Space Data Centers, Bubbles, and the Failure of Reporting
I really wish reporters did a better job dealing with obvious hype like this:
After Ars wrote a story on the potential of autonomous assembly to construct large data centers in space, Musk responded on X by saying that Starlink satellites could be used for this purpose.
“Simply scaling up Starlink V3 satellites, which have high speed laser links would work,” he said on the social media site X. “SpaceX will be doing this.”
This is almost certainly not going to happen on any reasonable timeframe for a variety of reasons that are not mentioned in the article at all:
Data centers give of a lot of heat, and dissipating heat in space is surprisingly difficult
Data centers running GPUs, need to replace equipment surprisingly often — 3-5 years if they are running under high load, which these will be. So you would need to replace the entire load out ever few yers — which means dragging an entire new data center worth of GPUs into orbit
Except, of course, that equipment does not all live and die on the same schedule. In practice, you’d lose some percentage of these things on a regular basis, not to mention all the normal stuff that can go wrong.
So you need to have the capability to replace/repair almost constantly. Robots? Maybe, but robots aren’t always great at fine repair work, and the communication lag might be an issue
So are you going to fly up people every time you need work? Or is there going to be enough work that you need to keep people there on a regular basis? All of that drives up the costs.
Never mind the cost of shielding and other space specific requirements.
Musk never delivers on his promises.
The article does mention once or twice that critics think these would not be economically viable, but it doesn’t dig into the why, which gives a muted impression of just how difficult this is going to be. And it spends a lot more time on famous rich people, like Bezos and Musk, claiming that they will in fact do this.
Okay, so why does this upset me so much? It’s not like reporters aren’t allowed to have bad days. And yeah, maybe I am just having a bad day. But I find this kind of reporting infuriating. Limiting the discussion to “famous person says” and “critics disagree” without really talking about the issue gives a distorted view of the issue. And it doesn’t take much to give at least a cursory overview of the whys of the critics. By limiting the discussion to he said, critics said, it creates an environment of less information, which can lead to poor collective decisions. We are in an AI bubble at least in part because the tech press has been reluctant, generally, to discuss how I works, what it does poorly, and how realistic the claims of the people whop promote it actually are.
Bubbles are bad. Reporting that discourages people from understanding well enough to be informed is bad. And I feel like, more and more, we just get reporting based on access and prestige instead of facts and reality. Maybe I am over-reacting to this article. Maybe I am not giving enough grace to a reporter trying to meet a deadline in a hard media market. But I don’t think I am expecting too much to ask reporters to spend more pixels on the details and fewer pixels on the press releases and unsubstantiated claims.
I don’t want dissertations, really. I just want some discussion of the reality of the claims. And I don’t understand why our media is so resistant, in so many areas, to doing something that should be the table stakes for any decent reporting.