5 ms·

I think a lot about how there's a very plausible alternate history where Elon Musk controls most of the frontier of AI.

by tptacek 6mo ago

sanderjd 6mo ago
There but for the grace of god go we...
dragontamer 6mo ago
You speak as if Elon Musk didn't buy tons of AI chips for full self driving (Dojo) and COMPLETELY flub it. It's the same as always. Musk himself is an awful business man. He relies upon buying the success of others and taking over. Outside of that, he's kind of awful. Initiatives started by Musk himself almost inevitably fail.
- fastball 6mo ago
  Tesla wasn't "successful" before he "took it over" (read: invested most of their seed capital and ran the actual day-to-day operations). SpaceX, founded entirely by him alone, is also the most valuable space technology company on earth, so...
  - dragontamer 6mo ago
    Well sure. But that's not AI chips or FSD. The subject of this whole topic. Within this realm of AI, Musk has constantly invested and failed. Dojo, xAI, Grok. His newest idea is leveraging SpaceX money to put data centers in Space. Good luck with that.
    - fastball 6mo ago
      Your previous comment was clearly not constraining itself to the realm of AI chips / FSD. Regardless: How do you measure Dojo, xAI, and Grok being failures? Dojo is training models for FSD, which is in operation today. The chips in Tesla cars are also taped in-house / vertically integrated, a lot of which is presumably shared investment with the chips needed for the Dojo training side specifically. Separating xAI from Grok (in a list like this) is kinda weird, but seems that Grok is actually a very capable LLM, even if it is not best-in-practice. Even if not beating out the top 3 labs, it certainly seems to be in the top 5 by most metrics. The xAI Colossus data centers are real AI training/inference infrastructure that were stood up in record time, and they are now selling capacity to other LLM providers. Is that a failure?
      - dragontamer 6mo ago
        https://finance.yahoo.com/news/tesla-shuts-down-dojo-ai-221950747.html https://finance.yahoo.com/news/tesla-shuts-down-dojo-ai-2219... Dojo is shutdown dude. Literally doing nothing but rotting. After spending $Billions on a custom chip, Tesla is now just buying NVidia chips and wasting their Dojo efforts entirely Team dismantled. Operations closed. Game over. --------- A similar event is playing out with xAI. Operations are transferring over to SpaceX. Out of money, out of time. They're going to try to grab more money (possibly illegally) across Musks enterprises but there's some severe legal questions on the legality of these moves. It's already sketchy as all hell that xAI bought Tesla's GPU fleet. Now SpaceX is buying it up under doubious circumstances. Whatever is going on with xAI / Twitter is now clear. It's not possible for it to stand on its own two feet and needs external investors to continue to survive.
        fastball 6mo ago
        Ok, but Dojo clearly has a spiritual successor in the form of: whatever Tesla is currently using to train FSD models? Like... Tesla kills teams all the time. Sounds like many were absorbed into the rest of the org. Seems like a bit of a nothing-burger without the follow-up that Tesla is shutting down FSD ambitions in general (or similar).
        dragontamer 6mo ago
        Yeah. It's called NVidia GPUs. IE: Tesla spent over a $Billion building a team and custom building Dojo chips only to fire the entire team, throw away their chips and buy NVidia.
        fastball 6mo ago
        Are they replacing their onboard chips with NVidia chips? AFAICT Tesla has increased investment in their own chip ambitions. Dojo specifically being cut is a massive failure if you are motivated to believe that, but otherwise it looks more like a subset of custom chip work that didn't pay off and the company has appropriately cut their losses. The onboard "AI2/3/4/5" chips are clearly the bigger investment (they can be found on every Tesla vehicle) and do not seem to be going anywhere at the moment.
  - cedws 6mo ago
    If you look at initiatives rather than companies most of what Musk has done has been a failure. Boring Company, Hyperloop, Roadster, Semi, full self driving, robotaxis, solar roofs, Optimus, mission around the moon, men on Mars. The list goes on. Tesla is an incredibly overvalued company rapidly losing ground to China that survives by Musk’s image of being a futurist, and by announcing fake products that never see the light of day. SpaceX has been headed by Gwynne Shotwell since the early days and usually when Musk inserted himself it’s been a mess.
- pizzafeelsright 6mo ago
  10 years returns S&P 500 (index of all those better than Musk): 261% Tesla: 2700% Disclaimer: My portfolio is 65% Tesla.
  - Calavar 6mo ago
    GME also beat the S&P 500 over the past 10 years. Is this evidence that Ryan Cohen is a business genius? Tesla has been a meme stock for about five years now, maybe more. Its valuation correlates with Musk's abilities as a showman and media figure, not a businessman.
    - filoleg 6mo ago
