Skip to content

Into the Fray Uber Leaps…

August 19, 2016

I read today that Uber is planning on introducing autonomous cars into the Pittsburg area soon.  This is key for them as a significant cost for Uber is the required human driver.  If they could eliminate the driver from the car Uber would be cheap, potentially cheaper than owning and driving your own car.

Sounds great, right?

Well, if you refer back to an earlier post about Google’s self-driving car you will note that I made a big deal out of the difference between a rules-based system and a Strong AI based system.  My opinion on this hasn’t really changed and I would suggest people in Pittsburg might want to take heed of this.

Today, with what we know about computing and computing systems, it is possible for an army of people (literally) to code up a system which is rules-based.  There are a number of examples of this having been done already on small and large scale.  One particularly interesting system that is rules-based is those that are used for high-frequency trading on various stock exchanges.  Yes, it is fairly simple to have something that follows rules that say “When conditions X, Y and Z are met, do ABC.”  That is the essence of a high-frequency trading system, and these systems are very, very effective at making money.

The problem is, the goal of the high-frequency trading system is to make money.  This is done by having lots of rules and lots of things that result from these rules.  All the time people are adding conditions to the list and extending how these systems behave.  To the outsider, these systems might be considered to be “smart” or even “intelligent”.  But that is far, far from the case.

The big difference with a self-driving car is the goals are different – and not just in magnitude.  You have multiple, often conflicting goals.  Sure, you want to move the car from A to B, but you also want to do so safely, following traffic laws and avoiding hazards.  The problem with a static rules-based system is the environment is changing all the time, and unlike the high-frequency trading situation, you can’t just ignore conditions you don’t understand until things change.  It is necessary to always have a safe response to every situation.

Where Google’s car fell down on the job previously was they didn’t have enough rules to account properly (and safely) for a bus.  Uber sounds like they are running into the same sort of limitations, although they are putting an engineer in the car to “help” it along to start with.  The problem is, with a rules-based system in a chaotic system there are never enough rules.  Whenever a unforeseen situation is encountered and there are no rules, a rules-based system fails to be safe because it has no fallback.  It sounds like Uber’s system simply stops and demands a human intervene when this happens.

An AI-based system would have a fallback and likely it would be safe.  The problem is, we aren’t there yet.  Nobody has built a AI system that is anywhere near capable of dealing with driving a car and we are likely decades (at least) away from being able to do any such thing.  The problems are legion, and we have been working pretty hard at it since at least 1970 if not earlier.  This isn’t a processor capacity or speed challenge, so faster computers are not the answer; sure they help, but that isn’t the real underlying issue.

What both Google and Uber are almost certainly doing is trying to win the race to having a self-driving car without having to wait for the AI folks to come up with the right answer.  So they are taking a shortcut and using a rules-based approach so they can get their cars moving on our streets.  Sounds like a real business-oriented approach, except this is an issue that is going to have consequences far, far wider than the boardroom or the accounting department.  When a rules-based car runs over a child (and this will almost certainly happen), we will begin to understand the consequences of letting a businessman decide they need to win this race without an AI system behind it.

No comments yet

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: