Just another example of technology for technologies sake.

There is no way that it can, in a vehicle, at the moment have access to, and process all the information that we humans do, to come to a decision in a fraction of a second on a particular course of action when facing a threatening situation on the road.

Automatic steering??? what is this lunacy, on today's roads it just cannot work, for the car to avoid a head on collision it will making a decision to take over the steering well before impact, and who knows what it will steer into, and as BNM said braking is not always the best option under a given set of circumstances. And if you are heading for a head on collision then someone is...

Just another example of technology for technologies sake.

There is no way that it can, in a vehicle, at the moment have access to, and process all the information that we humans do, to come to a decision in a fraction of a second on a particular course of action when facing a threatening situation on the road.

Automatic steering??? what is this lunacy, on today's roads it just cannot work, for the car to avoid a head on collision it will making a decision to take over the steering well before impact, and who knows what it will steer into, and as BNM said braking is not always the best option under a given set of circumstances. And if you are heading for a head on collision then someone is not in the correct place, so where was the tecno to avoid this in the first place???

I work with vehicles, big ones, and i can tell you that technology goes to far...

This sounds great! Just one more step towards an automated system.

If the system is monitoring collisions then it is doubtful that, like a human, it will reflexively react to avoid a collision only to turn the wheel into another. While the human brain is far more complex than todays computers they are not faster than the simple calculating speed of computers. As predicted, computers have double in speed on average of every 4 years. They are quite capable of monitoring road conditions and reacting to them much for quickly than any human could.

Today we can completely elliminate human control from transport simply by imbedding chips in the roads and adding a tracking system into each car t...

This sounds great! Just one more step towards an automated system.

If the system is monitoring collisions then it is doubtful that, like a human, it will reflexively react to avoid a collision only to turn the wheel into another. While the human brain is far more complex than todays computers they are not faster than the simple calculating speed of computers. As predicted, computers have double in speed on average of every 4 years. They are quite capable of monitoring road conditions and reacting to them much for quickly than any human could.

Today we can completely elliminate human control from transport simply by imbedding chips in the roads and adding a tracking system into each car that ties into satelite positioning and alignment by sensors. Of course the cost of such a thing makes this advance costly to alter every road and every car out there. But as the newer cars are made and sold the foundation for such a system becomes more and more feasable.

I think the trade off of a few malfunctioning systems crashing vs the hundreds of people hurt and killed in car accidents EVERY MINUTE is quite worthwhile.

The problem is that unlike humans computers do not have the capability of thinking, they are mere machines that do what they are programmed to do, and they do not have at the moment the capability of taking in all the information and reasoning the best course of action, they will simply do what they are told to do under a given set of circumstances, and that would imo be in many cases inappropriate action.

On the point of reliability, i can tell you from experience in dealing with this technology, that it does create (at the moment) more problems than it claims to cure, and when these problems do occur there would seem to be no-one who is able (on ocaision) (including the manufacturers) ...

The problem is that unlike humans computers do not have the capability of thinking, they are mere machines that do what they are programmed to do, and they do not have at the moment the capability of taking in all the information and reasoning the best course of action, they will simply do what they are told to do under a given set of circumstances, and that would imo be in many cases inappropriate action.

On the point of reliability, i can tell you from experience in dealing with this technology, that it does create (at the moment) more problems than it claims to cure, and when these problems do occur there would seem to be no-one who is able (on ocaision) (including the manufacturers) to give an adequate answer as to where the fault lies in the techno systems that are being incorporated into vehicles.

The correct approach to the problem of human error would be to incorporate the warning aspect of technological systems on vehicles rather than a corrective approach.

For example it would be relatively easy for a vehicle to simply warn that you are too close to another, or indeed that you are heading for a front end collision.

People need to be alert when driving and not be lulled into a false sense of security.

Which is my exact point. People are not only fallible, but irresponsible. You guys are talking like the machine has to think like a human brain, which it certainly does not. It need only avoid collision. Unlike a person who simply reacts, an automated system would be continually monitoring its surroundings. There is no avoiding a collision only to steer into another one (which a human mind is more apt to do because of reflexive actions). Instead a system which monitors collisions would steer clear of an oncoming car and into a direction that does not collide with anything else. It will also...

