Until we have MasPar chips or biological substrate chips,
True Strong AI such as at http://ai.neocities.org needs not more number-crunching chips for machine learning (a false goal) but simply better 64-bit von-Neumann-architecture chips. MindForth Mentifex AI at http://www.nlg-wiki.org/systems/Mind.Forth would benefit from a 64-bit native Forth CPU _lacking_ floating-point gates and simply providing execution of the most basic Forth instructions. Furthermore, the Wall Street Journal of 2016-10-27 on page B5 has an article by Sara Castellanos titled "Corporate IT Turns to Serverless Computing". It says that "serverless computing allows developers to focus only on writing code" such as http://github.com/PriorArt/AGI/blob/master/ghost.pl "without having to manage servers." The application of the customer (such as the two greatest AI's in the world, MindForth and the Perlmind q.v.) "is broken down into small pieces and placed into software shells...." Strong AI needs not fancier chips but more hospitable cloud environments. Deep Thinking is more important than Deep Learning. Mentifex AI Minds can already engage in automated reasoning with logical inference. ML is a detour on our way to True AI and the Technological Singularity.
[+] [-] mentifex|9 years ago|reply