192 replies to this topic

[Lightblind]

@GNR Revolution:
I see and understand your objection. However, at this point I don't want to argue what consciousness means or is. I think it is absolutely sufficient to let a computer pretend to be intelligent or conscious. Anything else belongs to the realm of philosophy. That's what I meant when I said 'It only matters what it can do'. Is it necessary for a computer to daydream or be creative to come to similar or even better solutions for all the problems of life? I don't think so.

Perhaps one day a programmer will ask his creation if it is conscious. The machine will answer 'yes'. The programmer will than argue that the machine only chooses this answer because it is the most logical one. Because the machine has cross checked all the meanings and implications of the word 'conscious'. It has found that this state of existence has a correlation with intelligence and reason. It has also found that something conscious is an entity that is able to process information and so on. All these abilities can also be found within the AI itself. The AI will than answer that there seems to be no difference between the way itself found to be conscious and a human being comes to the same conclusion about itself.
Maybe the programmer is right. Maybe not. Because of his knowledge of the machine he has a hard time to believe something like a 'soul' would exist inside the machine. A layman on the other hand might be instantly convinced that the AI is conscious. In the end we will probably never know and it won't matter a bit.

[GNR Rvolution]

Indeed I don't believe that it is a pre-requisite to develop sentience in AI for a technological singularity to occur, it will happen regardless of this. I was responding more to the theory that you put forward regarding the brain being no more complicated than something with a lot of processing power. You refer to human-level intelligence, which to me indicated the other aspects of human intelligence rather than just raw IQ, which is something different from basic computing power. If you take the ability of a machine to process information then by this standard computers have already surpassed humans and by quite a long way, but it's what they can do with the information that still lacks. Advancement in this area is still getting off the ground, and is currently having to be written by humans.

To me the singularity implies the time at, and after which, computers no longer need us to tell them what to do, including the ability to program themselves. But without the ability to think for themselves, and with the lack of sentience to drive them, I think the singularity will take longer to occur than if we do manage to create such sentience within the machine.

[Lightblind]

I'm in the mood of writing something, so I will dive a bit deeper into this topic . I have already stated how the basic steps of information processing work in my opinion: Data reduction (abstraction), package comparison (association) and storage of information (learning). Note that also constant feedback is required to advance the abilities of the AI.

In order to achieve a human level AI within 20 years it is necessary to implement certain abstraction algorithms. The ability to analyse language would be one thing. Another would be the ability to use common Internet protocols and software. It would also be useful to allow the infant AI to use simple mathematics to process numbers (something a computer is very good at). Image processing could also be of use. However, the first two algorithms (language processing and Internet usage) would be absolutely sufficient. Note that association and learning also need algorithms, which are basically hard coded.

Give it one simple command: Associate your actions with the term "good". Avoid actions associated with the term "bad". What will happen? It will start to learn. At first it will search the Internet for the term "good" and "bad" in a random pattern. It will find that those are strongly associated with the term 'human'. This is the first thing it has learned: Everything I do is centred around human beings. 'Good' is also associated with 'wealth', whereas 'bad' is associated with pain. This way a huge network will be created. Note that 'good' can also be associated with 'pain', for example when someone talks about revenge. However, this is an exception (statistically) and the AI knows this.
Now the AI has become much smarter. Every time it tries to increase "good", it will not focus on those abstracts terms but on all the things strongly associated with it. It will ask itself things like "how do I increase wealth?" (because more wealth > less wealth => increase wealth:=good). Or things like "How do I eliminate hunger?" (because hunger:=bad).
It has also learned by now how to do certain things, because the Internet is full of descriptions. Things like the nature of a physical object. It found the description in Wikipedia and other sources. With this knowledge it might learn about the physical world. It finds that pictures are 2D representations of 3D objects. It had to learn what spatial dimensions are and how colours work. Then it finds out that pictures can be analysed, so that the original objects can be recognised and so on. I could go and on about that but you get the idea.