      > GME also beat the S&P 500 over the past 10 years. Is this evidence that Ryan Cohen is a business genius? GME did not beat the S&P500 over the past 10 years, and it is just the evidence of you needing to verify your claims before making them. Over the past 5 years[0]: S&P500 up by 77%, GME down by 50%. Over the past 10 years: S&P500 up by 260%, GME up by 207%. GME performance in the past 10 years doesn't indicate that Ryan Cohen is a business genius. It indicates that he runs a company that has been underperforming the market for at least the past decade. 0. https://www.google.com/finance/beta/quote/.INX:INDEXSP?window=5Y&keymoments=false&comparison=GME%3ANYSE&type=line https://www.google.com/finance/beta/quote/.INX:INDEXSP?windo...
      - Calavar 6mo ago
        I did look up numbers before I made that claim: From Yahoo Finance GME Jan 1, 2016: $7.09, $5.49 adjusted (accounting for dividend disbursements) GME Jan 1, 2026: $20.09 266% or 365% return depending on how you count dividends. 365% for GME vs. 306% for S&P 500 over the same period (also using adjusted for dividend numbers).
        lotsofpulp 6mo ago
        For the previous 10 years, dividends reinvested, GME returned 14.59% per year, SP500 did 15.376% per year. Considering the much, much higher risk of owning a single stock, the risk adjusted return of SP500 is much higher than GME. https://dqydj.com/stock-return-calculator/ https://dqydj.com/stock-return-calculator/ https://dqydj.com/sp-500-return-calculator/ https://dqydj.com/sp-500-return-calculator/
- saulapremium 6mo ago
  I absolute abhor the man but I think this is silly. No doubt that he has had luck and help, but he is still a good businessman. He's certainly an asshole (almost certainly dark triad territory), but I think that can be a benefit when creating a business.
  - dragontamer 6mo ago
    It isn't helping him in this fight vs Sam Altman, AI investments or the like.
Aurornis 6mo ago
I've thought about that, too, but it would require that all of the key individuals at OpenAI would have been willing to stay at OpenAI under his ownership. That seems unlikely to me given how divisive he is. OpenAI already had one existential leadership crisis without Musk. I doubt it would have fair better under his notoriously difficult leadership. If he had wrestled control away, I would expect an exodus of employees going to new companies.
- Hikikomori 6mo ago
  They're willing to work under the current snake, even got him back after he was removed, so why not musk?
  - saynay 6mo ago
    Haven't they hemorrhaged a lot of the founding talent in the years since? Now it is full of ex-Meta ad-tech people trying to find a way to make it actually turn a profit.
  - serial_dev 6mo ago
    They are both snakes (IMO), but they are different kinds of snakes. I can see why to some Sam is acceptable while Elon wouldn’t be (and vice versa). It’s also convenient to be more lenient toward the snake’s shortcomings when you becoming extremely rich depends on it.
HardwareLust 6mo ago
And how much worse things would be if that had come to pass?
geek_at 6mo ago
why would he run Anthropic?
- tptacek 6mo ago
  As I understand it, Anthropic exists in part as a quirk of how Dario Amodei's experience at OpenAI went. In the world where Musk controls OpenAI, I don't think you can assume Anthropic splinters off the same way (for overdetermined reasons! I'm not saying Musk is a better manager.)
  - keeda 6mo ago
    Hmm, my understanding is that Dario split over insufficient focus on safety at OpenAI (which, admittedly, could just be the PR-friendly reason) -- something that Elon is also clearly not big on. Safety not only seems to be an anti-priority at xAI given what we've seen come out of it, even at OpenAI Elon seemed not that concerned with safety, leading to that entertaining incident with a "golden jackass" trophy: https://xcancel.com/abc7newsbayarea/status/2054698419382354319 https://xcancel.com/abc7newsbayarea/status/20546984193823543... Almost certainly Dario would still have split from under Elon too, but also very likely that Elon would have immediately thrown a barrage of IP / NDA / trade secret / non-compete violation lawsuits to cripple Anthropic and keep it from reaching the frontier.
    - tptacek 6mo ago
      It's not my argument here that Musk is necessarily more aligned with Amodei.
paulpauper 6mo ago
This is what he Grok hopes to become, but probably too late
- DANmode 6mo ago
  Fifth place is fine, when the numbers and power in question are that large.
HardCodedBias 6mo ago
The following dates for the start of the statue of limitations were the most plausible: 1. 2019 capped-profit restructuring + 1B MSFT investment 2. 2023 Microsoft expansion / reported 75%-then-49% economics 3. 2024/2025 PBC restructuring AFAIK it has not been reported as to exactly what the jury found, but IIUC the 2019 date is consistent with their findings. That's poor for Musk, but it makes sense. He was arguing 2023. I think it is a valid argument. But he had to know that 2019 was very much in play (and is likely the most logically consistent). This is very squishy law.

