Superintelligence not a ‘sci-fi risk’ and possible in ‘next decade’: Sam Altman
videogamer.com
videogamer.com
One of the big decisions: it will need rich inputs/outputs, human-like avatar with a direct feed on huge data, access to classic super computers, access to its own R&D aparatus (built by us for it)?
One training session will probably last for years and it's problematic as if it becomes self-aware, resetting the neural net or switching off the power will be murder.
It _will be built_ in the future if human sciences/maths have nothing to save us from extinction. Or it could become an "arm" race between the most powerful countries in the world. If the chinese military starts to build one, US/EU military will follow.
The outcome is totally unpredictable, only the soon-to-come end of humanity or a serious arm race could justify to build it.
Already there.
Will we never stop seeing this kind of self-serving pumping from tech entrepreneurs?