The technology doesn’t concern me that much. It definitely has uses. The thing that concerns me about it is how much money has been sunk into it. It’s basically the root cause of everything wrong with it.
So much venture capital has been pumped into the technology, and with the use cases where it shines alone, there is no way they’ll recoup their investment, let alone make money. So what they’re doing is trying to go all-in on this gamble by creating a ton of hype and putting it in everything to hopefully get back their investment.
They’re spinning up huge datacenters, that are sucking up unprecedented levels of water and power to run these things, and the end result is a tool that is not reliable enough for technical and scientific uses unless the training data is very specific about the topic in question. The generalized models are really only good for taking notes and composing emails.
The other big concern is how they’re trained. The tool sucks up truly staggering amounts of data online with no regard for anything. Privacy, copyright, and server bandwidth are simply not factored in, because the tool would be even less usable than it is now.
This is going to be the next dotcom bubble and things will get worse before they get better, because some rich assholes will need to lose some money. But after that happens, the industry will collapse to a point where things are reasonable again.
The technology doesn’t concern me that much. It definitely has uses. The thing that concerns me about it is how much money has been sunk into it. It’s basically the root cause of everything wrong with it.
So much venture capital has been pumped into the technology, and with the use cases where it shines alone, there is no way they’ll recoup their investment, let alone make money. So what they’re doing is trying to go all-in on this gamble by creating a ton of hype and putting it in everything to hopefully get back their investment.
They’re spinning up huge datacenters, that are sucking up unprecedented levels of water and power to run these things, and the end result is a tool that is not reliable enough for technical and scientific uses unless the training data is very specific about the topic in question. The generalized models are really only good for taking notes and composing emails.
The other big concern is how they’re trained. The tool sucks up truly staggering amounts of data online with no regard for anything. Privacy, copyright, and server bandwidth are simply not factored in, because the tool would be even less usable than it is now.
This is going to be the next dotcom bubble and things will get worse before they get better, because some rich assholes will need to lose some money. But after that happens, the industry will collapse to a point where things are reasonable again.