kaled - 8:25 am on Oct 11, 2010 (gmt 0)
The issue is not how many accidents occurred, the issue is how many interventions were required by the human supervisor - the statement with respect to accidents (or lack of) is just spin!
They said in test runs there was only one accident
You're thinking too narrowly kaled. Full autonomy is not required and AI is not necessary. I'm thinking about a modified infrastructure approach.
I'm going to be blunt - it's what I do best...
You're not thinking at all. We have electrical/electronic infrastructure on the roads already, have you never seen traffic lights? I'm guessing you have seen traffic lights but have you ever seen traffic lights that are guaranteed to never go wrong? Have you any idea how much it costs to install a set of traffic lights? Have you any idea what it would cost to install the infrastructure you're talking about? Have you any idea how long it would take? Your suggestion of using advanced infrastructure to automate driving is even more of a white elephant than what Google is trying to do.
Just for the record, I have a degree in Cybernetics (albeit a long time ago) so I do know what I'm talking about.