Eddie asked:

Is artificial ‘intelligence’ possible? It seems to me that no matter how sophisticated and powerful a ‘thinking’ machine nowadays could be, it is still a mere network of electronic circuits, i.e. it only deals with specific inputs and then gives the corresponding outputs as programmed. It does not actually think but just strictly executes programmed instructions. There manifests no free will at all. Isn’t free will a necessary element of intelligence?

Answer by Stuart Burns

You ask an interesting set of question, Eddie. The subject has attracted the attention of many philosophers, and many workers in the computer industry. Many of these thinkers would agree with you – that it is impossible in principle for a computing device, no matter how sophisticated and powerful, to ‘think’ or to be ‘intelligent’. Although I have not encountered the argument you make here, that ‘free will’ is a necessary element of intelligence.

But there are also many thinkers in this arena who would disagree with you. And I am one of those. Having spent 30 years of my working life programming computers of various sorts, and in my spare time pursuing my hobby of philosophy, I have learned to see in your comments four very challenging issues that must be addressed before any meaningful answer can be offered to your questions.

(1) The first of these issues is ‘Just what do you mean by ‘intelligence’?’ Rather than trying to provide a genus and species definition in terms of necessary and sufficient conditions that would identify ‘intelligence’, perhaps it would be easier to consider the issue in terms of ‘How would you recognize ‘intelligence’ if you encountered it?’ In 1950, Alan Turing provided an interesting idea known as the ‘Turing Test’. Suppose that you were closeted in one room, conversing with a pair of other conversationalists, each closeted in their own little rooms. It makes no difference how you imagine this conversation taking place, or where these two other conversationalists are located. How would you go about deciding whether either (or both, or neither) were ‘intelligent’?

Suppose, using whatever criteria might occur to you, you decide that at least one of these two conversationalists was ‘intelligent’. And suppose that you then discovered that this ‘intelligent’ conversationalist was in fact a computer. Would that not qualify as an ‘artificial intelligence’? If not, why not?

In order to pass a ‘Turing Test’, all a computer would have to demonstrate is an ability at free ranging English conversation at a competence level equivalent to any flesh and blood person who you would consider as ‘intelligent’. If you imaging this conversation taking place by keyboard, we already have sophisticated computer programs to read and ‘understand’ (within strict limits) English written conversation. But with modern speech recognition and synthesis software, and image manipulation software, your imagination need not be limited to the keyboard.

The ELIZA program was written at MIT by Joseph Weizenbaum between 1964 and 1966. ELIZA emulates a Rogerian psychotherapist. But it has almost no intelligence whatsoever, only tricks like string substitution and canned responses based on keywords. Yet when the original ELIZA first appeared in the 60′s, some people actually mistook her for human. (You can try talking to ELIZA at The illusion of intelligence works best, however, if you limit your conversation to talking about yourself and your life. As I said, ELIZA is not very sophisticated.)

Since ELIZA first appeared, much has been learned about English conversations and computer programming. Modern versions of these programs (now known as ‘chatterbots’) continue to fool people. ‘CyberLover’, a malware program, preys on Internet users by convincing them to reveal information about their identities. The program flirts with people seeking relationships online in order to collect their personal data. A much more sophisticated, and much more free-ranging, computer conversationalist is possible – although they are still very obviously not as competent in English as a young child (but we often classify a young child as ‘intelligent’). At least they are demonstrating that convincingly passing the Turing Test is most likely simply a matter of time. Consider the accomplishments of Watson – IBM’s contestant on Jeopardy.

(2) The second issue I see in your comments is ‘Why do you think that electronic circuits are any different from neurons?’ Of course, this question pre-supposes that you are something of a materialist, and do not think that the human mind (and human intelligence) is a result of some non-physical ‘soul’ (for lack of a better label). If you are, in fact, a mind-body dualist, then the issue of ‘artificial intelligence’ is probably moot. Most likely, you have already decided that anything without an immaterial ‘soul/mind’ cannot, by definition, be intelligent. But, if you believe that the human mind is a product of the biochemistry of the neurons in the brain, then why do you think that neurons are any different from electronic circuits. Neurons are just the same as electronic circuits – they only deal with specific inputs and then give the corresponding outputs as biochemically programmed.

It is true that we have not yet constructed a computer with a neural-net architecture that has anywhere near as many ‘circuit components’ as a human brain. So it is not surprising that those we have constructed have not demonstrated behavior we would characterize as intelligent. But we have constructed several smaller versions of such computers, and they have demonstrated remarkable mimicry of human capabilities in narrow fields (such as pattern recognition, for example). But if the neuroscientists are anywhere nearly correct in their characterization of the brain, we are just as equally constructed out of ‘circuits’ (neurons, in our case) that each ‘only deals with specific inputs and then gives the corresponding outputs as programmed.’ So on that basis, there is no evidence that a properly architected, and sufficiently large electronic computer could not evidence ‘intelligence’.

(3) The third issue I see in your comments is ‘Just what do you mean by ‘think’?’ This is not a superficial issue. For as long as computers have existed, some people have maintained that they do not ‘think’ because they cannot do some example of what humans do. And throughout the history of ‘artificial intelligence’ computers have repeatedly demonstrated the ability that was previously claimed as exclusively belonging to human ‘thinking’. Everything from Deep Blue’s defeat of the World Chess Champion Garry Kasparov, to the Jeopardy victory of Watson. From painting, to writing original musical scores, to poetry. So if you claim that computers just strictly execute programmed instructions but do not actually ‘think’, you need to come up with some notion of what constitutes ‘thinking’ that you suggest cannot be demonstrated by the computer.

