I find it so interesting because it makes my mind wonder. The subject touches upon the essence of what we think makes us human (as compared to, for instance, monkeys): The ability to reason logically and communicate about these 'neurological structures' on the one hand, and the fact that, whatever exactly that means, we are aware; I sit here right now being intrigued by the fact that I am sitting here right now and that I am writing this, wondering what it is the future will bring and enjoying the fact that I don't know.
When contemplating the concept of Artificial Intelligence the distinction between 'the ability to hold up an intelligent conversation' and 'awareness' is key. Two entirely different things, or are they? Food for thought indeed...
I am not the only one who's intrigued. And that's what this blogpost is about (as far as I can see right now, but, you know how it is, nobody knows what the future brings). There is a bet out there. Two very energetic scientists who have bet (lots of money involved too) on the question whether Artificial Intelligence will be achieved by the year 2029. Raymond Kurzweil and Mitch Kapor. Raymond thinks it will be done, Mitch thinks No Way. Now.
So what are they going to do? How are they going to 'test' Artificial Intelligence? They are going to use what's called the Turing Test. Back in the fifties of the last century there was this British Mathematician Alan Turing. He came up with this test 'in which one or more human judges interview computers and human foils using terminals'. Specifically what's going to happen is this:
In the year 2029 (and if so Raymond requests because he feels confident enough, earlier) three people (judges) will sit down behind a keyboard to have four different chat conversations (each lasting two hours). They can talk about anything. Three of these conversations will be with a human. One with a computer program. In the end the judges sit down and do two things: 1. They have to say which conversation was with the computer and 2. They have to rank each conversation (1 to 4, least human to most human).
The computer program 'passes' the test if: 1. at least two out of three judges thought they talked to a human and 2. if the median rank of the computer is equal to or greater than the median rank of two or more of the three human foils.
They have made this bet 'official' by publishing it online back in 2002. So please do read it for yourself, it's, well as I said, intriguing. So, let's go for the analysis: Hmm. First of all, the computer program is supposed to prove it can hold up an 'intelligent conversation'. The question whether an 'entity' is 'conscious' or 'aware' is an entirely different one. Secondly, so-called fictional histories may be used during the conversations. Since the computer program doesn't really have a history, it didn't really take that vacation last year and it definitely didn't 'grow up' chasing girls, it will have to fake. And that might even be harder then 'just being human and real'.
Imagine you are going to chat with 'someone' for two whole hours. Imagine all the questions you could ask. Imagine asking someone what exactly it is that arouses him or her, essentially, during foreplay, or how it felt growing up 20 years ago? A computer program that could fool you? For two hours? Conscious or not, that seems quite a challenge.
So what makes Ray so confident this can and will be done within the next 20 years? Well, lucky us. They put that online too. Here is why Ray thinks it will be done. Here why Mitch thinks No Way. And here where Ray says Yes Way! And here we go again: summary and analysis time >>
In order to explain Ray's confidence, let me start out by describing Moore's Law. It descibes an important trend in the history of computer hardware. Since the invention of the integrated circuit in 1958, the number of transistors that can be placed inexpensively on an integrated circuit has increased exponentially, doubling approximately every two years. This trend was first observed by Intel co-founder Gordon E. Moore in 1965. Almost every measure of the capabilities of digital electronic devices is linked to Moore's law: processing speed, memory capacity, even the number and size of pixels in digital cameras. So that's that >> computer hardware is developing real fast and computer hardware can be found in anything, so everything is going real fast (remember how big those arcade-machines were back in the eighties, on which you could play Pacman, and then the desktop pc's in the nineties, and then our current cellphones, and then...)
Exponential growth, ok. So what else? The second thing relevant here is what's called 'paradigm shifts'. Example: between 1946 and 1958 they used to build computers using vacuum tubes. These vacuum tubes were made smaller and smaller (exponential growth) until it was no longer feasible to 'maintain a vacuum', so it was time for something new: Tadaa! The transistor [check out these computer chronicles]. The point being: Each time one approach begins to run out of steam, research efforts intensify & the next source of exponential growth is found (that goes even faster than the previous). Within the next decade we'll have to go through a new paradigm shift because the distance between transistors on our current chips is soo small (check out this article about Intel's latest 32 nanometer chips) that new physical rules start coming into play. Hard to imagine how small? There are almost 2 billion transistors on this one chip! 2 billion!!! And still we want faster chips!? >> No problem, lots of new technologies (molecular based computing, 3D computing, etcetera) are eagerly awaiting to carry the relay baton. The transfer WILL BE there, let's just hope it'll be smooth too :)
Lastly, the phenomenon of ongoing exponential growth is far broader than computation alone; communication technologies, biological technologies, all subject to the same double exponential growth. And they influence each other too, causing things to develop even more rapidly. So... a whole mouthful, and then some.