I think a lot about how there's a very plausible alternate history where Elon Musk controls most of the frontier of AI.

by tptacek 6mo ago

I think a lot about how there's a very plausible alternate history where Elon Musk controls most of the frontier of AI.

sanderjd 6mo ago

There but for the grace of god go we...

dragontamer 6mo ago

You speak as if Elon Musk didn't buy tons of AI chips for full self driving (Dojo) and COMPLETELY flub it. It's the same as always. Musk himself is an awful business man. He relies upon buying the success of others and taking over. Outside of that, he's kind of awful. Initiatives started by Musk himself almost inevitably fail.

fastball 6mo ago
Tesla wasn't "successful" before he "took it over" (read: invested most of their seed capital and ran the actual day-to-day operations). SpaceX, founded entirely by him alone, is also the most valuable space technology company on earth, so...
- dragontamer 6mo ago
  Well sure. But that's not AI chips or FSD. The subject of this whole topic. Within this realm of AI, Musk has constantly invested and failed. Dojo, xAI, Grok. His newest idea is leveraging SpaceX money to put data centers in Space. Good luck with that.
  - fastball 6mo ago
    Your previous comment was clearly not constraining itself to the realm of AI chips / FSD. Regardless: How do you measure Dojo, xAI, and Grok being failures? Dojo is training models for FSD, which is in operation today. The chips in Tesla cars are also taped in-house / vertically integrated, a lot of which is presumably shared investment with the chips needed for the Dojo training side specifically. Separating xAI from Grok (in a list like this) is kinda weird, but seems that Grok is actually a very capable LLM, even if it is not best-in-practice. Even if not beating out the top 3 labs, it certainly seems to be in the top 5 by most metrics. The xAI Colossus data centers are real AI training/inference infrastructure that were stood up in record time, and they are now selling capacity to other LLM providers. Is that a failure?
    - dragontamer 6mo ago
      https://finance.yahoo.com/news/tesla-shuts-down-dojo-ai-221950747.html https://finance.yahoo.com/news/tesla-shuts-down-dojo-ai-2219... Dojo is shutdown dude. Literally doing nothing but rotting. After spending $Billions on a custom chip, Tesla is now just buying NVidia chips and wasting their Dojo efforts entirely Team dismantled. Operations closed. Game over. --------- A similar event is playing out with xAI. Operations are transferring over to SpaceX. Out of money, out of time. They're going to try to grab more money (possibly illegally) across Musks enterprises but there's some severe legal questions on the legality of these moves. It's already sketchy as all hell that xAI bought Tesla's GPU fleet. Now SpaceX is buying it up under doubious circumstances. Whatever is going on with xAI / Twitter is now clear. It's not possible for it to stand on its own two feet and needs external investors to continue to survive.
      - fastball 6mo ago
        Ok, but Dojo clearly has a spiritual successor in the form of: whatever Tesla is currently using to train FSD models? Like... Tesla kills teams all the time. Sounds like many were absorbed into the rest of the org. Seems like a bit of a nothing-burger without the follow-up that Tesla is shutting down FSD ambitions in general (or similar).
        dragontamer 6mo ago
        Yeah. It's called NVidia GPUs. IE: Tesla spent over a $Billion building a team and custom building Dojo chips only to fire the entire team, throw away their chips and buy NVidia.
        fastball 6mo ago
        Are they replacing their onboard chips with NVidia chips? AFAICT Tesla has increased investment in their own chip ambitions. Dojo specifically being cut is a massive failure if you are motivated to believe that, but otherwise it looks more like a subset of custom chip work that didn't pay off and the company has appropriately cut their losses. The onboard "AI2/3/4/5" chips are clearly the bigger investment (they can be found on every Tesla vehicle) and do not seem to be going anywhere at the moment.
- cedws 6mo ago
  If you look at initiatives rather than companies most of what Musk has done has been a failure. Boring Company, Hyperloop, Roadster, Semi, full self driving, robotaxis, solar roofs, Optimus, mission around the moon, men on Mars. The list goes on. Tesla is an incredibly overvalued company rapidly losing ground to China that survives by Musk’s image of being a futurist, and by announcing fake products that never see the light of day. SpaceX has been headed by Gwynne Shotwell since the early days and usually when Musk inserted himself it’s been a mess.
pizzafeelsright 6mo ago
10 years returns S&P 500 (index of all those better than Musk): 261% Tesla: 2700% Disclaimer: My portfolio is 65% Tesla.
- Calavar 6mo ago
  GME also beat the S&P 500 over the past 10 years. Is this evidence that Ryan Cohen is a business genius? Tesla has been a meme stock for about five years now, maybe more. Its valuation correlates with Musk's abilities as a showman and media figure, not a businessman.
  - filoleg 6mo ago
    > GME also beat the S&P 500 over the past 10 years. Is this evidence that Ryan Cohen is a business genius? GME did not beat the S&P500 over the past 10 years, and it is just the evidence of you needing to verify your claims before making them. Over the past 5 years[0]: S&P500 up by 77%, GME down by 50%. Over the past 10 years: S&P500 up by 260%, GME up by 207%. GME performance in the past 10 years doesn't indicate that Ryan Cohen is a business genius. It indicates that he runs a company that has been underperforming the market for at least the past decade. 0. https://www.google.com/finance/beta/quote/.INX:INDEXSP?window=5Y&keymoments=false&comparison=GME%3ANYSE&type=line https://www.google.com/finance/beta/quote/.INX:INDEXSP?windo...
    - Calavar 6mo ago
      I did look up numbers before I made that claim: From Yahoo Finance GME Jan 1, 2016: $7.09, $5.49 adjusted (accounting for dividend disbursements) GME Jan 1, 2026: $20.09 266% or 365% return depending on how you count dividends. 365% for GME vs. 306% for S&P 500 over the same period (also using adjusted for dividend numbers).
      - lotsofpulp 6mo ago
        For the previous 10 years, dividends reinvested, GME returned 14.59% per year, SP500 did 15.376% per year. Considering the much, much higher risk of owning a single stock, the risk adjusted return of SP500 is much higher than GME. https://dqydj.com/stock-return-calculator/ https://dqydj.com/stock-return-calculator/ https://dqydj.com/sp-500-return-calculator/ https://dqydj.com/sp-500-return-calculator/
saulapremium 6mo ago
I absolute abhor the man but I think this is silly. No doubt that he has had luck and help, but he is still a good businessman. He's certainly an asshole (almost certainly dark triad territory), but I think that can be a benefit when creating a business.
- dragontamer 6mo ago
  It isn't helping him in this fight vs Sam Altman, AI investments or the like.

