4 ms·
I am keeping my fingers crossed that there will be a 'Bloom Filter' moment for AI where we create an algorithm that can eliminate an expensive calculation in 99
by hinkley 3y ago
I am keeping my fingers crossed that there will be a 'Bloom Filter' moment for AI where we create an algorithm that can eliminate an expensive calculation in 99% of interactions and give us a 10x speedup in some larger problem domain.
Or some other 'trust but verify' situation where the suggested action is still validated against business rules that have some notion of consumer protection.
- usaar333 3y agoDropping numerical precision seems to have done that to some degree?
- hinkley 3y agoYou have a deterministic AI program to show me?
- dsign 3y agoI think the parent is referring to the fact that LLMs and generative models are, for the time being, “creating new markets” and “creating new problems”, but not really solving any of the hard problems we already have. I see it this way: you have this incredible piece of technology. You can use it to take the scraps away from a bunch of starving artists, and convert that survival income in wealth for some entrepreneur. Or you can use it to cure cancer or aging or make fossil-free airplanes. At the moment, the first application seems to be the one moving the market, and that’s really sad. On the side of hope, if somebody were to demonstrate how to use a bunch of GPUs to make the e-coli bacteria live for a little longer using an approach that has a semblance of generality, I guarantee that a lot of old people would be willing to part with their fortunes in exchange for a sliver of hope for them or future generations.
- soultrees 3y agoThat’s only because the companies with marketing budgets and marketing teams are the ones you are implying. There is serious research going down right now, but you don’t see you typical medical research department at your university buying ads on Instagram.
- pama 3y agoNot sure exactly what you mean here but most of the quantum mechanics calculations in material science and chemistry applications can now be done at dramatic speedups using such approximate models. It has been exactly this Bloom Filter moment a few years back and the complete transformation of how people in these domains work is well underway.