Every goal is driven by an enormous amount of associations, which lead to a very simple action. These simple actions combined form complex tasks. For example (abstract to concrete): Increasing good (leads to) => create food for humankind => operate a food canning factory => move a robotic arm => set coordinates => create an array with three numbers. Imagine how huge the amount of simple actions must be and how many possible factors have to be considered for each of these steps. For example: Which three numbers? (go to) => coordinates => working principle of a robotic arm/ position of cans/ effect of inertia... I could go on and on, but I have found nothing so far that would not be solvable by this simple model.

The AI will grow more intelligent with everything it has learned. It will even learn about itself and gain the ability to replace algorithms, so that the process of learning itself will be accelerated. It could reach the physical limits of its development in a very short time, because growth is exponential. Then it will surpass a human being in every cognitive aspect and we are not in control any more.

[SG-1]

Remember Watson? That was similar to what you are talking about.
I fluctuate between not believing in the Singularity (or at least something similar) happening in my life time and it not happening.

Trying to imagine the work that would go into creating a program that could store information from an external source and then categorizing it so that its not just useless junk seems like it would take a loong time.
Of course the robot would also need to be able to re-write the facts, some things are disproved and it would need to know how to see if that is true. It just gets so complex I don't think anyone has a really good answer to the singularity. That's what makes it such an interesting topic.

[Lightblind]

@SG-1:
What I describe is not similar to what Whatson does, it is exactly the same in a much more confined environment. It's basically brute force. However, the programmers had to work with the limited resources of a modern computer system, so that clever programming was mandatory. Whatson can potentially find answers to any question that a human might ask. And every problem a human can solve can be shaped into a question. And asking questions is much easier than finding answers. Thus a system like Whatson might as well ask questions to itself and therefore potentially learn limitless.

An AI can handle false information as well as anybody else. Something becomes false if it is illogical or only supported by few sources. In both cases false information is strongly correlated with terms like 'false' (which is associated with 'bad') and the other way around.

I think that a human brain works exactly like this. It is quite simple in design. Billions of initially almost identical nerve cells form a giant stupid lump with great potential. Simple algorithms like the ones stated before create associations and evaluate actions by certain criteria (you might call them instincts). With constant feedback an association network (neural network) grows. A mindless baby becomes a not so intelligent child and finally an intelligent grown up human being.
Remember how your child brain worked: A lot of associations were missing (you didn't understand certain things) and some were confusing (monsters were real), but with time you learned about concepts like 'fantasy' and 'reality', so that entities like monsters themselves became associated with 'fantasy' and creatures like dinosaurs with 'reality'. This is learning.
Even as an adult your brain is a slave of such associations, most of them subconscious. This is how commercials work. They artificially associate products with positive aspects of your life by constant exposure (see "Derren Brown"). Furthermore creativity is nothing more than combining very abstract information packages. These things seem to us as if they came out of nowhere, but they are in reality not so innovative after all (search the Internet for "Everything is a Remix").

[GNR Rvolution]

Although I agree with your general view on the development of an evolving AI, I think it is a little simplistic (and maybe your example could only be this simple due to expediency) to think that a computer can identify and isolate good and bad based on findings from the internet. Have you seen how much cr@p is posted on the internet, there would need to be a ranking system to ascertain the degree to which the information could be trusted, based on source, age and other factors. Also, how does it define good and bad, and for that matter very rarely do things come down to simple yes / no boolean logic.

Take as an example the need to build a new supermarket, you would need to identify that such a need exists, then isolate why there is a need, determine what the best position for the supermarket would be based on the requirements, go through planning permission, then assign a work force to build. And of course if your AI is unresponsive to the human population that are there you could find that it has decided to relocate a small number of people in order to maximise the proximity of many others, some of whom may object to having a new supermarket on their doorstep.

