> Nobody wants to pay for a trillion dollar cloud bill.
Buying dedicated hardware as a way to keep your AI bill down seems like a tough proposition for your average consumer. Unless you're using AI constantly, renting AI capacity when you need it is just going to be cheaper. The win with the on-device model is you don't have to go out to the network in the first place.
You misunderstood what I meant, I mean make models that run on potatoes, nobody wants to pay what chatgpt's subscription model probably SHOULD cost for them to make a profit.
No, not even sure how you arrived to that conclusion. The idea is that there are models out there that can run on small amounts of VRAM. If all it costs is charging your phone, as opposed to some subscription to some overvalued AI company, people will choose ‘free’ first. We have models that can google things now. They only need to know so much when online, and a specific subset when offline.
I think there are lots of advantages to running a model locally. Saving money is one of them, but that's only if you can keep the thing busy. You wisely put the "free" in quotes for a reason: you paid money for the hardware the model is running on, and you're paying for the electrical bill to power it too. Even if you pay a 100% markup to the cloud, unless you're keeping it busy 50% of the time, it's cheaper to rent.
Buying dedicated hardware as a way to keep your AI bill down seems like a tough proposition for your average consumer. Unless you're using AI constantly, renting AI capacity when you need it is just going to be cheaper. The win with the on-device model is you don't have to go out to the network in the first place.