Yliaster comments on Tech giants are spending $700 billion to make AI a utility — but they can't even get the power turned on. AI could become even more ubiquitous.

[–] Yliaster@lemmy.world 2 points 5 months ago (4 children)

I wonder if it's just a matter of time though.

Remember how computers used to be as big as a room and even then only stored 2mbs back when they were new? But then became more portable and accessible? It makes me wonder if, in a few decades, it'd be possible that happens with data centers in the future.

permalink

fedilink

source

parent

hideshow 4 child comments

[–] MyFriendGodzilla@lemmy.world 6 points 5 months ago

We can't do that though since we're already at the edge of the atomic (physical) silicon etching size limit. So no more steady increases in power, or decreases in size, unlike the last 30 years where we could do both. My hot take is that silicon chips as they are now defined are already at the end of the road.

permalink

fedilink

source

parent

[–] Flower@sh.itjust.works 2 points 5 months ago* (last edited 5 months ago) (2 children)

But in the same way we're now carrying a supercomputer in a pocket or purse. And they'll be a successor to LLM that's much more efficient. Most likely every laptop or phone puts those data centers to shame in a few decades.

permalink

fedilink

source

parent

hideshow 2 child comments

[–] Yliaster@lemmy.world 1 point 5 months ago (1 child)

Wouldn't that still eventually allow for an era(s) of monetizing token usage for universal applications? Even if it gets better over time and continuously replaces itself over decades

permalink

fedilink

source

parent

hideshow 1 child comment

[–] Flower@sh.itjust.works 2 points 5 months ago

Probably more for corporate use, digital autonomous companies and the like. If your home router is smart enough to do your taxes, you don't need much cloud computing. Although tech firms will try to get everything turned into a cloud subscription of course.

permalink

fedilink

source

parent