I suppose some folks were saying that about hydrogen-filled zeppelins too, but it turns out that you can’t just jam new tech down society’s throat that way.
We now have helium and hot air airships and the tech is used way more appropriately.
I do take your point that what whipped up AI frenzy isn’t just the tech, but the tech itself does needs to change before society will accept its ubiquity.
We always had helium. Those zeppelins were designed for helium. It was just Germany that used hydrogen during WW2, when the US (the largest supplier of helium) cut off their access to it.
Well, the power efficiency being much better is one problem to solve. Also various laws need to make pronouncements on copyright, etc. There is also figuring out what will happen to open weight models, what will become of challenges like deepseek.
Ultimately, ai will remain in some form, or catalyze another kind of technology, but it will certainly change tech landscape the way the printing press did.
I think a core part of it is going to be assigning the right queries to the right models.
Dont send recipe or ‘Google this for me’ queries to the latest most power hungry model, for example. Older models are already hundreds of times less computationally expensive than they were two or three years ago so the problem is already half solved in some sense.
I suppose some folks were saying that about hydrogen-filled zeppelins too, but it turns out that you can’t just jam new tech down society’s throat that way.
We now have helium and hot air airships and the tech is used way more appropriately.
I do take your point that what whipped up AI frenzy isn’t just the tech, but the tech itself does needs to change before society will accept its ubiquity.
We always had helium. Those zeppelins were designed for helium. It was just Germany that used hydrogen during WW2, when the US (the largest supplier of helium) cut off their access to it.
Yeah, instead of hydrogen zeps we use kerosene-powered planes that crash and explode way more frequently lol. Progress!
I’m not sure what you’re making here.
How so, what sort of change do you imagine?
Well, the power efficiency being much better is one problem to solve. Also various laws need to make pronouncements on copyright, etc. There is also figuring out what will happen to open weight models, what will become of challenges like deepseek.
Ultimately, ai will remain in some form, or catalyze another kind of technology, but it will certainly change tech landscape the way the printing press did.
I think a core part of it is going to be assigning the right queries to the right models.
Dont send recipe or ‘Google this for me’ queries to the latest most power hungry model, for example. Older models are already hundreds of times less computationally expensive than they were two or three years ago so the problem is already half solved in some sense.