Hacker News Reader

I do not, for a single second, believe that Musk is doing this for altruistic reasons.

it doesn't have to be luckily

I never said this was altruistic. The same goal can come from him trying to derail OpenAI-P.

Nobody is doing anything for altruistic reasons, including and especially Sam Altman.

Yeah sure, in the pandemic all people just did it for the money. Same with all the voluntary fire fighters and all the other voluntary work.

A lot of people did a lot of things for a lot of money during the pandemic. Even voluntary fire fighters do it for status, fun, companionship, extra money, experience, etc. Doesn't mean that volunteers aren't valued or important. It's important to note that there's limits to people's charity even among the most charitable. Those limits are driven by perceived costs outweighing the perceived benefits to the individual.

Could it be that part of the the Musk-led group's offer of $100 billion is a tactic to drive up the valuation of Open AI-P, so that it's more painful/difficult for Open AI-P to buy itself out of Open AI-NP?

That's why I said this:

> 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.

Interesting logic, but I think it's obvious to everyone that the reason they won't sell to Elon has nothing to do with the price of the offer.

I'm too smooth brained to see the obvious reason. Would love for someone with more folds to say what the reason is instead of hiding it behind its obviousness.

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.

Which number is bigger isn't the only calculation to make. Elon has shown himself to be a capricious and distracted steward of the companies he assumes control of. He has a lot of debt and is already over-leveraged on multiple companies as well as his new roles in government. X has plunged in value, Tesla sales are plummeting worldwide, Elon's scorched earth political agenda invites unending controversy that OpenAI already has enough of on its own. Even if you ignore those glaring concerns, there's a compelling argument to be made that OpenAI's current leadership has already proven itself more than capable of world-class innovation and vision when they effectively created an entirely new industry.

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.

For as "obvious to everyone" as your point is supposed to be, it focuses entirely on what the offer means to OpenAI-P and zero on what it means to OpenAI-NP. The decision is to be made by an OpenAI-NP that can effectively divorce itself from OpenAI-P and in either case will be divorced from control of OpenAI-P. They can either have $40B in OpenAI equity or $97B.

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!

> This isn't what I suggest at all

This you?

> They can either have $40B in OpenAI equity or $97B.

In other words: the big number trumps the small number - QED.

> and zero on what it means to OpenAI-NP

Why do you believe OpenAI-NP wouldn't care about all the issues I raised?

> This you?

> > They can either have $40B in OpenAI equity or $97B.

> In other words: the big number trumps the small number - QED.

Yes those are words I used to describe a factual choice they have. Following that are a lot more words that dive deeper into that discussion. Here's how I summarized my initial point in the TL;DR on my ultimate parent post:

"Musk is trying to force OpenAI to value OpenAI-NP's stake in OpenAI-P at a reasonable level which is far greater than the current $40B which will help Musk by potentially derailing OpenAI-P's goal to be spun off.

For people with smooth brains like me, I'll helpfully point out how this doesn't present Elon's offer as "an instant checkmate." The first word to note is "trying." This suggests it is an attempt to do something that is not guaranteed to result in a win (ie a "checkmate"). The second word is "potentially" which again reinforces the case that this does not guarantee Elon achieves the stated goal (ie is not "checkmate").

> Why do you believe OpenAI-NP wouldn't care about all the issues I raised?

My prior post raised all sorts of reasons and you seem to have decided to ignore them to build strawmen.

> Here's how I summarized my initial point in the TL;DR on my ultimate parent post:

Yes. I responded directly to that comment, acknowledged it as interesting, but stated that there are obvious reasons why this fact - that you are building your entire argument around - isn't the primary concern in this transaction. In exchange, you offered me feigned disbelief that all the insanity surrounding Musk isn't totally distorting the economic incentives at play.

> Following that are a lot more words that dive deeper into that discussion

Following that are a lot more words saying the same thing over and over again:

> ...either $40B of equity or $97B of cash.

97 bigger than 40

> in the scenario where they chose the $40B in OpenAI equity, they are eschewing $97B with which they can achieve their mission

if they choose 40 they are eschewing 97 (which is bigger than 40)

> but in this case they get to deploy $97B of capital to their goals.

In this case, rather than having 40 to put towards their goals, they have 97, 97 of course being higher than 40.

> My prior post raised all sorts of reasons and you seem to have decided to ignore them to build strawmen.

It's not a strawman, it's essentially the only thing you've had to say this entire thread. If an executive order is passed to make 40 greater than 97 your entire argument completely dissipates.

LMAO you literally just removed every word about the incentives around control of OpenAI-P and are claiming it's not a strawman to say that the only thing I say is 97 > 40.

It's not a strawman, it's the crux of your argument. The rest is just hand-waving away the chaos that Musk represents for OpenAI P and NP so you can repeatedly frame this is a simple question of which number is bigger than the other - hence why you have to repeat it over and over again.

> how is Elon uniquely special

This is the part that is obvious to everyone, but I already offered a few reasons why. If you can't see them, then I think we're at an impasse.

X plunging in value is outdated information. EBITDA is double what it was when it was public ($682m to $1.25bn). Revenue is down, but profit is up substantially, due to aggressive cost management.

XAI is valued at another $40bn as well.

Twitter’s interest payments are quite large though, something like $1.2bn/year.

The advertising business is rising again, with Disney, Comcast and IBM back. [1]

If revenue is rising, with the new cost structure it'll be a cash machine. The doomers who claimed Twitter would collapse were plainly wrong, there's been one minor outage with dramatically reduced infra and employees.

[1] https://www.adweek.com/media/advertisers-returning-to-x/

That may be all true. But maybe I'm not quite understanding the point or how it relates to Elon from the perspective of non-for-profit OpenAI. It could be "stranger X" and let me explain.

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.