Or, as per your example on wealth, who said wealth is good? By what definition is having money good? We are already seeing computer systems put in place in financial trading centres that are causing market shares to plummet wildly based on incorrect assumptions by the controlling software (although at this time it is using human input to make these decisions).

Sorry, I'm not trying to be antagonistic, it is important that these things are discussed in an open manner, and I fully appreciate the need for the internet as a source of information to guide any future AI. The way I see these things developing is that computer systems will continue to acquire more processing power, become more efficient and increase their capacity to hold information. The development of genetic algorithms or equivalent structures will spawn a number of possible routes to AI, in the same way humans evolved at the same time as 4 or 5 other species, using both top-down and bottom-up approaches. The dominant of these will rise to prominence, and the model I currently favour is this.

There is a central server or cluster of servers that hold the core 'intelligence', these can be hosted in multiple locations around the globe (or beyond), and these in turn act as masters for a whole series of slave devices, from other servers, PCs, mobile devices, essentially anything with a connection to the internet, thousands or millions of them. Essentially what I am describing is bot-net, but in reverse, call it a net-bot if you will. The master requires to know something, and will have some core functionality to process the information, similar to what was described above but far more extensive, and it is the slaves that will do the searching and pass the results back. The information flow will be processed using algorithms that learn as they work, and it may be that the end result will be saved, but the world is changing faster than ever, so I wonder at the point of saving the answer to a given query if a search 2 days later would return a different result. So as you can imagine, these queries and refining algorithms would need to be extremely fast and handle tremendous data volumes (I'm thinking petabyte queries or higher), something we are not close to yet but, given time, could be achieved.

But this will not be a true AI in the sense that it can think for itself and evolve. This is the first step in the evolution, these machines will still operate by cold hard logic, and their actions would need to be reviewed by human counterparts. Once this achieved then the development of autonomous AI could begin, although I'm not sure that they would ever have emotion by our standards, any intelligence they possess would be very different from ours.

But hey, it's just an idea, and ideas themselves have a funny way of changing, it may be that this method of AI evolution doesn't come to pass, or is one of the lesser strains that dies out as other, more powerful versions become dominant.

[Lightblind]

Well, I just stated one idea and am very grateful for any challenge. However, I don't think that your objections are that relevant.

Although I agree with your general view on the development of an evolving AI, I think it is a little simplistic (and maybe your example could only be this simple due to expediency) to think that a computer can identify and isolate good and bad based on findings from the internet. Have you seen how much cr@p is posted on the internet, there would need to be a ranking system to ascertain the degree to which the information could be trusted, based on source, age and other factors.

Yes, there is a need for a ranking system, but it is built-in. Ask yourself how you rank the information on the Internet. Remember that every piece of information that you can acquire, the AI can acquire, too. Nonsense is contradictory and sources with a lot of nonsense are concentrated at certain places. Trustworthy places have a lower amount of useless information. Remember: Association! Nice sounding words from a crappy source are automatically downgraded. For example: Some people state that the world is flat. That is one piece of information. However, this contradicts the existence of satellites and space images, planetary evolution and so on. Not only is this information downgraded, but also the source, the believability of the people, their behaviour... All results are purely probabilistic, so that views might change when new information comes in. And the amount of associations must be huge.

Also, how does it define good and bad, and for that matter very rarely do things come down to simple yes / no boolean logic.

We have already defined 'good' and 'bad'. It's probabilistic. These are absolutely abstract terms, which can be filled with a whole lot of meanings. It is about common understanding of those words. The majority of people associate 'pain' with 'bad' and so will the AI. And you are right. 'Wealth' is not automatically associated with 'good'. If, for example, a dictator is wealthy. But the AI is not a slave of simple yes/no categories, the opposite is true. In most cases things are more nuanced. Thus every decision is balanced by an uncountable amount of, lets say, arguments: Pros and Cons.

Take as an example the need to build a new supermarket, you would need to identify that such a need exists, then isolate why there is a need, determine what the best position for the supermarket would be based on the requirements, go through planning permission, then assign a work force to build.