People need to be alert when driving and not be lulled into a false sense of security.

Which is my exact point. People are not only fallible, but irresponsible. You guys are talking like the machine has to think like a human brain, which it certainly does not. It need only avoid collision. Unlike a person who simply reacts, an automated system would be continually monitoring its surroundings. There is no avoiding a collision only to steer into another one (which a human mind is more apt to do because of reflexive actions). Instead a system which monitors collisions would steer clear of an oncoming car and into a direction that does not collide with anything else. It will also apply the breaks in such a way as to halt the car as efficiently and quickly as possible to avoid skidding.

I hope that you can see that you reaction here is only a fear to new technology. I suspect that if you examine history you'll see the same arguments about cars that traveled faster than horses and how moving so fast was simply unnatural. Or the old addage "If man was meant to fly ...".

114 people DIE on average per day from auto accidents. That translates to almost 5 per hour or 1 person dying in a car crash every 12 minutes. How many people are only INJURED each second? The numbers are staggering. How many accidents happen each night in your area locally?

The majority of accidents are driver negligence or recklessness. A computer's attention never wanders and it is always scanning its surroundings. It isn't trying to show off, it isn't drunk, and it certainly never grows tired or bored.

When a human reacts to an accident it is generally jerking the wheel, slamming the breaks and stiffening their bodies. An automated system would not only react to an imminent collision but it monitoring the cars action through the entire process and controlling the car every second of the way, altering course, applying various levels of breaking, and seeking the safest direction of tr

Here's the problem: Computers are created by humans, so it is extremely unlikely that they will ever surpass us. You can't make a 10 foot box out of 5 feet of cardboard.

The greatest limitation of the computer is the digital nature of their programing. This makes them inherently finite, only capable of analyzing things in terms of dichotomies. On the other hand, analog systems work on an infinite gradient and thus can analyze organic conditions.

Without a fundamental change in the way computers are programed and even constructed, they will continue to be unable to analyze complex situations with ill defined parameters.

That is to say, a computer can only take into account the fixed elemen...

Here's the problem: Computers are created by humans, so it is extremely unlikely that they will ever surpass us. You can't make a 10 foot box out of 5 feet of cardboard.

The greatest limitation of the computer is the digital nature of their programing. This makes them inherently finite, only capable of analyzing things in terms of dichotomies. On the other hand, analog systems work on an infinite gradient and thus can analyze organic conditions.

Without a fundamental change in the way computers are programed and even constructed, they will continue to be unable to analyze complex situations with ill defined parameters.

That is to say, a computer can only take into account the fixed elements of a situation, but not the fluid or indistinct aspects. Humans can make judgment calls without information. Although this means we occasionally invade countries that don't have WMDs, it also means we can make split second decisions while driving without doing millions of calculations.

This is also why a computer can beat a human at chess but not at weiqi (Go). In Chess there are very strict, fixed parameters, and only a few million possible moves. But a weiqi board has 324 places to play, and begins with an empty board. This is too much to handle for a computer, which can't make the first move, and plays poorly until the game is past the half way point and the options become more calculable.

It is whimsy that makes us bad drivers, but it is also whimsy that makes us capable of driving at all.

@ ouyangwulong

I think your understanding of current computer technology is a bit off. There are breakthroughs being made every month. The biggest drawback of computers to mimic human thinking in the past was the in series method of calculations used. In other words it was a listed order of commands that had to be followed one after the other. This meant very fast computers would be needed to even come close to mimicking even the simplest of living things. But todays computers are being designed with true multiprocessing abilities, and neuronet understanding grows daily.

To say that computers will never surpass the human mind is very naive, to say the least. It was once said that computers...

@ ouyangwulong

I think your understanding of current computer technology is a bit off. There are breakthroughs being made every month. The biggest drawback of computers to mimic human thinking in the past was the in series method of calculations used. In other words it was a listed order of commands that had to be followed one after the other. This meant very fast computers would be needed to even come close to mimicking even the simplest of living things. But todays computers are being designed with true multiprocessing abilities, and neuronet understanding grows daily.