Aurornis 6mo ago

I've thought about that, too, but it would require that all of the key individuals at OpenAI would have been willing to stay at OpenAI under his ownership. That seems unlikely to me given how divisive he is. OpenAI already had one existential leadership crisis without Musk. I doubt it would have fair better under his notoriously difficult leadership. If he had wrestled control away, I would expect an exodus of employees going to new companies.

Hikikomori 6mo ago
They're willing to work under the current snake, even got him back after he was removed, so why not musk?
- saynay 6mo ago
  Haven't they hemorrhaged a lot of the founding talent in the years since? Now it is full of ex-Meta ad-tech people trying to find a way to make it actually turn a profit.
- serial_dev 6mo ago
  They are both snakes (IMO), but they are different kinds of snakes. I can see why to some Sam is acceptable while Elon wouldn’t be (and vice versa). It’s also convenient to be more lenient toward the snake’s shortcomings when you becoming extremely rich depends on it.

HardwareLust 6mo ago

And how much worse things would be if that had come to pass?

geek_at 6mo ago

why would he run Anthropic?

tptacek 6mo ago
As I understand it, Anthropic exists in part as a quirk of how Dario Amodei's experience at OpenAI went. In the world where Musk controls OpenAI, I don't think you can assume Anthropic splinters off the same way (for overdetermined reasons! I'm not saying Musk is a better manager.)
- keeda 6mo ago
  Hmm, my understanding is that Dario split over insufficient focus on safety at OpenAI (which, admittedly, could just be the PR-friendly reason) -- something that Elon is also clearly not big on. Safety not only seems to be an anti-priority at xAI given what we've seen come out of it, even at OpenAI Elon seemed not that concerned with safety, leading to that entertaining incident with a "golden jackass" trophy: https://xcancel.com/abc7newsbayarea/status/2054698419382354319 https://xcancel.com/abc7newsbayarea/status/20546984193823543... Almost certainly Dario would still have split from under Elon too, but also very likely that Elon would have immediately thrown a barrage of IP / NDA / trade secret / non-compete violation lawsuits to cripple Anthropic and keep it from reaching the frontier.
  - tptacek 6mo ago
    It's not my argument here that Musk is necessarily more aligned with Amodei.

paulpauper 6mo ago

This is what he Grok hopes to become, but probably too late

DANmode 6mo ago
Fifth place is fine, when the numbers and power in question are that large.

HardCodedBias 6mo ago

The following dates for the start of the statue of limitations were the most plausible: 1. 2019 capped-profit restructuring + 1B MSFT investment 2. 2023 Microsoft expansion / reported 75%-then-49% economics 3. 2024/2025 PBC restructuring AFAIK it has not been reported as to exactly what the jury found, but IIUC the 2019 date is consistent with their findings. That's poor for Musk, but it makes sense. He was arguing 2023. I think it is a valid argument. But he had to know that 2019 was very much in play (and is likely the most logically consistent). This is very squishy law.