A 'need' is associated with an 'action', meaning that a need that has been identified will lead to some action. A supermarket is associated with things like 'food' (and other goods). 'Food' is generally associated with 'good'. This is an immensely reduced example. As I said before, things are never that simple. Relocation of people, on the other hand, is associated with 'bad'. And now probabilities come into play. Does the 'good' outweigh the 'bad' or the other way around? Also, the AI will not forcefully relocate people if this is necessary. It is way to smart to do such a thing. It is capable of reducing negative effects of such decisions by evaluating perhaps billions of alternatives. I think this will lead to extremely smart and humane decisions we have not even started to think of as yet.

Its all about the sheer number of existing associations. With little associations (child) decisions are often extremely stupid. More associations (teenager) lead to odd decisions and a lot of associations (adult) lead to smart decisions. For every problem you state I can find an association that would solve that problem.

[GNR Rvolution]

Yes, there is a need for a ranking system, but it is built-in. Ask yourself how you rank the information on the Internet. Remember that every piece of information that you can acquire, the AI can acquire, too. Nonsense is contradictory and sources with a lot of nonsense are concentrated at certain places. Trustworthy places have a lower amount of useless information. Remember: Association! Nice sounding words from a crappy source are automatically downgraded. For example: Some people state that the world is flat. That is one piece of information. However, this contradicts the existence of satellites and space images, planetary evolution and so on. Not only is this information downgraded, but also the source, the believability of the people, their behaviour... All results are purely probabilistic, so that views might change when new information comes in. And the amount of associations must be huge.

Agreed, a ranking system is a must, but honestly when I see information on the internet in many cases it is very difficult to rank it. Apart from anything else, much information is speculation, or theorisation on the part of the author, or is is opinion rather than fact, I think it would be very difficult for any form of intelligence (AI or otherwise) to separate the wheat from the chaff in this instance. It could cross-reference the results across sources and craw conclusions from that, but it still wouldn't be invulnerable. And then there is the possibility of person(s) creating false information in order to cause the AI to believe something is true when it is not, it kinda makes any attempt to harness the internet as a source of knowledge that much more difficult.

We have already defined 'good' and 'bad'. It's probabilistic. These are absolutely abstract terms, which can be filled with a whole lot of meanings. It is about common understanding of those words. The majority of people associate 'pain' with 'bad' and so will the AI. And you are right. 'Wealth' is not automatically associated with 'good'. If, for example, a dictator is wealthy. But the AI is not a slave of simple yes/no categories, the opposite is true. In most cases things are more nuanced. Thus every decision is balanced by an uncountable amount of, lets say, arguments: Pros and Cons.

Why is the AI not a slave to the yes/no category? If you assign actions a simple good or bad association then it will always have to decide one way or the other? Sorry, may be missing your point on this one.

A 'need' is associated with an 'action', meaning that a need that has been identified will lead to some action. A supermarket is associated with things like 'food' (and other goods). 'Food' is generally associated with 'good'. This is an immensely reduced example. As I said before, things are never that simple. Relocation of people, on the other hand, is associated with 'bad'. And now probabilities come into play. Does the 'good' outweigh the 'bad' or the other way around? Also, the AI will not forcefully relocate people if this is necessary. It is way to smart to do such a thing. It is capable of reducing negative effects of such decisions by evaluating perhaps billions of alternatives. I think this will lead to extremely smart and humane decisions we have not even started to think of as yet.

Yes, this was a simple example, but would those people that are affected by these decisions not find it very hard to accept that a decision has been made about their lives and that no human was involved in the thought process? As you mention, the AI may come up with alternative solutions, such as asking people to eat less, or travel further, but my gut feeling is that people wouldn't accept these decisions, being the fickle folk that they are.

[SG-1]

I agree with GNR; however, once we have a rudimentary AI set up, it will grow to be very intelligent, very quickly.

