[FoRK] Dude, where’s my job?

Stephen Williams sdw at lig.net
Fri Jan 16 08:53:08 PST 2009


J. Andrew Rogers wrote:
>
> On Jan 15, 2009, at 9:02 PM, Stephen Williams wrote:
>> Autonomous vehicles are already solved.
>
>
> Wait, what?  That is an area near and dear to my business and that is 
> the first I have heard of it.  Yes, people are working on it, but 
> "solved" is not the term I would use.  It is solved if we restrict 
> "solved" to only include very narrow and specialized cases.  There is 
> a lot of work left to do, and it is a *lot* trickier than it sounds to 
> come up with a generally useful solution.  There are still fundamental 
> theoretical computer science questions that have to be answered first 
> such that we contrive test cases that allow us to avoid asking them.  
> The technology is getting better, but is still pretty stupid in ways 
> that become obvious in more edge-like cases. It would be cool, but the 
> tech only exists to do Disneyland versions. When you used the word 
> "solved", you forgot to add a bunch of asterisks and attached caveats.

Of course it's not ready.  Of course it's not safe enough yet.  But it's 
far more usable than a "Disneyland version".  There, nothing's real and 
cheating for effect and illusion is just fine.  The last Darpa 
challenge, paired with new vision techniques like one I just read about 
to avoid colliding with pedestrians, is getting pretty close.  Certainly 
there are edge cases and liability issues.  But I hear that the 
technology is already being integrated into tractors and military 
vehicles.  I'm sure that rock haulers and similar are next.  In-factory 
vehicles.  The point is that several instances of a very general purpose 
solution have already passed a fairly rigorous test that required full 
California driving rules in an unpredictable multi-vehicle city 
situation.  It's clear that current technology would work most of the 
time.  That's a huge step.  When could it come to market?  Perhaps as 
soon as the savings are significantly more than the insurance needed to 
pay claims for accidents.  Multiple methods for minimizing accidents 
will probably be required.  Quite a bit of safety could come from simple 
rules like: lose GPS lock unexpectedly -> stop.

Specifically, the current DSLAM algorithms are a huge improvement, and 
involve insight that can be used in other problems.  And they are built 
on probabilistic relational reasoning and learning (Bayesian / Markov 
with approximate reasoning algorithms for efficiency).

We already have armed autonomous fliers, and that is an arena where you 
could have some serious failures and vulnerabilities.
>
> Also, I would observe that the commercial infrastructure for such 
> things that is being worked on includes deep and pervasive real-time 
> tracking of individuals, but I will leave it to the civil libertarians 
> as to whether or not that is a good thing.

That is something that is, or could be, coming from several directions.  
It doesn't really need any autonomous vehicle AI.  If you have a cell 
phone, or for that matter, a face in a camera-infested area, you're 
already trackable.
>
>
>> Hard problems have been falling fast.
>
>
> Not fast enough, but I will take what I can get. :-)

sdw
>
> Cheers,
>
> J. Andrew Rogers



More information about the FoRK mailing list