|
Post by You Must Enter A Name on Oct 14, 2014 20:55:23 GMT 10
Brad I agree with you on everything except your comment on Basic, it was the first programming language I ever learned and to be honest it is just that BASIC and everything you learn in it is universal throughout all "Programming" languages. That is you must define EXACTLY what will happen and under what circumstances. You can say,
IF inputVideo1 = (Target) Then ' Where target is determined from a range ' Call execute() Else Call Scan()
A little bid vague and a touch of modern in the statement but MEH! True AI is not possible with existing programming languages in my opinion, not when you must define every circumstance and reaction, aka logic. Now the movie transcendence is probably much more onto the path.
|
|
brad
Senior Member
Posts: 156
Likes: 288
|
Post by brad on Oct 14, 2014 21:13:58 GMT 10
one word Daniel, Transformers! are you telling me they are not real? if so that really P's me off as I wanted a car like that. When I say basic code I mean the code to do it is pretty basic/simple(not BASIC the language, my example was SAS code simplified as it was the first language I thought of). That doesn't mean it would be programmed in a day, it would take much longer (year or two before real testing) to complete a program as like you explain above, each step has it's own variables that need to be set. A predictive model will have about 4 separate programs to define what each variable means, for something like this there would be a heap of programs for each variable and there will be a lot of variables. A lot of programming and a lot of testing however the code used isn't that hard to work out.
|
|
|
Post by You Must Enter A Name on Oct 14, 2014 21:51:01 GMT 10
I see, but how does a program exceed it's programming while keeping within the limits of it's language?
|
|
brad
Senior Member
Posts: 156
Likes: 288
|
Post by brad on Oct 14, 2014 23:41:00 GMT 10
it doesn't, the program learns to write new variables based on what you have put in. Once it has been through the learning stage by saving results and analyzing what happens if one leg is not in front of the other it will have the ability to insert macros or code then enhance that movement to make it smoother.
If the program says left then right foot then to jump it has to have both feet side by side it will work out through trial and error that it is better to have one foot back and the other in the air and forward to jump a puddle. The code is there, the program just has to insert new variables based on the current situation.
So if a robot walked up to a creek and wanted to cross without getting wet it would have to first look at the situation, recall what it has been told and work out the above is the best way to do it.
So the first stage of "learning" is just taking code or macros you have already put in and inserting and saving them into different situations.Robots are already partly self aware and have created their own language to communicate with each other, I forget the country and the lab that has made them however there are two robots the same, and they can tell each other to do things like raise your left arm, if the other robot doesn't do the right thing they say no in their own language and then demonstrate to the other robot what they want. This is the program writing new code as it goes as they are creating their own language. So when robot A says "Do nee" and the other robot gets it wrong, robot A will show robot B, robot be will then write a new line of code like: "Do nee" = left arm up.
So they are always within the means of the standard code language however they are using variables based on a language they created.
I don't really see limits on programming languages as you are only limited to the variables you use. I think it is pretty dam exciting when you think about the possibilities. My company will be joining challenge to build a robot in the near future however the programming is what I am working on now. There are endless variables that need to be created however it is still based on the very simple principles of if/then/else. I have a lot of other algorithms in there like law of cosines to assist with location to a pin point and accurate geo location at all times, I have geo risk factors in there so based on location the program will have a base to make sensible actions. Biggest problem is the computing.
For the past 5 months I have been working off the laptop, mid next month I will have my computer arrive. It has 32 gig of ram, 2 tb in solid state hard drives. GFX graphics and speced out to the max yet it still will not be fast enough to run all the code at the same time in real life situations. I can test code, I can have it run situations in a controlled environment however having the program plugged into multiple senses like, geo, sound, temperature, vision, and smell will more than likely crash the system as there is a lot to compute at the same time, just in the front yard you have cars going past, animal noises, wind, light and a heck of a lot it has to look at and take in. So controlled environment is all that this program will be able to handle on a PC. Quantum computer on the other hand can and does automatically take in multiple streams of information at the same time, and also breaks up all that information, looks at all possible variables and calculates the best action within milliseconds.
