3 ms·
Very weak article. It prepends several statements with “myth:” and does very little to debunk them. Example: AI won't spin out of control because Moore’s law
by tambourine_man 11y ago
Very weak article.
It prepends several statements with “myth:” and does very little to debunk them.
Example: AI won't spin out of control because Moore’s laws is nearing its end.
First, is it really? Just because we can't move electrons reliably much faster through much smaller sizes? What's stopping new materials, spintronics, photonics and what have you to take us to the next decades? I've been hearing the end to Moore’s laws is “coming real soon” since I was a kid. Lots of clever explanations guaranteeing there was no way we could move past 200nm or so due to physical size of the wavelength used in lithography.
And yet here we are using crazy stuff like phase shifting masks and interference patterns routinely.
Second, the brain seems to be a very slow and massively parallel machine, so maybe transistors are plenty small and fast already, we just need to ditch the Von Neumann architecture.
Third, we only need a single strong AI to emerge for it to be a problem. I don't see the clouds from these megacorps getting smaller any day.
I don't really know what I'm talking about, of course, but it takes stronger arguments than because quantum tunnelling and the speed of light to make this case.
- Qwertious 11y agoThe fundamental driver behind Moore's Law is shrinking the circuits. The problem is that the circuits are made up of atoms that deliver electrons, and electrical circuits fundamentally cannot go smaller than a single atom, because electricity is movement of electrons between atoms. We could, in theory, use other technology (some sort of photon-based computer?) to get smaller (and therefore faster) computers, but that wouldn't have the Moore's law mechanism.
- sawwit 11y agoThough, aren't there also plenty of other mechanisms that affect it? For example increasing speed of manufacturing procedures and improving knowledge about how to design complex systems. The end seems to be extremely dense blocks or spheres of highly optimized computation substrate, where each atom sits very precisely at the right place. But the state of the art is merely a few layers of crude photo-lithography. Imagine nanobots being programmed to move atoms around to assemble 3D circuits at the 10nm scale. I think this is where we are heading, and this will likely grant a couple of decades of continuing exponential progress. And by then our computation power will be insane. An equivalent of the simplest human brain models will likely fit into pea.