So, a guy with probably one of the largest incentives in the world to trash xAI is trashing xAI?
The information content of this is rather minimal. Even if everything he says is literally true it's hard to tell through the massive, massive vested interest he has.
And it doesn't help that...
'Hoffman, who is an investor in both Anthropic and OpenAI, pushed back firmly on the narrative that the two companies are in a zero-sum race. “We tend to want to tell these stories as cage matches,” he said, as in two companies enter and only one leaves, but “in fact,” he claimed, “there’s a lot of room for both of them to win incredibly.'
that's clearly a very self-interested gloss on the flip side of the situation. Yes, that's in the possibility space. No, I would not consider "both companies do fantastically for many many years" as a terribly large part of the possibility space. Look to all of the many past instances of industries starting up. It is a very common case that if you take the two early leaders you aren't looking at who is going to be the two biggest companies in 10 or 20 years. It is in fact a common case that neither of those companies are the leaders in 10 or 20 years. The sheer staggering size of the AI training moat at the current time may lock in the possibility that no other business could possibly overtake them... but what if somebody solves that massive training gap? It probably isn't mathematically fundamental; I can't help but observe that humans do not get to their level of capability by pouring the entire Internet through their head several times.
He probably does know a lot of things most of us don't know, but I doubt he's sharing very many of them in this article. This is just trash talk.
The AI labs have somewhat the same problem that publishers have. Essentially their asset is a static piece of IP: a huge file full of numbers. They’re like a publisher only letting you read their book through some scuzzy Flash reader UI because they have to protect that file at all costs. At some point the weights get out/get reproduced and then what they have a is bunch of sunk costs.
I think the theory is that it’s not purely static - you need to keep training and tuning the model (even just for general knowledge upkeep with current architecture) and so the infra/data is a contributory moat. Exfiltrating weights would get you a depreciating asset (plus we have all the lovely legal and regulatory frameworks to further protect them, which IS more like publishing)
Is xAI being used by any professionals? I see them acting as a data center rental service for the others, but that doesn’t justify their valuation imo. They seem to be behind on everything and don’t seem to have any relevance. The cursor purchase may change that but for how long?
I see no reason any portion of SpaceX justifies its current valuations. I also think taking what is probably ultimately a successful and profitable company in SpaceX, even if it is maybe not as successful and profitable as Elon might say, and tying it at the hip to the AI bubble, while being clearly on the losing end of that AI bubble so far, could well kill SpaceX in the process. I hope not, because no matter how HN may feel about Elon, SpaceX has some great tech and is definitely moving space tech forward. Rather harder to say that about xAI.
But then I'd say I don't understand the valuations of a lot of companies right now. It seems to me the stock market has written into its structure the idea that United States companies will be claiming something like 500-1000% of all TAMs in the entire world in the next 10-20 years, which seems unlikely to be the case. SpaceX's claimed TAM of "pretty much the entire United States GDP, you know, why not" is merely the most blatant instance of this.
I'm not defending SpaceX or xAI. Billionaires don't need my help. But this article is still pretty pointless. Hoffman isn't a dispassionate observer, he's one of the players. Of course he's telling everyone he's going to win and the other guy is going to lose. Even when a coach is completely objectively correct when he says in his pre-game press conference that he has every confidence that his team will win in the end, it's still an information-free statement.
There is also a political aspect to this. Reid Hoffman is a big democrat donor and hater of Trump, going so far as to funding a law change so Trump could be privately prosecuted for sex offences in New York.
Which is ironic, because Reid was a friend of Epstein and visited his island.
>So, a guy with probably one of the largest incentives in the world to trash xAI is trashing xAI?
He's just stating the obvious, so I really don't see this as contentious.
xAI is irrelevant. It's so irrelevant that after being relegated hardware from Tesla, then pushed into Twitter to try to make that have value, then pushed into SpaceX because Elon Musk somehow gets away with hilarious levels of securities fraud, now it's basically reduced to renting out hardware.
Yes, xAI is irrelevant, and Hoffman is just pointing out the blatantly obvious. Its only value is in renting out hardware that can be better used by more capable orgs. It is basically a scalper that happened to get loads of nvidia hardware pre-orders in just before the AI run-up, and the entire SPCX scam relies upon everyone trying to buy usage of it.
XAI is very relevant. They have one of the best video models, they own cursor which has a good percent of the coding market. Anthropic runs thier ai on xai’s data centers.
In porn and deepfakes. Yeah, they should be regulated to obliteration.
>they own cursor which has a good percent of the coding market
They don't own cursor. They announced an acquisition immediately after the insane SPCX valuation to desperately scrabble to lock in some of that laughably nonsensical hype valuation (as the entire US equities market has become firmly detached from reality, and at some point is going to catastrophically crash to reality). It doesn't close for months. And FWIW Cursor is rapidly declining, and anyone still foolishly on it should probably find an offramp now. Before Elon Musk starts fiddling with it and you find your code focused on the genocides of whites in South Africa or something.
I mean, the fact that you had to cite that as their credibility fully demonstrates how completely worthless it is. GPUs that were originally Tesla's (before the whole robo-taxi scam fell apart), then shuffled to Twitter, then to SPCX.
>Anthropic runs thier ai on xai’s data centers
That was literally the foundation of my comment. xAI is so worthless that they get better value become another vanilla rent-a-GPU operation.
xAI is a joke. Somehow Elon's pathetic Matryoshka doll routine keeps suckering fools.
I truthfully don’t know the answer, but if I had to guess, his connections and positions provide him with an unusual amount of knowledge and perspective. Another might be that his opinions often are correct in hindsight.
The information content of this is rather minimal. Even if everything he says is literally true it's hard to tell through the massive, massive vested interest he has.
And it doesn't help that...
'Hoffman, who is an investor in both Anthropic and OpenAI, pushed back firmly on the narrative that the two companies are in a zero-sum race. “We tend to want to tell these stories as cage matches,” he said, as in two companies enter and only one leaves, but “in fact,” he claimed, “there’s a lot of room for both of them to win incredibly.'
that's clearly a very self-interested gloss on the flip side of the situation. Yes, that's in the possibility space. No, I would not consider "both companies do fantastically for many many years" as a terribly large part of the possibility space. Look to all of the many past instances of industries starting up. It is a very common case that if you take the two early leaders you aren't looking at who is going to be the two biggest companies in 10 or 20 years. It is in fact a common case that neither of those companies are the leaders in 10 or 20 years. The sheer staggering size of the AI training moat at the current time may lock in the possibility that no other business could possibly overtake them... but what if somebody solves that massive training gap? It probably isn't mathematically fundamental; I can't help but observe that humans do not get to their level of capability by pouring the entire Internet through their head several times.
He probably does know a lot of things most of us don't know, but I doubt he's sharing very many of them in this article. This is just trash talk.