| ▲ | s08148692 3 hours ago | |||||||||||||
The future is going to be fascinating If we have enough fabs theres a world where every single problem has a completely custom chip to maximise efficiency of whatever problem needs to be solved. End to end AI - Ask to solve a problem, AI designs the chips, fab delivers it, AI writes the software in a completely custom language. Nobody will understand how anything works other than knowing that it does work | ||||||||||||||
| ▲ | ywvcbk 3 hours ago | parent | next [-] | |||||||||||||
AI so far can't really do anything that humans can't though? (obviously it's infinitely faster) if building extremely customized chips so far has proved to be suboptimal (even if we partially exclude the engineering salaries) why would AI change that? This approach also implies that optimization and performance will be come highly important? But it's not (outside of a few use cases) and the software that even the SOTA AI labs are building is often an unoptimized mess just like human written software used to be (if not worse) | ||||||||||||||
| ||||||||||||||
| ▲ | Tuna-Fish 3 hours ago | parent | prev | next [-] | |||||||||||||
Right now, the cost that limits doing small custom chips is not designing the chip, it's the cost of making the mask sets for the chips and other foundry NRE, which is well above $50M for a single leading edge chip these days. This money is generally not spent on things that AI can easily make cheaper. | ||||||||||||||
| ▲ | edwinfredy 3 hours ago | parent | prev | next [-] | |||||||||||||
Might allow further growth in chip performance now that simply making smaller transistors is seemingly approaching its limits. | ||||||||||||||
| ▲ | FridgeSeal 3 hours ago | parent | prev | next [-] | |||||||||||||
Surely this also potentially unlocks the scenario where “good enough” chip fabrication becomes easier, because not everything needs to rely on the single-most-cutting-edge node-fabrication capabilities, because they can now gain more from architectural customisations in return. Could be verrry interesting . | ||||||||||||||
| ▲ | bentobean 3 hours ago | parent | prev | next [-] | |||||||||||||
Isn’t there an old episode of Star Trek where they visit a planet with a population that has become utterly dependent on a supercomputer? And as a result everyone has become very stupid? Will have to track that down… | ||||||||||||||
| ||||||||||||||
| ▲ | chii 3 hours ago | parent | prev | next [-] | |||||||||||||
3D printing is the beginnings of this sort of capability. But current "home" lab 3D printing is incapable of producing chips - may be one day it will. | ||||||||||||||
| ||||||||||||||
| ▲ | toyg 3 hours ago | parent | prev | next [-] | |||||||||||||
That "fab delivers it" does a lot of work, in that concept... | ||||||||||||||
| ▲ | exe34 3 hours ago | parent | prev | next [-] | |||||||||||||
Unless it infringes on big tech patents or other wealth extraction measures, then it'll just be computer says no. | ||||||||||||||
| ▲ | techpression 3 hours ago | parent | prev | next [-] | |||||||||||||
Wouldn’t AI design a completely programmable base that can become any type of chip instead? It seems limiting that you’d need custom chips for different tasks, a chip could morph on demand, even include different architectures that change during runtime dynamically, or even reinvent itself as required. If we use the very abstract term “the future” that is. | ||||||||||||||
| ▲ | chrisjj 3 hours ago | parent | prev | next [-] | |||||||||||||
Uh, how will they know it does work? | ||||||||||||||
| ▲ | tourmalinetaco 3 hours ago | parent | prev [-] | |||||||||||||
Which is, effectively, magic. A terrifying and yet exhilarating potential. | ||||||||||||||
| ||||||||||||||