the opposite: If you want to work according to your ideals, and not simply for money, you should absolutely do whatever 'wacky' thing protects that.
> I'd personally prefer we just not work on AGI at all, but I'd rather a non-profit dedicated to safe AI do it than a for-profit company dedicated to returns for shareholders.
You seem to be under the impression that OpenAI is a nonprofit. For the most part, it's not: it was founded as a non-profit, but it subsequently restructured into a for-profit company with the nonprofit owned under the same umbrella company. This is indeed an unusual corporate structure.
That's likely what OP is referring to as "wacky".
Wikipedia says the for-profit part is owned by the non-profit, not under the same umbrella company.
Mozilla Foundation/Corporation does this too IIRC. It's what allows them to to pursue serious revenue streams with the for-profit part, while still steering their mission with the non-profit in charge, as long as they keep a separation in some kinds of revenue terms.
EDIT after 56 minutes: Hell, even IKEA does this type of ownership structure. So it's quite cool, but probably not all that "wacky" as far as enterprises that want to be socially responsible go.
> Serious revenue streams like having Google for a patron yes? I feel like the context is important here because […]
For that specific example, Mozilla did also go with Yahoo for as-good revenue for a couple of years IIRC, and they are also able to (at least try to) branch out with their VPN, Pocket, etc. The Google situation is more a product of simply existing as an Internet-dependent company in the modern age, combined with some bad business decisions by the Mozilla Corpo, that would have been the case regardless of their ownership structure.
> Which is great and possible in theory, but […] is ultimately only sustainable because of a patron who doesn't share in exemplifying that same idealism.
The for-profit-owned-by-nonprofit model works, but as with most things it tends to work better if you're in a market that isn't dominated by a small handful of monopolies which actively punish prosocial behaviour:
https://en.wikipedia.org/wiki/Stichting_IKEA_Foundation
https://foundation.mozilla.org/en/what-we-fund/
> people are trying to defend OpenAI's structure as somehow well considered and definitely not naively idealistic.
Ultimately I'm not sure what the point you're trying to argue is.
The structure's obviously not perfect, but the most probable alternatives are to either (1) have a single for-profit that just straight-up doesn't care about anything other than greed, or (2) have a single non-profit that has to rely entirely on donations without any serious commercial power, both of which would obviously be worse scenarios.
They're still beholden to market forces like everybody else, but a couple hundred million dollars in charity every year, plus a couple billion-dollar companies that at least try to do the right thing within the limits of their power, is obviously still better than not.