> this is fundamentally about: stopping or fundamentally changing OpenAI's plans to be spun off.
If OpenAI actually values OpenAI-NP at $40B then they will surely sell it for $97B. The point is the $40B is obviously a dramatic undervaluing and the fact they won't sell it for $97B shows that. This is the wrench.
"We believe an OpenAI-P owned by Elon is fundamentally against our mission."
While I'm not saying Sam's being an angel here, I do think it's reasonable to say that llms aren't an existential risk and spinning them off to a consumer thing while xrisk and super alignment stays in a nonprofit is plausibly mission aligned, and certainly moreso than giving any amount of control to the guy who's proven his devotion to care and safety in progress by checks notes taking a sledgehammer to the US Treasury.
But OpenAI-P of course wouldn't be owned by Elon, he would just own $40B worth of OpenAI-P according to their own valuation of it. A tiny minority stake of the $157B enterprise.
OpenAI-NP, could just go back to other OpenAI investors and buy back their "$40B" stake plus have $57B leftover to apply to AI safety. Silly Elon, overpaying by almost 2.5X for something of such little value!
It's a hard point that they would likely have to make in court! Harder still with Elon having originally been involved in founding OpenAI-NP.
Everyone hates Elon reasonably but few people seem to care the Sam Altman is conducting a theft on the order of $57+B.
If OpenAI-NP has a reasonable case to believe that Elon will be a bad steward of the technology and work against their charter (https://openai.com/charter/), they shouldn't sell to him, even if his offer is more than another. Unlike a normal board, the duty-to-shareholder isn't the only duty, a hostile buyout like with twitter doesn't necessarily work.
> But OpenAI-P of course wouldn't be owned by Elon, he would just own $40B worth of OpenAI-P according to their own valuation of it. A tiny minority stake of the $157B enterprise.
IDK, Musk's offer is apparently for the nonprofit, not for a partial interest in the not-yet-extant for-profit subsidiary.
If the OpenAI-NP board sees such existential risk from a change in leadership of OpenAI-P, how can they in good faith eschew control of it? By removing themselves from control, they would see the exact same risk anyhow. Elon could make a tender offer for the private for-profit entity that its board would be bound to accept in its fiduciary duty to shareholders. Plus, they would have to make the case that both Elon is specifically bad AND that he's so bad that it's not worth having $57B more to deploy for their AI goals. What's more, you would have to consider whether or not preventing Elon from controlling OpenAI even prevents whatever risk you see from Elon as he is already building xAI in parallel.
> IDK, Musk's offer is apparently for the nonprofit, not for a partial interest in the not-yet-extant for-profit subsidiary.
You can't buy a non-profit. It's their assets he's trying to buy. If OpenAI-NP values their OpenAI-P claims at $40B then that's what Elon would own according to their own beliefs.
I think you added this after I replied. Even if this is the case, better to do so with $97B and no control in OpenAI than $40B and no control in OpenAI!
OpenAI-NP can either have a $40B from spinning out OpenAI-P or $97B in cash. In theory, they could literally just go and buy back their stake in OpenAI-P and have $57B left over for other AI endeavors if the valuation of it at $40B were remotely reasonable! If I'm on the board and I believe the $40B number, I'd do this in a heatbeat.
It wasn't long ago that the company almost fell apart because senior folks who actually care about the company tried to oust sama - the chaos that an Elon takeover would introduce is a critical risk that cannot be ignored.
"I'll pay you double" isn't the instant checkmate you're suggesting.
If the existing assets are worth $40b then for $97b as a board you could sell it, build another OpenAI (in fact you could build 2 new competitors) and still have change left over to pursue your mission. If your charter is to "produce lots of OpenAI accessible to everyone" this has to fit that mission - more AI competitors, more access to AI, etc. No matter what your feelings are of the characters in this story... by definition given their charter they have $40billion worth of assets (by their valuation) to achieve their mission - which includes the PV of all internal advantages/models/patents/etc they have developed. By accepting this deal they are ahead.
So:
- Either the assets are worth $40 billion and to further their charter they should take the deal (and by definition have more assets/firepower to do their mission) OR - $40 billion is significantly undervalued which then raises other questions around governance, internal dealings, etc to try to convert to a for-profit by buying the non-profit's assets. If they go with the internal plan with the other deal on the table it could be argued as self-sabotage. I'm not sure of US law, but it could give rise to the validity of the deal legally. A non-profit converting to a profit is already a bit "hazy" from my understanding.
In the scenario where they chose the $40B in OpenAI equity, they are eschewing $97B with which they can achieve their mission. They will be less effective by $57B. They still have no control over OpenAI in this context. They don't have control over OpenAI from the perspective of safety/benefit of humanity. All they have is $40B worth of equity.
Or, they can get $97B and go about completing their mission. In this case, what happens to OpenAI doesn't really matter. They lose control over OpenAI from the perspective of safety/public benefit in both cases, but in this case they get to deploy $97B of capital to their goals. The success of OpenAI-P is irrelevant to OpenAI-NP.
Now, maybe you would claim that there is some kind of existential risk from OpenAI-P being controlled by Elon. But if that's the case, how is Elon uniquely special? How does OpenAI prevent Elon from gaining control of OpenAI anyway when they won't have control to prevent it? How can they be sure Sam Altman is okay to lead when the board already removed him due his duplicity once. To eliminate their control OpenAI-P, they open themselves up to this same risk regardless. They can take that risk with either $40B of equity or $97B of cash.
> "I'll pay you double" isn't the instant checkmate you're suggesting.
This isn't what I suggest at all and I would've expected someone who sees through such complex issues to rationales that are "obvious to everyone" would've been able to see that. This does however pose real problems for OpenAI-P as they try to go private!
XAI is valued at another $40bn as well.