To say that computers will never surpass the human mind is very naive, to say the least. It was once said that computers would never prove to have any public value because the smallest ones would still be the size of cars and only be affordable by businesses or government. Many people, even developers, could never have conceived what we use computers to do today. They already park cars!

Computer technology has vastly increased even in the last few years. They can identify patterns and objects even more effectively than that of the human eye. They recognize emotions and can respond to how they are treated. They can learn through experience and change their own programs in order to overcome obstacles and challenges.

I guess that it really surprises me that anyone could say computers or robots could never surpass humans.

I should clarify, I'm not saying they could never surpass human. I'm saying I don't believe that a digital system will surpass the human brain. What is more, although I would certainly conceed that computers, or something like them, may one day surpass us, I would hardly say it with the same certainty that you have. After all, as often as technology has surpassed expectations, it has also fallen short. There are some things that computers do exceedingly well, and some that they don't. Technology is not a pancea, nor is anything else for that matter.

I'm also aware of multi-processing abilities, but that's not what I'm talking about. You used one very key word here: mimic. The computers can ...

I should clarify, I'm not saying they could never surpass human. I'm saying I don't believe that a digital system will surpass the human brain. What is more, although I would certainly conceed that computers, or something like them, may one day surpass us, I would hardly say it with the same certainty that you have. After all, as often as technology has surpassed expectations, it has also fallen short. There are some things that computers do exceedingly well, and some that they don't. Technology is not a pancea, nor is anything else for that matter.

I'm also aware of multi-processing abilities, but that's not what I'm talking about. You used one very key word here: mimic. The computers can model pre-designated behavior, and indeed can do it very well in some cases. This is cool, but if we set asside our awe for a little bit, there are some serrious practical limitations, the most important being the necessity of firm definition. Computers, unlike people, can't deal with ambiguity beyond certain parameters, and uncertainty can't be resolved with out a programmer's input.

If you are involved in some highly secret project working with technology that won't hit for the market for 20 years, then perhaps you can shed some light on how a computer might overcome uncertainty without human input. Otherwise, if you work in programing and have a little spare time, try out this challenge:

Write a program that can play the game Go (wei qi) starting from an empty board. Getting it to play the first 10 moves is extreemly dificult, because the computer essentially can't make subjective decisions. The normal solution is to program to mimic the subjective decisions of a human role-model, but that's not the same as thinking for itself.

If you write a program that can beat even the lowest rank professional player, there is something like a 2,000,000 British Pound prize that has yet to be claimed.

My concern is that driving conditions would present a similar problem for computers.

Oh, and on the other hand, it occurs to me that I should be clear: I'm not talking about the feasibility of a fully automated system. What I'm doing is exploring a conceptual area in the development and limitations of current computers. My goal is, of course, to think up a solution to surpass these limitations. And I'm sure we can both agree that computers as they are have certain finite limitations right now, even if we haven't settled exactly what those are.

On the other hand, the question of a fully automated system, as I see it, is totally different. My concern is not that computers are incapable of driving a car, but that the attempt to combine human driving with selective computer dri...

Oh, and on the other hand, it occurs to me that I should be clear: I'm not talking about the feasibility of a fully automated system. What I'm doing is exploring a conceptual area in the development and limitations of current computers. My goal is, of course, to think up a solution to surpass these limitations. And I'm sure we can both agree that computers as they are have certain finite limitations right now, even if we haven't settled exactly what those are.

On the other hand, the question of a fully automated system, as I see it, is totally different. My concern is not that computers are incapable of driving a car, but that the attempt to combine human driving with selective computer driving is dangerous.

In a fully automated system, then ideally, a single controller is operating all the cars in a vicinity. This provides the finite limitations to the situation necessary for computers to efficiently make good decisions. Parameters can be easily programmed, and the whole thing can go into a fail-safe mode if one of those parameters is exceeded. This grounds the situation in the realm of tangible numbers and finite calculations.

On the other hand, when having to interface between a digital system that sometimes operates the car, and an organic system (you) that operates it the rest of the time, there are huge gaps in the information you are using to drive, and the information that the car has. The largest problem, as I see it, is that a computer will never be able to efficiently predict what human driver might do.

As I see it we need to keep it all human, or go all automated. Trying out somewhere in between (like Volvo is doing) seems to be asking for trouble.

Also, my original concern is not that the software is impossible to create, but that it would be really really inconvenient to have to work out bugs in the system post release, given it can control the direction and velocity of the car.

As a company, just imagine the risk of ridiculous law suits.

Ben Goertzel is an interesting guy, for sure, and as far as I'm concerned, a step in the right direction. I'm going to try finding an article of his to clip so that the philosophical question of artificial inteligence can be separated from the practical questions of Volvo's countermeasures.

First, and foremost, though, Goertzel isn't talking about the same thing as Vovlo is. He's talking about Artificial General Intelligence, meaning a computer that is self-aware and understands its surroundings in relation to itself. This is different from task-oriented "narrow AI" which is what Vovlo is trying to use to drive cars.

What is more, work on AGI is highly diversified, in non-digital fields. ...

Ben Goertzel is an interesting guy, for sure, and as far as I'm concerned, a step in the right direction. I'm going to try finding an article of his to clip so that the philosophical question of artificial inteligence can be separated from the practical questions of Volvo's countermeasures.

First, and foremost, though, Goertzel isn't talking about the same thing as Vovlo is. He's talking about Artificial General Intelligence, meaning a computer that is self-aware and understands its surroundings in relation to itself. This is different from task-oriented "narrow AI" which is what Vovlo is trying to use to drive cars.

What is more, work on AGI is highly diversified, in non-digital fields. For an outlandish introduction to how broad they cast their net, just watch the chemist Eamonn Healey's monologue in the Richard Linklater film Waking Life. These people are talking about integrating the organic directly with the digital, or even creating organic "computers."

The later is what I would favor as the most likely bet, because I believe that as we have constructed our hardware today, we have already surpassed the human brain in some aspects, but in others we have reached a dead end. Further progress, I believe, will require a total paradigm shift.

My central criticism of Goertzel and others is that they have become enammored with similarity, while overlooking the persistent differences. They have focused on how to make a computer approximate a brain, but only in result, not in process. Quantum computing and other theoretical technologies are good examples of this - trying to accelerate the computing systems to the point where they can produce human-like results, but using a fundamentally diferent system. To my mind, this is like trying to use a shoe as a glove. It might keep your hand warm, but it doesn't work the same way.

I am coming at this from a philosophical perspective, instead of a mechanical one. (A major difference between me and most of the people working on AI). Unlike most people taking the philosophical approach, I believe AGI is possible. However, unlike most people taking the mechanical approach, I believe that the fundamental structure of our current computing hardware constiutes a significant limitation (or "bottleneck" as Goertzel calls it) in the development of AGI.

I also don't believe in a technological singularity like Ray Kurzwiel. Although a "narrow" singularity might be possible, a general singularity won't, because as the evolution of technology accelerates, that will simply focus it...

I am coming at this from a philosophical perspective, instead of a mechanical one. (A major difference between me and most of the people working on AI). Unlike most people taking the philosophical approach, I believe AGI is possible. However, unlike most people taking the mechanical approach, I believe that the fundamental structure of our current computing hardware constiutes a significant limitation (or "bottleneck" as Goertzel calls it) in the development of AGI.

I also don't believe in a technological singularity like Ray Kurzwiel. Although a "narrow" singularity might be possible, a general singularity won't, because as the evolution of technology accelerates, that will simply focus it down to a smaller and smaller point, culminating in a singularity where all further development possible takes place in a infinitely small ammount of time. However, that is simply the stagnation of a single branch of technology, and I believe new branches to persue will be discovered once they are no longer obscured by our current limitations.

But even more, I think that we shouldn't confuse what AGI is and can be. AGI can be an extreemly advanced and self-sufficent computer, which is aware of itself, and can manage itself with minimal human supervision. It can even exceed a human brain in given situations, though I see no logic to suggest it will necessarily by universally superior.

The larger problem is that their is a difference between self-awareness and volition. Truthfully, nobody really understands volition these days. People who try to mathematically model the brain over look this key problem: free will. In philosophy some people try to slide by this question by applying Karl Popper's interpretation of Propensity to salvage Probablistic Determinism, but this is not substantially connected to reality, merely one of many plausible explanations. As Popper would be the first to point out, it is utter pseudo-science because it is not verifiable through a testable antithesis

Although I hate to fall back on it, because I think he is too narrow as he tries to refute AI, we shouldn't forget John Searle's "Chinese Room" argument differentiating AI from organic intelligence. I particularly like this example because I speak Chinese, and in teaching Chinese people to speak English, I run into this exact problem quite often.

The argument goes like this. Suppose you have a man, let's say he's Hungarian. Nothing against Magyars, but we need to start somewhere. This Hungarian speaks neither English nor Chinese. However, he is placed in a room full of books. In the books there are lists of every Chinese character ever used, and next to it is a single English word that pres...

Although I hate to fall back on it, because I think he is too narrow as he tries to refute AI, we shouldn't forget John Searle's "Chinese Room" argument differentiating AI from organic intelligence. I particularly like this example because I speak Chinese, and in teaching Chinese people to speak English, I run into this exact problem quite often.

The argument goes like this. Suppose you have a man, let's say he's Hungarian. Nothing against Magyars, but we need to start somewhere. This Hungarian speaks neither English nor Chinese. However, he is placed in a room full of books. In the books there are lists of every Chinese character ever used, and next to it is a single English word that presumably represents that character exactly.

Then, through a slot in one door, a note slides in with a Chinese character on it. The man then looks at the books full of gibberish until he finds it, writes a new word with the unintelligible English counterpart, and then passes the English word back out the door, although he speaks neither English nor Chinese, he appears to have translated between the two languages.

Searle argues that the man is like the computer, and although it can process the information, simulating the translation, it is not self-aware. Goertzel is trying to solve this problem, but I see there as being a different problem inherent in this model.

I differ from Searle, primarily in that I speak both Chinese and English. Thus I have noticed certain key aspects that make this computer model fundamentally different from natural intelligence.

1. Definition

First is the idea that the mind understands and stores knowledge in terms of definitions. This is not in fact the case. If you ask a person to define any given word, even though they use it regularly, they are often at a loss for words. Why is this? Because unlike current computers, the brain does not function based on defintions of concepts. Instead, it would appear to store knowledge as a series of creative webs. You can take advantage of this to rapidly acquire knowledge and for...

I differ from Searle, primarily in that I speak both Chinese and English. Thus I have noticed certain key aspects that make this computer model fundamentally different from natural intelligence.

1. Definition

First is the idea that the mind understands and stores knowledge in terms of definitions. This is not in fact the case. If you ask a person to define any given word, even though they use it regularly, they are often at a loss for words. Why is this? Because unlike current computers, the brain does not function based on defintions of concepts. Instead, it would appear to store knowledge as a series of creative webs. You can take advantage of this to rapidly acquire knowledge and foreign languages. Abailart just had an interesting clip on this.

2. Correspondence

The other key difference is the idea of correspondence, that one mental concept corresponds singularly to another tangible concept. This is not necessarily the case. Take the languages for example. In Chinese, there is no word for "off." The word is simply replaced by different words depending on the context, such as "xia" (approximately down) and "guan" (approximately close). I say approximately because if we were to draw a vin diagram, we would see that the Chinese word and English word, although overlapping in meaning and usage, are not the same.

What is more, our mind does not seem to store them as equivalences, but as unique words in a single mental language the encompasses all the spoken languages one has learned.

The result of overlooking these two problems is the often hilarious Chinglish, but it illustrates exactly what happens when you take a digital system that relies on concrete definitions and equivalences, and apply it to an organic system with subjective definitions and ambiguous generalities.