Being involved in computers and having spent some time on Decision Support Systems (DSS) I have managed thousands of man hours writing algorithms which would enable the computer to make predictions that would allow our company to maximise profits. I have been told that these ‘intelligent’ systems are the first stage of Artificial Intelligence (AI). I’m always arguing that we are nowhere near AI at the moment. I don’t even class the systems as intelligent. They just give the appearance of intelligence because they do things we generally cannot do easily.
You see human intelligence and what is currently called AI are actually so far apart they are not the same thing. Sure we can make it look like they are but that is the intelligence of the development team rather than the intelligence of the computer. Computers simply follow one instruction after another as programmed by the programmer. No more no less.
I’m sure you remember the blonde joke;
A young blonde woman is distraught because she fears her husband is having an affair, so she goes to a gun shop and buys a handgun. The next day she comes home to find her husband in bed with a beautiful redhead. She grabs the gun and holds it to her own head. The husband jumps out of bed, begging and pleading with her not to shoot herself. Hysterically the blonde responds to the husband, ‘shut up … you’re next!’
Even the most powerful computer in the world will switch itself off when it hits the switch off instruction even it there are a million other instructions following it. Yet, even children would question an instruction to go to sleep which was followed by an instruction to do your homework. Computers simply process sequentially. Humans process concurrently and in abstract.
Now, as we know computers are fast. So fast that they can perform billions of instructions per second. We can also augment them so that they can be interrupted, like humans, and be given new instructions which can run in parallel giving the impression they are doing two things at once or processing concurrently but it is not so. It’s called time slicing or multitasking, if you prefer, and, unlike humans, computers are very good at it because they do not forget anything nor do they become fixated on one task to the detriment of another. On the other hand if one task needs more processing time to complete the computer will not adjust its timeslicing like a human unless it has been instructed to in its programming. In short a computer is only as smart as its programming team.
Now, when you show a face to a computer it scans it for several variables, proportions between the eyes and the chin, width of the head etc. It then searches a database one record at a time until it finds a match. Sure, it does this very quickly but the process is still the same. One record at a time. Now, it may pass this task on to several computers to speed this up, when dozens of computers are looking for the same thing in the data it finishes faster but it is still processing the same way. Read a record, check if it matches, display it if it does, and move on to the next record. If at the end then finish. They never forget, they always perform the search in the same way and if you fed the same face in as the next face to search for it would do it exactly the same and in the same time.
Now humans on the other hand. You show some one a face. Some abstract processing takes place we cannot replicate in software yet and we either recall the name, or we recognise the face but the name is forgotten and you have to think about it. We also do this differently we try and remember where or when we saw the face and hope that will be a trigger that will make us remember the name. Sometimes we just can’t pull the memory back but some time later the name pops into your mind. Show you that picture next and we immediately say it’s the same one. Now, some of this we can also improve on in the computer by programming them in, as an enhancement, but the computer would never do that on its own. That is the fundamental difference.
In addition if the face is viewed from an angle or is only seen in profile it will skew the readings. Humans automatically compensate and still recognise the person while it can fool the computer if it is not programmed in. Recognising text and numbers not in a regimented format is why Captcha works so well. Humans recognise the numbers and letters at any angles and even with lines through them but computers only can see what pixels are there or not. Of course soon spammers will be able to process them but it is a hurdle they can’t jump at the moment.
Computers also handle events differently, if you were putting wheel nuts on cars and a fire broke out next to you you would immediately stop what you were doing and attempt to put out the fire. If successful, you would then report the incident, it would be checked and something would change either something to stop it happening or tools to enable you to handle it better. With a computer the fire would be detected by a system set up for that purpose and the robot would continue on as if nothing happened. That detection system would trigger something, automated or human, to process the fire and it would be handled. It would be logged and someone would be tasked to checking the cause, fixing it and identifying any remedial action which a computer would be unable to do.
There are advantages and disadvantages to both methods. Obviously the computers are consistent, they never get bored, they never forget and they never need to stop for breaks. Humans on the other hand are adaptable, can think beyond what they have been taught and reprogram (teach) themselves safely. Computers can only handle what their programmers have thought of where humans can handle unexpected events.
This is why we can replace humans with robots for mundane repetitive tasks but when it comes to dealing with unknowns humans will be better every time. This is good news of course because it means that while production is automated and controlled by very fast computers humans can be innovating, creating art, exploring, or even relaxing and all the things we are good at and want to do.
Now programmers get smarter, they think of new ways to make computers look intelligent they make processors faster and develop ways to multitask better and automate the process but it still all boils down to sequential processing. However, systems improve, computers can handle new tasks and our lives are made easier.
So we can give the impression of AI for small subject areas like traffic management, logistics, behavioural prediction software, etc. For true AI, in the sense of intelligence as seen in the movies we actually need the computer system to be self programming across all areas. We are a long way away from that at the moment with computer processors and I don’t see us attaining it any time soon. However, let us remember that we are now looking at processing with real brain cells. Backward engineering and improving on this we may be able to create our first real brain. One that can be taught as we are and has the ability to adapt. Proving we can back that up at different stages we could end up with something we can use. However, on bad bit of programming or incorrect input to the system and we could damage the lot. Imagine if our AI soldier droid was scared of loud bangs because someone dropped something whilst it was being grown. We could end up with processors that don’t process correctly. Of course even our own brains don’t always process data correctly. That is how we end up with Socialists.
Personally though I think we are so far from a learning machine that could think and act like us that I see AI splitting into two branches. One where we use fast and parallel processing to perform tasks and interact as a robot with specific tasks in mind. We can create robots for all our mundane tasks and free up our time. Remember though I want robots that do a specific task, like clean the house, and nothing else. Not something that can do everything. This keeps the programming simpler and stops them taking over the world. The second branch is where we work on an artificial brain which we could use in dangerous situations to evaluate and resolve issues that humans would normally do. Eventually of course we would end up with something that would be, in reality, a new life form. Although I think we are a long way away from this I still actually wonder if there would be any real benefit of us ever getting there. I see it as another Pandora’s Box with significant implications for our species.
It’s one area where would could be too smart for our own good. Don’t forget KISS. A philosophy to live by.
AI, a step too far?
Being involved in computers and having spent some time on Decision Support Systems (DSS) I have managed thousands of man hours writing algorithms which would enable the computer to make predictions that would allow our company to maximise profits. I have been told that these ‘intelligent’ systems are the first stage of Artificial Intelligence (AI). I’m always arguing that we are nowhere near AI at the moment. I don’t even class the systems as intelligent. They just give the appearance of intelligence because they do things we generally cannot do easily.
You see human intelligence and what is currently called AI are actually so far apart they are not the same thing. Sure we can make it look like they are but that is the intelligence of the development team rather than the intelligence of the computer. Computers simply follow one instruction after another as programmed by the programmer. No more no less.
I’m sure you remember the blonde joke;
Even the most powerful computer in the world will switch itself off when it hits the switch off instruction even it there are a million other instructions following it. Yet, even children would question an instruction to go to sleep which was followed by an instruction to do your homework. Computers simply process sequentially. Humans process concurrently and in abstract.
Now, as we know computers are fast. So fast that they can perform billions of instructions per second. We can also augment them so that they can be interrupted, like humans, and be given new instructions which can run in parallel giving the impression they are doing two things at once or processing concurrently but it is not so. It’s called time slicing or multitasking, if you prefer, and, unlike humans, computers are very good at it because they do not forget anything nor do they become fixated on one task to the detriment of another. On the other hand if one task needs more processing time to complete the computer will not adjust its timeslicing like a human unless it has been instructed to in its programming. In short a computer is only as smart as its programming team.
Now, when you show a face to a computer it scans it for several variables, proportions between the eyes and the chin, width of the head etc. It then searches a database one record at a time until it finds a match. Sure, it does this very quickly but the process is still the same. One record at a time. Now, it may pass this task on to several computers to speed this up, when dozens of computers are looking for the same thing in the data it finishes faster but it is still processing the same way. Read a record, check if it matches, display it if it does, and move on to the next record. If at the end then finish. They never forget, they always perform the search in the same way and if you fed the same face in as the next face to search for it would do it exactly the same and in the same time.
Now humans on the other hand. You show some one a face. Some abstract processing takes place we cannot replicate in software yet and we either recall the name, or we recognise the face but the name is forgotten and you have to think about it. We also do this differently we try and remember where or when we saw the face and hope that will be a trigger that will make us remember the name. Sometimes we just can’t pull the memory back but some time later the name pops into your mind. Show you that picture next and we immediately say it’s the same one. Now, some of this we can also improve on in the computer by programming them in, as an enhancement, but the computer would never do that on its own. That is the fundamental difference.
In addition if the face is viewed from an angle or is only seen in profile it will skew the readings. Humans automatically compensate and still recognise the person while it can fool the computer if it is not programmed in. Recognising text and numbers not in a regimented format is why Captcha works so well. Humans recognise the numbers and letters at any angles and even with lines through them but computers only can see what pixels are there or not. Of course soon spammers will be able to process them but it is a hurdle they can’t jump at the moment.
Computers also handle events differently, if you were putting wheel nuts on cars and a fire broke out next to you you would immediately stop what you were doing and attempt to put out the fire. If successful, you would then report the incident, it would be checked and something would change either something to stop it happening or tools to enable you to handle it better. With a computer the fire would be detected by a system set up for that purpose and the robot would continue on as if nothing happened. That detection system would trigger something, automated or human, to process the fire and it would be handled. It would be logged and someone would be tasked to checking the cause, fixing it and identifying any remedial action which a computer would be unable to do.
There are advantages and disadvantages to both methods. Obviously the computers are consistent, they never get bored, they never forget and they never need to stop for breaks. Humans on the other hand are adaptable, can think beyond what they have been taught and reprogram (teach) themselves safely. Computers can only handle what their programmers have thought of where humans can handle unexpected events.
This is why we can replace humans with robots for mundane repetitive tasks but when it comes to dealing with unknowns humans will be better every time. This is good news of course because it means that while production is automated and controlled by very fast computers humans can be innovating, creating art, exploring, or even relaxing and all the things we are good at and want to do.
Now programmers get smarter, they think of new ways to make computers look intelligent they make processors faster and develop ways to multitask better and automate the process but it still all boils down to sequential processing. However, systems improve, computers can handle new tasks and our lives are made easier.
So we can give the impression of AI for small subject areas like traffic management, logistics, behavioural prediction software, etc. For true AI, in the sense of intelligence as seen in the movies we actually need the computer system to be self programming across all areas. We are a long way away from that at the moment with computer processors and I don’t see us attaining it any time soon. However, let us remember that we are now looking at processing with real brain cells. Backward engineering and improving on this we may be able to create our first real brain. One that can be taught as we are and has the ability to adapt. Proving we can back that up at different stages we could end up with something we can use. However, on bad bit of programming or incorrect input to the system and we could damage the lot. Imagine if our AI soldier droid was scared of loud bangs because someone dropped something whilst it was being grown. We could end up with processors that don’t process correctly. Of course even our own brains don’t always process data correctly. That is how we end up with Socialists.
Personally though I think we are so far from a learning machine that could think and act like us that I see AI splitting into two branches. One where we use fast and parallel processing to perform tasks and interact as a robot with specific tasks in mind. We can create robots for all our mundane tasks and free up our time. Remember though I want robots that do a specific task, like clean the house, and nothing else. Not something that can do everything. This keeps the programming simpler and stops them taking over the world. The second branch is where we work on an artificial brain which we could use in dangerous situations to evaluate and resolve issues that humans would normally do. Eventually of course we would end up with something that would be, in reality, a new life form. Although I think we are a long way away from this I still actually wonder if there would be any real benefit of us ever getting there. I see it as another Pandora’s Box with significant implications for our species.
It’s one area where would could be too smart for our own good. Don’t forget KISS. A philosophy to live by.