How far are we from them?
Self-driving Cars
reminder that this completely BTFOs any self-driving car for the foreseeable future.
Loss, right?
>How far are we from them?
They already exist.
>Being less than perfect is a problem
Sure glad we have human drivers on the road keeping those perfect 0% injury and fatality stats year after year.
Meanwhile in the real world Uber already bought a fleet of self-driving Volvos for 2019.
Also:
arstechnica.com
>To repeat: these kinds of attacks worked on the specific machine vision system the researchers trained, and the altered signs in the gallery above would not fool any cars on the road today.
Ready now. The thing is you run it mostly AI but have remote human error handlers. Waymo is already deployed and running in Arizona.
There is not a huge rush though as the big tech behind it doesn't care about hitting market asap as much as getting it perfect.
That's because all the real money is placing it into self driving semi trucks and eliminating the trucking industry.
>Self-driving
programmers wrote their programs
general intelligence doesn't exist you morons
Good point. OTOH we might not need to rely on visual driving signs in the future.
That's actually not the problem. The problem is that we'll likely have the same algorithm working in every car in the self-driving fleet. Which means if you can break one, you can break all of them. (As opposed to humans, who just randomly crash.)
Your point of self-driving cars being safer than humans is well-taken, but is not relevant to the issue of adversarial attacks.
When that semi accident happens, the company will lose all support.
...