Ask HN: AI on the open web? Really?
It is a full-blown OS-in-a-browser. After 2+ years working on the nuts and bolts of the system and a few silly apps, I have recently gone all-in with the AI concept that initially turned me on to programming.
Using dynamically generated SVG faces, and the Chrome JS APIs for voice recognition and synthesis, I have been able to create a kind of "Hello World" example of AI. To jump right into it with Chrome, see https://www.urdesk.net/desk?intro=bertie
I am currently working on integrating the "Brain" module that parses the natural language input with the terminal, so users will really be able to dig into the internals.
My goal is to move towards full-blown run time configurability so that developers will be able to provide their own SVG faces, response templates, and even entire ontology schemas in order to develop domain-specific AI's.
Right now, the AI obviously only has a lone sense of hearing. The next sense to work on might be seeing, by way of capturing webcam stills and running them through a machine vision algorithm. I really doubt that there will be a JS API for vision any time soon... maybe a Native Client module could do the trick (at least for very basic seeing).
Given the AI prototype that I've developed that uses only Web technologies (even if they are non-standard/Chrome-only) what do you think about starting to turn the online world upside down? For instance, it is pretty simple to connect all of this to the outside world via local node instances that can be talked to very efficiently with websockets, which can then send control signals to real world objects using modules like johnny-five (a JS robotics interface).
All I'm saying is that I am so infinitely sick-and-tired of the lame MVC/Backbone/React/Docker/whatever silliness that just doesn't do anything, if by "do", we mean having some effect on the world above and beyond painting some pixels on a boring LCD!
Who's with me?