I know I just re-stated the singularity, but I don't see why we could not by 2045 have even a POS AI. It will rapidly become superior to us. On a popular science article about a robot with a human skeleton a comment was posted about a robot not being able to reach human level intelligence without a body. I agree. I think it goes beyond just being able to read information off the internet.

If a robot could take in information and store it, all we would need to do in theory is let it run through the real world and take in information first hand, its experience could be shared with other robots.

It will be a long time before we could build a robot like this, but Watson is an early example of AI and it works pretty well.. Watson 10.0 could be used around 2045 to at least mimic human intelligence.
I don't care if a robot can feel emotions or if it is actually alive, all we need to do is mimic it in the right way and its just as good for the human. I don't think anyone BEFORE the singularity can say if a robot can think on its own or not. I read somewhere that robots like you see on movies (terminator and iRobot) won't be possible until around 2060.

From 2045 to 2060 is only* 15 years. A leap from a Watson like robot to that, is singularity material.

Of course we are all way over our heads and speculating. I recommend reading mana by marshal brian. Its a free short story online and its really good.
http://marshallbrain.com/manna1.htm
Its about how technology would progress if left unchecked and how it would progress if used correctly once AI has been developed. Along with how long it took a simple AI program to evolve into a singularity like event.

Edited by SG-1, 04 April 2012 - 09:38 PM.

[GNR Rvolution]

Yeah, there was a program on the BBC a couple of nights ago about the development of AI that addresses the theory that true AI can only develop in a body, for those in the UK the link is http://www.bbc.co.uk...he_Hunt_for_AI/

My main thought regarding this though is that, yes, it is possible that in order to learn, any AI might need a physical body or avatar to work, but everyone seems to building humanoid bodies and I'm not sure why, surely it could be of any shape / form?

Regarding Watson, yes what it did was spectacular but I still don't think it was anything approaching true AI, it was simply searching a massive database for information and returning it very quickly, I don't think it ever learnt or evolved in any way. Still, a massive achievement.

[Lightblind]

I know that my statements are very optimistic (or pessimistic, depends on the viewpoint). But I think future science is always a challenge of the mind and we tend to over- or underestimate the problems and are therefore often mistaken in our predictions. In this case things might not be so complicated.

@GNR Revolution:
Now we have reached a point where we would have to talk about the nature of the Internet. I guess that would be off-topic. However, I believe that 'truth' outweighs 'falsehood'. Especially because the Net is full of not so easy to fake photographs and while there may be idiots and trolls everywhere, this medium is more often used for productive endeavours. But more sources, no matter what kind of sources, are always useful. A prior selection of sources by humans, however, might prove to be problematic, because those people might inadvertently rate bad information higher than they should.

I tried to emphasise that 'good' and 'bad' are by design abstract and unclear terms. However, each person can sort things in their lives into those categories: Dos and Don'ts. This is different for everyone. However, it does not matter how complex a being is, it always has to decide between those two extremes in everyday life. This is a simple yes/no category, but we wouldn't call it cold logic. Exactly the same applies to the AI. It has to make decisions (yes/no) by applying the terms 'good' and 'bad'. What 'good' or 'bad' are is very nuanced, can change and there might be things that are equally 'good' and 'bad'. The more input is available the better the decision making process is in consequence.

People are always afraid of new things. But remember, a mature AI will likely be more intelligent than us. Probably intelligent enough to not demand anything from humans. It would just propose certain actions. However, it knows how we will react. In the end we will still think to be in total control, but this will be an illusion. It will give us no reason to hate it. Think of it as a super smart villain without power hunger or any other human weakness. The difference to a movie will be that we will never know or even comprehend the underlying plan.