So back to the original question: What makes Ray so confident? Well, there is hardware and there is software. First the hardware. In order to build a system that has the capacity of fooling us into thinking it is human, let's assume that that system has to be at least 'as complex' as the human brain (i.e. it has to have the same Information Processing Capacity as the human brain or more). I just told you there are around 2 billion transistors on Intel's latest chip. Well, the human brain is estimated to have around 100 billion neurons, each with a thousand connections & each connection being able to 'handle' 200 'digitally controlled transactions' per second (so 20 million billion operations per second) >> According to Moore's law by 2020 the power of the human brain could be on your desk for let's say a 1000 dollars >> 10 years later, at the same price a system 100 to 1000 times as powerful as the human brain! That's surely impressive, but without the proper software, even a system like this amounts to little more than a very powerful calculator. The trick's in the software. Trick! Ha! We're talking about reverse engineering the human brain here! Quite a trick indeed. Surely, that won't be possible, will it?
Well, Raymond thinks it will be. He is convinced that we will understand the principles of operation of the human brain and that we will be in a position to recreate its powers in synthetic substrates before the year 2030. Now to most of us that sounds ridiculous, doesn't it? I mean, come on, we're going fast, but THAT fast?! No Way! But that's the funny thing about Raymond. Everytime somebody says No Way, he says Yes Way and then he comes up with compelling arguments...
...In this case the Human Genome Project is a good example. In the late eighties a bunch of enthusiasts came up with the idea to map the complete human genome (identifying all 20.000-25.000 genes in human DNA, determine the sequences of the 3 billion chemical base pairs that make up human DNA and storing all this in a big database). They wanted to do this in 15 years! Back then, ridiculous! Science Fiction! Impossible! And so the journey began in 1990. By 1997 they had mapped out 1%. One percent! Half the time's up and 99% to go. You see, they should have listened to the sceptics. Impossible! Ha! >> 3,5 years later they were done!!! Ha! What's the moral of this story >> exponentiality! 1 + 2 + 4 + 8 + 16 + 32 + 64 + 128 + 256 + 512. Get it?! (remember that book: The King's chessboard).
Our brains think linear, we are very much incapable of truly imagining the nature of exponentiality and it's impact.
Having said that, back to reverse engineering the brain & Raymond >>
>> First of all, make no mistake, the effort to reverse engineer the brain is further along than most people realize (did you see my previous article: Erasing memories no longer Science Fiction)
>> Second, just like the Human Genome Project, our capability to 'scan the brain' is growing exponentially too. Things get smaller and smaller and smaller; scanners the size of blood cells that are observing the connections between neurons. What, 15 years?! Having your complete genetic code mapped for less than $10? What, 15 years? Your T-shirt = the computer, what 15 years?!
Previous examples are powerful enough by themselves, however, it's when thinking about the fact that all these technologies come together and what that means, that's where even my imagination (and I don't need to tell ya, I am a genuine nutcase) goes heywire. Imagine going back to the eighties and asking someone to imagine a portable phone. Then give 'm an Iphone or a Nokia N96. You think you could explain to them why you need to carry 100 LP's of music with you, on your phone (and if you are thinking 'why not a 1000', then you might catch my drift)? Then touch the screen. Ah! Touch-screen, what?! Finally explain how easy it is to update your weblog, using that phone and that the Internet access is fast enough to watch Youtube or 'general television'. And oh, please don't mention the GPS system, you wouldn't wanna give anybody a heart attack...
Catch my drift? Do I myself? Ok, things will basically change unimaginably fast the coming two decades, but will that lead to a machine that can display or 'fake' (whatever that means) emotional intelligence? Raymond thinks it will because although human emotional intelligence is complex, it nonetheless remains a capability of the human brain. Ultimately the human brain is made of the same small list of proteins that all biological systems are comprised of, so there is little basis to expect that the brain relies on some nonengineerable essence for its capabilities.
And that brings us to the last part (can you believe that?): the before-mentioned subject of consciousness/awareness. As I just said 'little basis to expect the brain relies on something non-engineerable'. And there we go, the essence. A lot of people out there who are not willing to accept the previous statement. People like my mother; 'There has to be something else', 'humans have a soul', 'humans have a spirit'. Now that could be true, just as much as it could not be true. Problem is, so far we have no way of proving nor disproving. If you want to go about doing things scientifically you have to at least be able to either prove or disprove something (don't look at me here, check out this philosopher of science, Karl Popper). And this is just my first shot...
Secondly: The assumption humans and only humans would have this specific, non-tangible soul type-of-thing, that ONE thing that makes them special, that assumption implies, if you ask me, a rather black-and-white worldview. We have the spirit, animals don't. Ok. humor me and look at that chimp at the top of this article. He seems rather 'aware'. Maybe not to the extend we are...
