Comment on Nvidia insists it isn’t Enron, but its AI deals are testing investor faith
vacuumflower@lemmy.sdf.org 1 week agoWhat if it’s not a bubble?
So I tried using some AI chatbots to find a movie recently, it made up a few, none being the answer.
(The question was about a historical movie, made in perhaps 1970s by the feeling, set someplace in southern France somewhere around 1650s, has a few beautiful views of nature and castles ; one scene where a guard captain enters a room, asks a question, as a power gesture drinks a glass of wine on the table and a minute later falls ; another scene where for whatever reason a rapier fight happens in something like a tavern, two women in pastel dresses are descending by an open ladder from the second floor, seeing the brawl take our pocket pistols, one of them is stabbed with a rapier ; another scene where a guy is getting question with his feet over the fire ; another when another guy is climbing a tower clinging at brick mortars outside and hears guards’ boots on the ladder very loudly ; when I was a kid and saw that, someone said it’s an adaptation of something by Lope de Vega, but I’m not sure that’s correct ; that’s just in case someone reading this knows such a movie.)
But some googling sessions they do optimize, without you the user ever having to browse a webpage, and just getting a textual answer. That’s a valid use.
And some other processes. They don’t have to be useful for all things they are applied to, just some profitable.
andallthat@lemmy.world 1 week ago
Being a bubble does not mean the service they provide is useless. It means that the service is never generating enough profit to repay for the huge cost of providing it.
Would you pay 500 dollars a month to have the possibility to do your movie searches? Or alternatively, would you like your LLM of choice to counter that, having read all your emails and browser history, you are probably interested in a totally different movie that just happens to be playing now at a nearby cinema?
Because these AI companies are currently burning through literally a good chunk of all the cash in the world and they will eventually need to make even more gigantic profits to repay that cash. And the only one that is currently making money from AI seems to be NVIDIA, by selling the hardware that powers the AI giants.
I’m not saying that it IS all a bubble, by the way, as I can’t read the future and these gigantic profits might well materialize in the future. I’m just saying that “bubble” and “useless” are different.
vacuumflower@lemmy.sdf.org 1 week ago
There might be a more direct parallel than originally intended in this with the explanation how one person works hard all day and makes less than another person who pushes a few buttons. The latter knows which buttons to push.
This technology is useless for my movie searches, but it might be useful in the same way as radar was for air defense.
BTW, I’m not sure what I’d choose if offered to pay 500 dollars for knowing what that movie is. There’s one girl, if she’d be interested too to find that movie, perhaps I would.
So if such an expensive technology would allow this kind of nuanced search, and more seemingly efficient wouldn’t, then we have a use case.
Or a model allowing to predict actions of other people sufficiently well, based on seemingly not precise enough data. However much it would cost, that would be justified, similarly to high-frequency trading, because it would operate on all existing value, not just what it generates.
I know, I was making two points, one is that everything is relative (what you’ve just agreed to), another is that it might not at all be a bubble.
CheeseNoodle@lemmy.world 1 week ago
On the air defence note, specialist systems are already amazing at tasks like this and can usually run fine on high end consumer tier hardware (up to a single server rack) rather than data centres with the power requirements of small countries. AI is here to stay but LLMs are absoloutely a bubble.