My sad prediction:
AI won’t get much better. Its current state is the peak, since it’s already been trained on everything the internet has to offer, regardless of copyright, and is already consuming unsustainable amounts of compute, and therefore energy.
But it also won’t go away. We’re just not living in a world where a product that sucks will fail anymore, as long as it’s pushed by big corporations.
And they’ll keep pushing it, cause if they stop, they’ll go bankrupt.
So where does it lead? A shift in culture! AI will keep being pushed down our throats until we start to accept it. It’ll become an “art style”. I think it already is.
Just look at this year’s Coca-Cola christmas ads. One of the largest corporations in the world uses AI for its most important ad campaign of the year, regardless of all the obvious errors in it. This is the new normal, and anyone criticizing images of humans with crab claw hands will just be called a luddite.
“AI” that actually works is much more modest than you think: it’s trained on small, focused, curated, legally licensed datasets mixed with synthetic data. It’s small and task focused. It’s heavily augmented and constrained to actually do what it’s supposed to, and not a bajillion other things.
It’s getting lower power. If bitnet takes off, inference costs will plummet, which means a lot will probably just be done on phones, with more control from the user.
Don’t get me wrong, AI is still going to be sloppy shit. This kind of stuff is going to be the norm, and way worse.
…But I think a “race to the bottom” will be a mixed bag, not pure cyberpunk. There’s a point where some uses get so cheap and “good enough,” it’s just too cheap to sell. Like someone trying to charge money for basic text editors. This is why the tech bros push the existential AGI angle so hard, as “devolving” into sets of specialized tools means their profit scheme implodes… Hence, hopefully, lazy landlords will have something a little less shitty to make an ad with than ChatGPT.
My sad prediction:
AI won’t get much better. Its current state is the peak, since it’s already been trained on everything the internet has to offer, regardless of copyright, and is already consuming unsustainable amounts of compute, and therefore energy.
But it also won’t go away. We’re just not living in a world where a product that sucks will fail anymore, as long as it’s pushed by big corporations.
And they’ll keep pushing it, cause if they stop, they’ll go bankrupt.
So where does it lead? A shift in culture! AI will keep being pushed down our throats until we start to accept it. It’ll become an “art style”. I think it already is.
Just look at this year’s Coca-Cola christmas ads. One of the largest corporations in the world uses AI for its most important ad campaign of the year, regardless of all the obvious errors in it. This is the new normal, and anyone criticizing images of humans with crab claw hands will just be called a luddite.
That’s a little extreme.
“AI” that actually works is much more modest than you think: it’s trained on small, focused, curated, legally licensed datasets mixed with synthetic data. It’s small and task focused. It’s heavily augmented and constrained to actually do what it’s supposed to, and not a bajillion other things.
It’s getting lower power. If bitnet takes off, inference costs will plummet, which means a lot will probably just be done on phones, with more control from the user.
Don’t get me wrong, AI is still going to be sloppy shit. This kind of stuff is going to be the norm, and way worse.
…But I think a “race to the bottom” will be a mixed bag, not pure cyberpunk. There’s a point where some uses get so cheap and “good enough,” it’s just too cheap to sell. Like someone trying to charge money for basic text editors. This is why the tech bros push the existential AGI angle so hard, as “devolving” into sets of specialized tools means their profit scheme implodes… Hence, hopefully, lazy landlords will have something a little less shitty to make an ad with than ChatGPT.