Driverless Dilemma

[Markymark]

Interesting, do you have a source?

I have read about amazons pricing which is now done automitically and does not follow a set of rules. It is constantly striving for an efficient and maximising pricing. I understand stock markets employ similar software.

 

[PhilDawson8270]

I have read about amazons pricing which is now done automitically and does not follow a set of rules. It is constantly striving for an efficient and maximising pricing. I understand stock markets employ similar software.

Surely though, the system responsible for the pricing. Follows an algorithm based upon several external factors to come to it's price. Or learns the market, but the process of learning that data, and how to process it, will be programmed?

note: I am not trying to pick an argument, I am genuinely interested.

 

[Markymark]

But how it processes that data, how it sorts it, how it identifies key parts of what it senses. Is programmed. At the moment, no machine can "think", they are not capable of abstract thought.

Although impressive, it is simply following pre-determined algorithms to process data based on sensors to the outside world.

Not predetermined. It's does start with inputs but it tests and retests milioons if scenarios and adapts the algorithm itself

 

[Markymark]

Surely though, the system responsible for the pricing. Follows an algorithm based upon several external factors to come to it's price. Or learns the market, but the process of learning that data, and how to process it, will be programmed?

note: I am not trying to pick an argument, I am genuinely interested.

I am no expert. It starts off with a crude algorithm. Runs a scenario. Tests it. Adapts the algorithm. Repeat millions of times. It will access large amounts of inputs, give weights to each, correlated not individually. There's no fixed answers but it will deduce from a thousand inputs the 'best' course of action after they're weighted against each other.

 

[Markymark]

A good exmaple is you amking a sandwich. I bet you;ve never been taught the exact routine. You had an ideo fo the inputs (filing, spread, bread, taste). You then tried it, it was horrid (who wants marg in a sandwich), adapted it, and created a complex algorithm that makes a nice sarnie. I doubt there's a set of rules, but you know exactly how to do it. It's also adaptable...make in in someone's kitchen? Reckon you could. Sliced turkey instead of ham,? You;'d make it ok as a 'best case scneario' with what yuo have.

Computers would learn exactly the same way. They not teach it how to mkake a sandwich, but teach it the basics, teach it what tastes nice, and let it learn. It would take an awful lot longer and have some horrbie sandwiches during testing but the end result with a perful enough computer would be some yummy sandwiches.

 

[mustang1]

The reason a human asks such a question is the reason why I'd trust a driverless car rather than one driven by a human. Having said that, the driverless car is programmed by humans but I have more faith in logical thinking nerds than regular humans.

 

[Markymark]

[QUOTE 4251097, member: 9609"]But how will it know whether something was a good or bad result,; for instance if the car is regularly killing birds, will it learn what sort of roads these killings were occurring on and modify its speed in future, or will it have been told that killing birds doesn't matter. Somebody has to have told the car what its goals are.[/QUOTE]
Yes. Minimising death and injury would be the goals.

 

[winjim]

I've had one of those for the past 2 years

Oh dear.