/sci/ - Science & Math

>>9984466
>also, humans wont be able to program something smarter than themselves
You're a moron, every day at work I program things smarter than myself. All my job consists of is writing programs to complete tasks that no humans could do because they're too tedious or difficult. Lately I've been building fail-safes into the system so that it can detect internal failures and immediately rectify them and notify humans to come take a look. Please note that this is an AI system I am talking about which has a pipeline of new data being continually fed to it to augment its intelligence. This includes everything from new algorithms defined in a generically loadable way, to new scores for various subsets of inputs that the AI system can receive.

It's disturbing how /sci/ is so dismissive of this because none of them work in the industry and they don't fucking get it. They have no future vision because they haven't worked on the types of things that people at major corporations leveraging AI have worked on. You have no idea how fast this is all going to pick up once things get going. My team at work has been achieving exponential gains every year, and the humans running the system haven't been getting exponentially smarter or faster workers, that's for sure. So where are these gains coming from? From increased coherence between each subsystem of the AI system and enhanced optimization techniques