disclaimer: i am stupid, this is not my field of expertise, all of my information is secondhand and melted together like silly putty in my brain and none of this should be taken as anything more than frustrated sentiment. this is not informational posting
I'm looking at this techradar piece about the fundamental horror contained within nvidia's computex keynote
and i'm just. ruminating.
the software-hardware industrial tcomplex is wile-e-coyoteing off the cliff because we already hit Peak Computer in a practical and operational sense, but since we haven't yet hit the limits of physics that prevent us from throwing more silicon at a problem
we have reached a point where throwing silicon at the problem to Make Computer Work Good has solved the problem, and now they're trying to create new problems to solve with this functionally limitless ability to burden sand with purpose.
For the end user, Computer is a solved problem. For almost everyone, Computer is a solved problem. But the stone could be milked for more blood, we still have more compute power that can be drawn out of it, beyond anyone's reasonable needs for it.
RTX is flashy but it's a problem created for the existing solution of "hardware fast what do to sell it because that doesn't matter anymore"
and following in that. crypto, nfts, "AI" of all flavors. Problems created to sell the solution of "hardware more gooder than anyone needs"
So nvidia unveils a compute board where a single unit draws 15 kilowatts and they want to sell them to pack datacenters full. One single chip of the 8 on that board draws twice as much power as my entire house in the less summery months. To do what.
It's just to sell ideas to people in suits and by the time they look down and the cliff face is twelve miles behind them, it won't matter, the money has been made and it's the rest of us that lose because they reactivataed a whole lot of coal plants to power datacenters full of systems doing useless work to create fictional value
Hardware has always been the salvation of the tech industry. It made sense early on when there were obvious improvements adding a math coprocessor, but you're right: we know now at this point what computers can do, and do well. The only frontier left is what computers can't do well.
And that will always have takers. That will always have suckers. It's perfect in that useful programs have tangible metrics that can be analyzed and altered, but AI apps only have vibes.
We're not really seeing an AI grift. It's more accurate to see it as a hardware grift instead. The people who sell the tools are the ones who get rich in a gold rush.
