I know what—I have an even worse business plan!
-
@cstross My assumption would be that Nvidia asks for a giant bucket of money and eventually gives Musk an off-the-shelf chip with a “SPAAAACE" sticker on it. But I don't know. The kabuki theater could get deep on both sides.
@zarfeblong @cstross Or they tell him that technically all data centers are already in space.
-
@cstross Okay, let me see if I have a handle on this…
⁃ Musk wants to IPO SpaceX.
⁃ So he wants to merge SpaceX and his AI company, because SpaceX looks better if he can glue in the $N-billion investments that he's getting for AI stuff.
⁃ He shouts “AI datacenters in space!" This has no business or economic meaning; it's just stringing VC buzzwords together for a SpaceX/XAI merger.
⁃ Nvidia says “Sure, boss, we can do you an AI-in-space chip.”@zarfeblong @cstross also, disrupting the market for solar panels
-
@cstross Okay, let me see if I have a handle on this…
⁃ Musk wants to IPO SpaceX.
⁃ So he wants to merge SpaceX and his AI company, because SpaceX looks better if he can glue in the $N-billion investments that he's getting for AI stuff.
⁃ He shouts “AI datacenters in space!" This has no business or economic meaning; it's just stringing VC buzzwords together for a SpaceX/XAI merger.
⁃ Nvidia says “Sure, boss, we can do you an AI-in-space chip.”@zarfeblong I think it's the other way around. SpaceX IPO as-is would be fine, a lot of people want to buy it, they have actual tech, product and revenue. Meanwhile xAI is, even by AI company standards, a piece of shit burning through money and recently also losing quite a few top people with not that much to show for it.
Musk would still like his AI though, so the idea is to have people who want to buy SpaceX subsidize xAI. He is also astroturfing the numbers and pulling strings to get the SpaceX+xAI monstrosity into one of the big indexes so index funds would effectively prop up the price. See yesterday's piece on Alphaville: https://www.ft.com/content/59adbe42-ca30-47f3-9cda-5415945e9368 (registration probably required) @cstross
-
RE: https://infosec.exchange/@jerry/116242195809861609
I know what—I have an even worse business plan! Why don't we build all our new fission reactors IN SPAAACE? That way we'll be safe from the radiation and any meltdowns!
Also me: I am a teapot, short and stout: this is my handle and this is my spout …
/s
@cstross Next we know Big Orange will give up Greenland and lay claim to all possible orbital slots.
-
@cstross My assumption would be that Nvidia asks for a giant bucket of money and eventually gives Musk an off-the-shelf chip with a “SPAAAACE" sticker on it. But I don't know. The kabuki theater could get deep on both sides.
@zarfeblong @cstross Probably it'll have an optimised sleep instruction, given that's what it'll spend most of its time doing because of heat buildup
-
@zarfeblong I think it's the other way around. SpaceX IPO as-is would be fine, a lot of people want to buy it, they have actual tech, product and revenue. Meanwhile xAI is, even by AI company standards, a piece of shit burning through money and recently also losing quite a few top people with not that much to show for it.
Musk would still like his AI though, so the idea is to have people who want to buy SpaceX subsidize xAI. He is also astroturfing the numbers and pulling strings to get the SpaceX+xAI monstrosity into one of the big indexes so index funds would effectively prop up the price. See yesterday's piece on Alphaville: https://www.ft.com/content/59adbe42-ca30-47f3-9cda-5415945e9368 (registration probably required) @cstross
@redsakana @cstross Hm, I did not think of the other way around. Makes sense though.
I will keep an eye out for financial commentary.
-
@redsakana @cstross Hm, I did not think of the other way around. Makes sense though.
I will keep an eye out for financial commentary.
@zarfeblong @redsakana @cstross yes, SpaceX is a home for X AI (and X) because it's got revenue and demand already, so it can paper over some of the financial mess. But SpaceX is already way overvalued because launch demand from non-Starlink sources has not picked up, and Starlink while profitable is close to its maximum addressable market already (there are not a billion extra Starlink users to be found.)
I think Patrick Boyle did an OK job walking through it in this video: https://youtu.be/8rS3fTbC7TE
The combination of Space and AI keeps retail interest high to try to get into the indexes in an accelerated schedule before reality hits.
-
R relay@relay.infosec.exchange shared this topic
-
@cstross My assumption would be that Nvidia asks for a giant bucket of money and eventually gives Musk an off-the-shelf chip with a “SPAAAACE" sticker on it. But I don't know. The kabuki theater could get deep on both sides.
@zarfeblong @cstross Except that Musk wants to build his own chip fab (with blackjack and hookers (probably literally with blackjack and hookers)), so he’d have no need of Nvidia in his completely baked fantasy world.
-
@cstross I seem to remember an Asimov (?) story about something similar with ultra dangerous for power generation but even Asimov picked the back side of the moon not deep space.
It would be kind of poetic justice though if a space based Musk data centre got taken out by a stray starlink.
@etchedpixels @cstross I remember one about a bio research lab in space that's deemed too dangerous, so they send it into the sun, but because the researcher kills himself inside his experiment it starts leaking space-resistant spores just when it crosses earth's orbit...
-
@cstross Okay, let me see if I have a handle on this…
⁃ Musk wants to IPO SpaceX.
⁃ So he wants to merge SpaceX and his AI company, because SpaceX looks better if he can glue in the $N-billion investments that he's getting for AI stuff.
⁃ He shouts “AI datacenters in space!" This has no business or economic meaning; it's just stringing VC buzzwords together for a SpaceX/XAI merger.
⁃ Nvidia says “Sure, boss, we can do you an AI-in-space chip.”@zarfeblong So wait, AI in space chip. So that’s just a chip without the need for the probabilistic “temperature” factor used in LLMs because you’ll get that for free from cosmic radiation?
-
RE: https://infosec.exchange/@jerry/116242195809861609
I know what—I have an even worse business plan! Why don't we build all our new fission reactors IN SPAAACE? That way we'll be safe from the radiation and any meltdowns!
Also me: I am a teapot, short and stout: this is my handle and this is my spout …
/s
@cstross What about a giant fusion reactor that is so big WE orbit IT? You could use photovoltaic cells to capture visual radiation and generate power, without worrying about microwave transmission dangers.
Just a thought. -
@cstross What about a giant fusion reactor that is so big WE orbit IT? You could use photovoltaic cells to capture visual radiation and generate power, without worrying about microwave transmission dangers.
Just a thought.@Dtraslerwriting Psst: go back and read the whole thread from the start. (Before me.)
-
@Dtraslerwriting Psst: go back and read the whole thread from the start. (Before me.)
@cstross What, HOMEWORK? Damn! And I thought I was first....
-
@landley For a while, I figured Twitter was paying its way by providing a pre-sorted data feed for the AI company. (The tulip-craze money comes easier if you can say “Look at all the training data users give me out of the goodness of their hearts!”)
You're probably right that that's run its course though.
-
@Mschatelaine @zarfeblong @cstross lol. Here's your 500 nm chip with 3x ECC running at 800 MHz. $$$
-
@cstross I seem to remember an Asimov (?) story about something similar with ultra dangerous for power generation but even Asimov picked the back side of the moon not deep space.
It would be kind of poetic justice though if a space based Musk data centre got taken out by a stray starlink.
@etchedpixels @cstross Back side of the moon?! Do you want to get Space: 1999?! Because that’s how you get Space: 1999!
(And as I’m typing this I realize that people once thought we’d could have extensive moon bases before the end of the 20th century…)
-
R relay@relay.mycrowd.ca shared this topic