Yeah, I think that's the critical observation here. Good job cutting through the noise and spotting it.

> 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.

> "We believe an OpenAI-P owned by Elon is fundamentally against our mission."

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.

The issue isn't OpenAI-NP having some stake (which will obviously be the case no matter what), it's Elon having some stake.

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 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

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.

> 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.

They can make the argument that this offer is not in good faith and that Elon would not make an equivalent offer on an open market. That is, Elon's purpose in this offer is to prevent the sale of OpenAI, not to actually acquire it. Perhaps one could get around this if the cash were put in escrow, but also Elon likely doesn't have 100B in cash and it appears part of the valuation that Elon is putting up is x.ai, so it's quite reasonable to say that x is overvalued in this deal and that it isn't really 90B anyway.

> 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.

From what I can gather you're just mistaken here, Elon seems to be offering 100B for the thing recently valued at 157B (per https://archive.is/59jZ5#selection-637.0-648.0), while what is being sold for 40B is not the entire $157B entity.

Either that, or elon really is getting a controlling interest, and then the argument about Elon being a terrible steward still applies.

> They can make the argument that this offer is not in good faith and that Elon would not make an equivalent offer on an open market.

Yes this an angle they could take. This would probably be a huge mistake. By doing this they would in part be conceding that it's a worthwhile offer to be made and Elon also is likely more than able to acquire the funds.

> From what I can gather you're just mistaken here,

I'm not. You and WaPo are. For one, it should be obvious that that's the case otherwise this offer would be worthless since it's less than the $157B most recent valuation. There would be no point in discussing any of this and everyone would be laughing at Elon. For another, here's the original reporting from the WSJ which is unsurprisingly much clearer: https://archive.is/stI0V

> 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

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!

Drive up or value it fairly? If Sam and the board are unwilling to even consider the $100B offer then it seems to be strong evidence that it is worth much more than $100B.

This is a well known unsolved problem in theoretical investment science.

I probably misunderstand you but why should it be fundamentally about that and not fundamentally about controlling more valuable technology? The example of Twitter clearly demonstrates that despite the Dear Leader's altruistic claims, he's after money and power just the same as anyone else.

I think some of it is just a grudge. From everything I've read about Musk he seems big on them. So while some of it may be for power, a large portion is probably plain old revenge.

He would probably also be happy to get OpenAI-NP for $97B! However, I don't think he believes this is likely at all. His goal is to force OpenAI's hand and stop the comical; undervaluing of OpenAI-NP by forcing OpenAI to value OpenAI-NP at >$97B.

I probably haven't explained this clearly enough in my original post which I wrote too quickly.

Reportedly [0], SoftBank is looking to invest $40B for a share of OpenAI.com that would value it at $260B.

Given that OpenAI.org currently wholly owns OpenAI.com (doesn't it? even Msft didn't get a stake?), I have trouble seeing (a) how the $40B valuation would stand up in court if anyone could challenge it (and I believe Musk is trying to do that?), but also (b) how Musk's $100B pushes up the valuation more than a SoftBank investment would do.

Scratching my head...

[0] https://www.nbcwashington.com/news/business/money-report/sof...

I thought about this for a bit and I'm not sure how you would even counter. A legitimate offer has been tendered. It's pretty much checkmate here. A certainly well played move. This will definitely derail things. For OpenAI to best move would be to actually unconditionally accept, as it's likely this offer isn't serious in the sense that it's the offer they'd prefer to give.

Great thread, thanks. This is kind of what I was thinking and makes the most sense. That this is a play to block OpenAI plans to go full for-profit.

OpenAI gonna prove P=NP is what you are alluding to?

I think Musk (probably using his Saudi admirers yet again) wants to buy OpenAI because he needs a working chatbot right now. For example, he is gutting the General Services Administration and wants to install a GSAi chatbot to replace the workers:

https://www.wired.com/story/doge-chatbot-ai-first-agenda/

Perhaps "big balls" and the other geniuses he hired haven't managed to come up with a working solution.

Musk needs chatbots for his technocracy movement because they are great for faking progress in efficiency without understanding any of the systems they replace and destroy.

He may also need them for online bot campaigns to convince MAGA voters that layoffs, deregulation, H1Bs and supporting billionaires are really in their best interest.

That said, considering he overpaid for Twitter, perhaps OpenAI should consider his offer.

> I think Musk (probably using his Saudi admirers yet again) wants to buy OpenAI because he needs a working chatbot right now.

But I think Musk does have a working chatbot right now? Haven't used it so I won't speak from experience, but Grok by xAI is working and is not that much behind OpenAI, maybe like half a year behind, judging by LLM Arena[0] and other benchmarks that I've seen.

[0] https://lmarena.ai

Putting aside that LLMs can’t replace people that way, it would be infinitely cheaper to just license one vs buying OpenAI. And you could say “but Elon wants government money”, but it’d take so long to make it back because inference is so cheap (and probably running at a loss if you’re OpenAI).

This makes no sense. Republicans work for the government too. How are they going to get re-elected if they fire their base?

Nonsense. He's ignoring judges orders to cease and desist in his funding freezes and he's simultaneously attempting to seize control of the AI company through an aggressive acquitision, all the while avoid seemingly all oversight.

This can only be described as techno authoritarianism.

> and he's simultaneously attempting to seize control of the AI company through an aggressive acquitision

Kind of, yes! If OpenAI actually believes that the true value of OpenAI-NP is $40B then surely they will sell it for $97B! But, they of course don't believe that the proper valuation of complete control over OpenAI-P is worth only $40B. That's the point of why this works.

you can add data, transportation and transhumanism to the list of things he'd love to do.

You go work at his company, living in his government, in his vehicle, using his data network, thinking thoughts through his mind link.

He's laid it out bare, quite honestly.