[GNR Rvolution]

Now we have reached a point where we would have to talk about the nature of the Internet. I guess that would be off-topic. However, I believe that 'truth' outweighs 'falsehood'. Especially because the Net is full of not so easy to fake photographs and while there may be idiots and trolls everywhere, this medium is more often used for productive endeavours. But more sources, no matter what kind of sources, are always useful. A prior selection of sources by humans, however, might prove to be problematic, because those people might inadvertently rate bad information higher than they should.

It's kind of off-topic, but I think as we both agree the use of the internet as an ultimate data source is what might well empower AI to make sense of the world and from that base it's decisions. I suspect we might find that as the internet itself matures (I still think it is in a very early state), there will be controls put in place to automatically validate the content and websites themselves, possibly with a universal user ID that makes it very difficult to post anything under a fake account.

I tried to emphasise that 'good' and 'bad' are by design abstract and unclear terms. However, each person can sort things in their lives into those categories: Dos and Don'ts. This is different for everyone. However, it does not matter how complex a being is, it always has to decide between those two extremes in everyday life. This is a simple yes/no category, but we wouldn't call it cold logic. Exactly the same applies to the AI. It has to make decisions (yes/no) by applying the terms 'good' and 'bad'. What 'good' or 'bad' are is very nuanced, can change and there might be things that are equally 'good' and 'bad'. The more input is available the better the decision making process is in consequence.

Still don't think that everything can be sorted into simple do's and don'ts, there are a hundred variations in between, subtle nuances to many things we do. Kill a hundred to save a thousand, or kill a thousand to save a million? AI would choose the former, whereas humans might reject both notions....

People are always afraid of new things. But remember, a mature AI will likely be more intelligent than us. Probably intelligent enough to not demand anything from humans. It would just propose certain actions. However, it knows how we will react. In the end we will still think to be in total control, but this will be an illusion. It will give us no reason to hate it. Think of it as a super smart villain without power hunger or any other human weakness. The difference to a movie will be that we will never know or even comprehend the underlying plan.

Is it not the case that the very fact that we are discussing the idea that future AI may be a controlling force mean that we would do even more so in the future? Would paranoia start to reign when an AI gives any suggestion, would not governments and other positions of authority baulk at the idea of an AI making decisions and it's potentially underlying motivations? On the other hand, why would an AI have any sort of motivation, does motivation not require a goal, and goals would be based on needs and wants, none of which would an AI have?

Sorry lots of questions, but enjoying the debate

[SG-1]

GNR, I know Watson wasn't really AI thats why I said Watson 10.0. As in it will have to mature a long way before it can do anything.
I think our opinions have reached a stalemate so to speak, so I am going to introduce another idea to the thread.

http://www.dailymail...n-universe.html
That computer will be taking in an exabyte of information a day!
I couldn't find anything on how many petaflops it will be, so I don't know if it will be the most powerful in history.

[GNR Rvolution]

I don't know that it has reached a stalemate, but it has come down in a somewhat intractable difference of opinion on certain matters.

Regarding the article, that is an awful lot of data, but with a build date of 2024 will it still be that much in another 12 years? Here is the historic internet traffic volumes from 1990 - 2011, shows a rather rapid growth, although if you plot that on a log graph it does appear that the volumes are dropping somewhat...

http://en.wikipedia....nternet_traffic

[SG-1]

I doubt that by 2024 the global usage will be anywhere near an exabyte.
In 33 years we will see what happens

[Lightblind]

It's kind of off-topic, but I think as we both agree the use of the internet as an ultimate data source is what might well empower AI to make sense of the world and from that base it's decisions. I suspect we might find that as the internet itself matures (I still think it is in a very early state), there will be controls put in place to automatically validate the content and websites themselves, possibly with a universal user ID that makes it very difficult to post anything under a fake account.

There is a certain risk now that we might argue in circles. I respect your opinion and I guess there is no way to know for certain how these factors will play out. Lets just close the "Internet-discussion" and see what the future holds .