...and that's exactly it right there. That last sentence reflects the essence of how I think about consciousness/awareness. I don't believe in black-white thinking. everything is a grey area. It reminds me of my time as an exchange student in Davis, California (back in 1998). I had a wonderful girlfriend there, Silvana Renteros (and please rest her soul) and she had an autistic daughter, with the beautiful name Aurora Renteros. She was 5 years old. I remember going to the playground with her, remember 'trying to control her' in the alleys of Safeway. She just loved smacking everything she saw onto the ground. With passion! I remember nights, so many nights, where Aurora had gotten up, redecorated the room by means of the contents of the fridge and then she would always, (always!) end up at the bathroom sink, staring at the flow of the water. If it rained outside, she would run outside and go nuts! nuts I tell you. She just loved water. Jarno! Yes? What's your point? My point? Hmm. You couldn't approach Aurora. she would hit you. She seemed 'completely absorbed' in her own reality. She loved water, but we'll never know why. Was it the look, the sound, the touch? They say autistic people perceive the world as one big amalgam of stimuli. video, audio, touch, all blended into one big mix.
Question: Are they aware they are here?
Answer: Yes, no and everything in between.
I read many books on autism. One of them: 'Thinking in pictures' by Temple Grandin. On the back of the book and I quote:
'I hardly know what to say about this extraordinary book. This remarkable woman gives us much insight into animals and into the world of autism. The explanations of consciousness and her unique experience of it provide a way to understand the many kinds of sentience, human and animal, that adorn the earth" - Elizabeth Marshall Thomas, author of the hidden lives of dogs.
The point being: I think 'consciousness/awareness' is nothing more than the combination of all electrical signals that are flying through our brains. Whatever brain: Cow, monkey, human (autistic, non-autistic or simply nuts like me, whatever). There is no black and white, there are different 'levels'. As I said before, look at the face of that chimp at the top of this page, you know he is conscious to some extend. Am I right? Does it feel that way? If it does, doesn't exactly that sentence, 'to some extend', prove my point?
Oh, and there's just one more thing! Conscious not conscious... does it even matter? When there's interaction between two 'beings', behavior of one being depends, in part, on the behavior of the other being. Not so much on the question whether that other being is conscious.
Did you ever see people talking to their dogs? Exactly, that's what I mean. Dogs are dogs. They want to run around, sleep, shit, mark their territory by pissing onto any tree they can find, fuck other dogs, and eat. That's it. And Oh yeah, the hierarchy. They need to have a boss. Well, goody, the human is the boss. But talking to him? Talking?! There are psychological treatments out there..., for dogs. Clinics. Stores with dogg-stuff. Vast amounts of dogg... stuff...
...Have you seen this video of children playing with this robotic toy 'Pleo'. A robotic dinosaur. It was treated like any other animal with 'feelings'.
All I mean to say is this: The question whether another 'being' is thought of as being 'aware'/'conscious' is irrelevant.
Well then. Who will win this bet? Raymond believes it will be possible, Mitch thinks the essence of what it is to be human cannot be captured into some program (prior to 2030), hence it's not possible. Hmm. I'll side with Ray when it comes to his belief that our imagination falls short significantly when predicting the future, due to the nature of exponentiality. Does that mean I believe the Turing Test will be passed by 2029? No way!
Although I believe firmly that consciousness need not to be a prerequisite for passing the Turing Test, and although I also do believe the next two decades come with breathtaking change, when it comes to having 'an intelligent discussion' with a program, if you ask me, seriously, there's not yet even a glimmer of hope; Two weeks ago I wrote this article about the Loebner Prize competition. I was pretty excited back then about the outcome and I have had several chats with Elbot myself since then. I agree with Luciano Floridi. He was one of the judges at this years' competition & he is an influential thinker in the field of philosophy of technology and ethics. On his blog he wrote he had great fun and he was intrigued by the whole competion, but in the end concluded that the computer failed miserably.
"It was the usual, give-away, tiring, Eliza-ish strategy, which we have now seen implemented for decades."
Sounds kindda harsch, but indeed, I felt the same way about the online version of Elbot. Although its linguistic analysis abilities are in some respects impressive, it really doesn't 'get' anything. Let alone if I would subject the program to some questions as proposed by Luciano >>
1. 'If we shake hands, whose hand am I holding?'
2. 'I have a jewellery box in my hand. How many cd's can I store in it?'
3. 'The four capitals of the UK are three, Manchester and Liverpool. What's wrong with this sentence?'
So, we currently have a worldwide network of pretty powerful computers, I recently blogged about the fact that, this year alone, something like 370 exabytes of data has been created, and indeed technologically we've come quite far > they've even managed to erase specific memories in mice. Yet, a few simple questions like the ones above cannot be answered...
...I think we're on the wrong track here. And the nature of acceleration doesn't change that. Things have been accelerating for a while now, so I would have expected at least that glimmer of hope by now, if I were to believe the Turing Test could be passed by 2030 (or, at all). So No Way Ray! But I do hope, you'll prove me wrong.
Thank you for tuning in and,
Bye bye now,
By the way: Try this Chatbotgame.
Jarno de Vries