Thanks for the Tensorflow reference, roofy. I'll look into that SW library. Thank God for open-source coding libraries! In general, I'm a big fan of coding in a language which has open source libraries as this can avoiding a huge amount of reinventing the wheel / boilerplate coding.
I think AI can be capital intensive if a team is trying out different ideas quickly in a 'fail fast' approach (which GSW seems to be doing, based on Noto's approx. monthly turnaround for features adds / bug fixes).
However, I agree with you in principle that AI coding generally needs i) time and ii) lots of data. Think about the Junior RoboCup Soccer comps. The runway is about 3yr+ in coding in the lead up to comps. I remember during my undergrad in around 2010-2012 and I seem to remember people were talking about a 2015 competition! (Don't quote me on that.) Regardless of the exact runway, it is evident to me that coding logistics AI would be a lot simpler than coding the brains of a robo dog team to win a soccer game against other robo dogs!
Nevertheless, it seems intuitive that tweaking logistics AI will take time and lots of data if you are aiming for it be 'Google Maps' quality in terms of app experience. Remember, Google Maps has been around for a while now, and it wasn't so long ago that they added rerouting features based on live traffic feedback (e.g. traffic jams, roadworks). Behind the scenes, they may have been collecting a ton of data over quite a few years before they released that feature.
GSW Price at posting:
53.0¢ Sentiment: Sell Disclosure: Held
