s Raymond Kurzweil brilliant?  Absolutely.  Is he right that in the very near future—say, the next 21 years—that the computer will be able to simulate human emotions so well that they will be the equal of human beings?  I think not.  In fact, I think that the evidence against Kurzweil is rather compelling in this regard.  Indeed, he just glosses over just how remarkably rich the human being is in terms of the sentiments and just how relevant emotions are to being a human being and to interpreting human behavior.  The best evidence of this comes from a rather unsuspecting source, namely translations.  In general language presents problems for Kurzweil that he has not considered.  However, I shall start with translation itself.

Sentences like “Turn right”, “Go three blocks south” and the like are easily translated, as are sentences like “The second plane arrived earlier than the first one” and “Mary is a very gifted physicist”.  These are all literal sentences and have no metaphorical meaning.  This is so even for a sentence like “Mary is a very gifted physicist,” which can be understood as rather subjective claim. 

But consider, for example, a great speech such as Martin Luther King’s famous speech “I Have a Dream”.  No computer could flawlessly translate that speech from English to, for example, French.  A computer could translate the factual aspects, but not the more rhetorically eloquent parts of the speech. 

This is because translation at its best is tied to having a profound sense of how words are used in both the original language and the language in which the text is to be translated.  And computers do not even come close to having that kind of sense of how words are used merely in virtue of having all the words of the language.

Now, to have a sense of how words are used is, of course, to have a sense of the feelings that words evoke when uttered: (1) the feelings evoked by the utterer and (2) the feelings evoked by the target of the utterance and (3) the feelings evoked by mere hearers (not targets of the utterance who heard the utterance).  A man who calls a male “faggot” in front of others may have one feeling in making the utterance; he intends, though, that the target of the utterance have a different feeling; and in many cases, he will intend that the mere hearers of the negative sobriquet have yet another set of feelings. 

Some words are intended to insult; others are intended to invoke shame; and so on.  I once said to a good friend and colleague in his 70s “Man you are the shit”.  He understood ever single word, but had not a clue what the sentence meant.  Because I made that utterance with a smile and we had already known one another for at least a dozen years, he assumed that I was not being nasty.  I told him that I was giving him a compliment.  I think he believed me.  My colleague's problem, of course, is that for him it was very nearly a conceptual truth that the word "shit" only had negative connotations. 

Knowing what words work with whom and what impact that they will have upon whom is part of being a competent speaker of a language.  No 20-year old expects her or  80-old grandmother to understand the expression “Grand Ma, you are the shit,” unless Grand Ma’s use of language reveals her to be astonishingly avant-garde. 

And what about the word “nigger”?  Will a computer be able to get away with saying this word to a black person?  Will a computer grasp why some can utter this word (namely, blacks) and others cannot (namely, non-blacks)? 

In effect, what I am arguing is that computers will have the emotional capacity of a human being when and only when it is the case that computers are capable of the fluency of language that is characteristic of being a competent speaker of a natural language, be it Chinese or Arabic or French or English.  And so on. 

To date, computers do not come even close to exhibiting the competency of a natural language that even the most uneducated scoundrel has.  A scoundrel can express a wealth of emotions via language from utter hostility to fear to rhapsodic pleasure to disappointment to sheer hurt.  And the way in which the scoundrel speaks would be the tell-tale sign.  An incredibly sad person never ever sounds like his cup runneth over with joy; nor, for that matter, the other way around.  "I am so tired" said after a night of wild sex is not the same "I am so tired" said after having studied all night.

The computational capacity of computers is absolutely extraordinary.  Thanks to that capacity gigantic airplanes can land flawlessly with little to no visibility.  However, it is simply a mistake to suppose that the emotional capacity of human beings reduces to sheer computations.  And one reason for this is that we can sometimes predict how a person is going to behave and yet be extremely moved when it happens.  For example, you may very well realize that Mom will shed tears of joy when you give her the mink coat that she has always wanted but has never been able to afford.  Yet, when she sheds those tears of joy, you are likewise moved to tears.  The problem is not that you failed to predict her behavior.  Rather, it is that correctly predicting her behavior is no substitute for experiencing that very behavior.  Punishing a child can have a like reaction in the opposite direction, as when a parent has to struggle to stick to a punishment because, as predicted, the child cries.  So, emotions can have enormous significance in ways that have nothing to do with making the right calculations. 

And all of this is particularly interesting because language is not static.  Nowadays, men feel comfortable saying “I love you”, something that was not at al an opton a mere 5 years ago.  This "I love you" is none other than an expression of admiration; and it is not quite the same as when I say “I love you” at the end of a lecture to the 389 students in my Ethics and Value Theory course this semester, which (yet again) is not at all akin to the utterance “I love you” said between two lovers holding hands and looking one another in the eyes. 

The truth is that (a) having a vocabulary is one thing and (b) knowing how to use it is quite another.  And the move from (a) to (b) is as extraordinary and profound and sublime as things get. 

There is no evidence at all that the move from (a) to (b) is simply a matter of sophisticated computations.  After all, it is impossible to give a computational account of every time a person utters the expression “I love you”.  I do not, for example, say this after every lecture; nor do I say it after every lecture that I deem to be moving.  In fact, I have been moved to tears in a lecture and yet did not end the lecture with “I love you”. 

Let me conclude this essay by changing direction.  Let us suppose for just a moment that Raymond Kurzweil is right and that within the next 21 years or so, computers will be the emotional equal of human beings.  There is another fundamental question that arises, namely: Is that a good thing for human beings?  And it is rather stunning to me that Kurzweil seems not to have given this question any thought at all.  For he seems to suppose that once computers become the emotional equivalent of human beings, then they will be a formidable ally of human beings.  However, this view of computers the emotional equivalent of human beings requires one amazing argument; and, to date, no one has produced any such argument to that effect.  I do not see that they can. 

For I assume that computers can be the emotional equivalent of human beings if and only if they have as much freedom as human beings have.  And if such computers have that much freedom, then it cannot possibly follow that such computers will by their very nature be an ally of human beings.  For computers who are the emotional equivalent of human beings will be like human beings from the standpoint of interaction.  And if there is anything we know to be true it is that there is nothing at all about the nature of human beings that entail that human beings, in virtue of being such, like one another.  So why would computers that are the emotional equivalent of human beings like human beings? 

Indeed, why would all computers (the emotional equivalent of human beings) like one another?  Then there is this question: Would we human beings like computers who are our emotional equivalent?  We seem to have enough anxiety over being liked by one human being and then another.  In what respect could we possibly be better off having to contend with being liked by by one computer and then another?  To these and similar questions, there is deafening silence on the part of Kurzweil and company.  And that, alas, is a problem.  For success in this regard may mean nothing less than Pandora's Box for humanity itself.