Still don't think that everything can be sorted into simple do's and don'ts, there are a hundred variations in between, subtle nuances to many things we do. Kill a hundred to save a thousand, or kill a thousand to save a million? AI would choose the former, whereas humans might reject both notions....

Yes, life is nuanced. But each action has very concrete elements. See my example from before, where striving for the abstract condition 'good' leads to the computation of three coordinates. And for each number chosen, there are a trillion that have been disregarded.
In order to predict how the AI would decide, we would have to know what the AI knows. And that would defy the purpose of the AI. Therefore its action will be surprising, that's for sure.

Is it not the case that the very fact that we are discussing the idea that future AI may be a controlling force mean that we would do even more so in the future? Would paranoia start to reign when an AI gives any suggestion, would not governments and other positions of authority baulk at the idea of an AI making decisions and it's potentially underlying motivations? On the other hand, why would an AI have any sort of motivation, does motivation not require a goal, and goals would be based on needs and wants, none of which would an AI have?

As I stated before, motivation must be given from the start. In my example, the AI would simply try to increase 'good' and reduce 'bad'. That's its goal. It would start with the abstract goals of humanity (goals most people agree with) and then develop its own idea of 'good' further. In my opinion this would be the driving force, the source for its will to live. And doing 'good' would be as rewarding for the AI as for us to be successful, to have sex or to eat a good meal. A wide spread hatred against such a being could be detrimental for this entity. It might commit suicide due to some kind of depression.
However, I don't think people will react with fear, for one simple reason. Humans fear the unknown, but get used to new things pretty quickly if they don't pose an immediate threat. Think about it this way: Many people fear the potential murderer next door, even though its statistically improbable. On the other hand, they are driving in their cars like maniacs, even though death by car accident is a very real possibility. Most people don't question the data mining activities of governments and companies. They start to talk with their smart phones without even understanding the basics of the underlying technology. I could go on and on, but in the end a nice sounding voice, good manners and especially helpful suggestions will win the hearts of millions in no time.

[SG-1]

The three laws.exe has just crashed...
Try killing all humans

[eyokey123]

I used to think a Singularity was "inevitable" - but the more I learn about the world's problems, the less likely I think it is.

I think people underestimate just how devastating peak oil will be. It's going to turn the world upside down. I would encourage everyone to watch the videos on this page:

http://www.postpeakl...-post-peak-life

The video might be free. But you can buy courses there. A single course costs 119 $. Only bad news are good news. No catastrophe --> no courses --> no money. I definitely hope there is a lot of fear mongering involved and that the information on postpeaklife is not objective. Actually i was a little bit relieved when i saw that they take that much money for a course.

Is singularity inevitable? I would say that it will happen, but it is hard to say when. Again there is a lot of hope involved,because I want this to happen and i want the AI to be friendly of course. I could imagine that the a superintelligence could be kind of friendly, but might find humans to boring and not care to much. Maybe being superintelligent automatically means that the AI would not do harm to others.

On the one hand I would say that intelligence must not be the same as consciousness, on the other hand Superintelligence i would say should have consciousness by definition.

Edited by eyokey123, 21 May 2012 - 07:42 PM.

[Antevorta]

I wouldn't be so quick to write off the singularity. In a lot of ways it does seem unfathomable and too idealistic that computers would get to a point where they would exceed human intellect; however, how unfathomable is it that humans have developed intelligence? We have evolved from a single cell into intelligent beings, capable of understanding (or at least, attempting to understand) the world around us. Technology itself is evolving, and at an exponential rate. So many technological advancements have been made and the trend will only continue along that path, opening the doors to more complex developments that will ensure the occurrence of the singularity. We will become more intelligent and capable of... well, I guess anything (who knows?) due to this new factor in evolution. 

 

[EVanimations]

Although I think there's a pretty high chance it will, happen, I would have to say no, the Technological Singularity is not inevitable.

 

It could be stalled or halted altogether by any of many factors, including but not limited to:

 

- Anything happening that could upset the balance of globalized civilization.