This means I have to do a lot of teaching so the program doesn't have to learn as much by just walking out the front door. Fun stuff, very basic when you break it down however a lot of code.
|
|
|
Post by You Must Enter A Name on Oct 15, 2014 4:49:50 GMT 10
Cool stuff Brad, looks like my knowledge is still stuck in the 90's
|
|
wolfstar
Senior Member
Posts: 1,020
Likes: 917
|
Post by wolfstar on Oct 15, 2014 8:02:35 GMT 10
That's what I mean, they have a few projects currently going that they have forced to learn everything independently. No coding used at all. That's why it's taking so long. It had to even learn language. I'll have to find the show I was watching on it for you.
|
|
brad
Senior Member
Posts: 156
Likes: 288
|
Post by brad on Oct 15, 2014 13:57:02 GMT 10
There is always code used wolfstar, robots are just shells bits or metal or plastic with actuators and parts the program tells to move. Self learning robots are writing code all the time, this is how they remember what works and what doesn't work.
|
|
Frank
VIP Member
APF Life Member
Posts: 1,864
Likes: 2,739
Email: frank@ausprep.com
|
Post by Frank on Oct 15, 2014 20:18:22 GMT 10
I must admit I am simpleton when it comes to anything programming related. Computers are purely a tool to get what I need done. This article was and all the subsequent discussion has been interesting, thanks for all the input. Now, how long until I can get my terminator??
|
|
brad
Senior Member
Posts: 156
Likes: 288
|
Post by brad on Oct 15, 2014 23:42:51 GMT 10
lol Frank, don't hold your breath with my company, we are only at the programming and virtual testing stage however Google may have something for you a lot sooner. If all else fails you can always get a furby.
|
|
Frank
VIP Member
APF Life Member
Posts: 1,864
Likes: 2,739
Email: frank@ausprep.com
|
Post by Frank on Oct 16, 2014 7:36:57 GMT 10
Killer furbies would be terrifying......creepy little bastards
|
|
wolfstar
Senior Member
Posts: 1,020
Likes: 917
|
Post by wolfstar on Oct 16, 2014 8:21:17 GMT 10
|
|
overlord
Senior Member
Posts: 614
Likes: 720
|
Post by overlord on Oct 16, 2014 12:47:00 GMT 10
Self-learning systems have been around for quite some time. The Japanese have developed some systems that are based on that.
Before, Neural Networks (something I enjoyed reading and contemplating) were the fad and norm for Artificial Intelligence. Nowadays, it is something different and more efficient
|
|
Frank
VIP Member
APF Life Member
Posts: 1,864
Likes: 2,739
Email: frank@ausprep.com
|
Post by Frank on May 29, 2015 3:49:44 GMT 10
Another article on AI. These kinds of things keep cropping up lately www.news.com.au/technology/innovation/academics-warn-killer-robots-will-leave-humans-utterly-defenceless/story-fnpjxnlk-1227373294083AN ACADEMIC from California has warned that killer robots currently being developed by the United States military will leave humans “utterly defenceless”. There are currently two programs being commissioned by the US Defense Advanced Research Projects Agency (DARPA) that are attempting to create drones that can track and kill targets without any human intervention.
“Autonomous weapons systems select and engage targets without human intervention; they become lethal when those targets include humans,” professor Stuart Russell wrote in the journal Nature.
“Existing AI and robotics components can provide physical platforms, perception, motor control, navigation, mapping, tactical decision-making and long-term planning. They just need to be combined.
“In my view, the overriding concern should be the probable endpoint of this technological trajectory.
“Despite the limits imposed by physics, one can expect platforms deployed in the millions, the agility and lethality of which will leave humans utterly defenceless. This is not a desirable future.”
The robots themselves are called lethal autonomous weapon systems or, if you want them to sound even more bad arse, LAWS. They’re expected to be armed quadcopters or mini-tanks that uses an artificial intelligence to determine whether or not to kill someone.
Professor Russell said: “LAWS could violate fundamental principles of human dignity by allowing machines to choose whom to kill — for example, they might be tasked to eliminate anyone exhibiting ‘threatening behaviour’
“Debates should be organised at scientific meetings; arguments studied by ethics committees. Doing nothing is a vote in favour of continued development and deployment.”
Elon Musk is well known to share similar fears with killer robots and artificial intelligence, famously saying, “Summoning the demon” of self-learning artificial intelligence would be “potentially more dangerous than nukes”.
|
|
|
Post by graynomad on May 29, 2015 10:04:24 GMT 10
This sort of thing is a big mistake IMO, nothing good can come from it, especially for us minions.
|
|
Frank
VIP Member
APF Life Member
Posts: 1,864
Likes: 2,739
Email: frank@ausprep.com
|
Post by Frank on May 29, 2015 12:03:14 GMT 10
Yep, dangerous territory
|
|