2 ms·
Personally I think that computer programming AIs that can really replace humans (beyond just programming automation) will need to be AGIs. Those AGIs will be a
by ilaksh 9y ago
Personally I think that computer programming AIs that can really replace humans (beyond just programming automation) will need to be AGIs. Those AGIs will be able to do philosophy just as well as they can program.
I also personally believe that we currently have most if not all of the ingredients necessary for AGI. My prediction is that in 2018 or 2019 we will see public demonstrations of AGI. These initial demos will be underwhelming and not have the capacity of humans or even animals -- but nonetheless will be truly general intelligences with significant human- and animal-like abilities.
It will take another few years before the technology is widely recognized to be human-equivalent or better, but the initial systems will be trained in numerous fields quickly. There will be very powerful general systems available by 2021.
Some of the pieces necessary for general intelligence as I see them are truly general purpose inputs and outputs (like a body or virtual body), highly efficient processing of high bandwidth input and output, fast online (immediate) learning, hierarchical learning and computation, the ability to learn and recombine flexible sequences and time-based data. Advanced neural net-based systems, especially ones that are inspired by animal cognition systems, are providing all of these features. Its mainly a matter of integrating existing leading edge neural-net research with a good understanding of AGI.