In general parlance, ‘thinking’ is just the human process of evaluating alternatives and deciding which alternative is ‘best’ according to some established criteria of ‘best’. But that, of course, is just what computers are very good at – given the appropriate data. What humans are currently clearly better than computers at doing, is collecting the appropriate data. But that is not what is usually meant by ‘thinking’. And might reasonably be explained by the fact that we do not provide computers with general purpose ‘data collection’ devices.

(4) The fourth issue I see in your comments is ‘Just what do you mean by ‘Free Will’?’ At the level of physics, the electronic circuits that make up a computer are clearly deterministic. But then, according to the neural scientists, at the level of biochemistry, so are the neurons that make up our brain. At the most basic level, we are a deterministic biochemical computing device. Both the computer and the brain, at the most basic level, only deal with specific inputs and then yield the corresponding outputs as programmed (or learned). So where is there an opening for Free Will? Remember that you can’t rely on random processes here, because it is inherent in the concept of Free Will, that if anything other than ‘You’ make the choices, then the choice is not a product of your Free Will.

But what, then, are ‘You’? Are you not a product of your nature (your construction) and your nurture (the things you have learned)? When you make a choice, when you exercise your Free Will, you bring to that choice your experiences and memories, your character and desires, your goals and your values. That is what ‘You’ are. And given who you are and what you are made of, in any particular circumstance the only way that you could choose other than how you do choose, is if something external to you forced the issue. In many circumstances, your friends can predict the way that you will choose, because they know who you are. And if you were to choose other than you would normally choose, your friends would look for some unusual reason for this unusual choice. And is this not the same as for a sufficiently sophisticated computer? (If you would like to peruse an essay I have written on the compatibility of Free Will and Determinism, please see .

So what is your notion of Free Will? And how is it related to your concept of ‘intelligence’? I would be very interested in your thinking on this issue. As I said above, I have not seen before an argument linking Free Will to Intelligence as you do here.

In Summary then, before a meaningful answer can be provided for your very interesting questions, I think that we must address just what you mean by ‘intelligence’, ‘think’ and ‘Free Will’, and just why you think that neurons are different from electronic circuitry when it comes to those notions.

 

Answer by Jürgen Lawrenz

You’ve practically answered your own question. Artificial intelligence is not a philosophical issue by any means. Stringently regarded it is merely a vocabulary of puns based on analogies with human intelligent behaviour and used in a restricted technological sense.

Human beings are vulnerable to taking their own puns seriously, however, and it is a problem in this case because of the social ramifications of misuse of this verbiage in education. Widespread public interest leads to the puns generating their own paradoxes, which are then debated for as long as interest can be sustained. While in the short term the chief beneficiary is the entertainment industry, the possible damage to social value systems must not be disregarded or underestimated. One insalutary side effect can be contempt of the evolutionary acquisition of human intelligence, based on the delusion that artificial intelligence is of a ‘higher’ grade. The potential for such doctrines to exert a detrimental effect on the self-perceptions of humans is not to be dismissed.

The point you have made can be also be expressed another way. The only form of intelligence known to us–involving free will–is creature intelligence. The ‘trick’ used in innumerable debates (even learned works) to persuade us that this is all illusory is at bottom quite simple. There are severe restraints on the exercise of free will. Other wills exists and resist or cancel mine; social impediment like laws and customs intervene or obstruct it; and there are also physical constraints. But once you understand this, it is no more than saying, ‘I have muscles and a certain strength’ while acknowledging that a crane can lift greater weights. When you are in chains your strength cannot be exercised. The strength exists nonetheless; likewise free will is a basic feature of humans (and animals). Impediments to their exercise don’t render them illusory of non-existent.

The example of the crane is, of course, central here. A computer is no more intelligent than a fire alarm. So, just as we design cranes to lift great weights, using technological resources, so we can design computers to execute certain mechanically reproducible thinking tasks. Where the AI brigade comes off the rails is in the belief that a mere sham is real, because superficially regarded it looks real. But you already nailed it.

Let me congratulate you on your independent thinking. You will have great difficulty maintaining your attitude. But I give you a nice bon mot to remember for all occasions of debate (sorry I forgot the author of this quote): ‘The greatest danger is not that computers will ever think like us, but that we will start to think like computers.’

 

Answer by Helier Robinson

First of all, free will is not a necessary element of intelligence: it is possible that the entire Universe is determined, in which case there is no free will, but there is still intelligence; in other words, there is intelligence whether or not there is free will, so free will is not a necessary condition for intelligence. But this is a small point. The major point in your question is whether an electronic network can be intelligent, and the answer is yes because our brains are electronic networks. To be sure, no one has produced artificial intelligence yet, but that does not mean that it cannot be done.

The key point here is emergence. When a structure gets sufficiently large new properties emerge in it, to form the next higher level structure. For example, life emerges out of sufficiently complicated molecules, and consciousness emerges out of brains. Our brains are enormously complicated networks; in fact they are the most complicated structures that we know of. A structure of inter-connected transistors that was equally complicated could well be both conscious and intelligent.

 

About these ads