John McCarthy is a living legend. Yes, this is a tired cliché and it makes you sigh and think ”oh no, not another one...” But in this case it happens to be true. Let's review the facts:
1) McCarthy coined the term ”Artifical Intelligence”. There. Ok? No?
2) McCarthy invented LISP. Now?
3) McCarthy has been a leading researcher in AI for more than 50 years. Get the picture?
4) McCarthy has, as a side project, started one of the most passionate and useful collections of arguments to prove that material progress is sustainable.
You know, we could go on. But lets drop the clasifications for a while. When I knew I was going to Stanford, one of my aims of course quickly became to meet with Professor McCarthy. He kindly assented and I had the privilege to spend two hours with him discussing the future of AI, sustainable material progress and science fiction. At almost 80 he is as mentally agile as ever, and is still working on research in AI, he has also recently started his first science fiction novel.
The future of AI
Professor McCarthy is confident that AI will continue to develop. Or, as he notes: ”There was a hundred years from Mendels discoveries until we charted the genome – we still have time.” This does not mean that he downplays the problems with AI, however. Brute force, speed, for example will not be enough. We definitely need something more than mere speed, he says. But exactly what that is, is not certain.
Now, mind: we have come far. Computers routinely beat great chess players. They create swarm-like intelligences in computer games. They create the vast majority of non-player characters in virtual worlds. AI helps with everything from word processing to traffic planning. The applications are out there and they continue to develop. But there are still things that we cannot do. One of the things that remains hard is to allow a computer program to fram context. Look at another, more complicated game, like the Japanese game of go (or Chinese game, depending on your historical views). Computer programs play very poor go, and the reason, professor McCarthy thinks, is that go forces the players to think about the board in regions and to subdivide and find reasonable regions to compete in. What regions are reasonable, however, seems to be a hard problem to solve.
Sure, in chess we have kingside and queenside regions – players subdivide the field into these regions – but it turns out that this division is unnecessary for a computer. A computer does not need to think about these regions at all because of the speed with which it can explore the tree of possible moves. Not so in go.
One of the current research projects that McCarthy works on is related to this problem. McCarthy is trying to develop a logic of contexts and approximate concepts. How do we work with these as human beings? How do we develop concenpts that are in a basic sense vague? We seem to be able to handled vague concepts with ease, and this increases our intelligence.
When I suggest that this seems like thinking from the later Wittgenstein, especially when Wittgenstein speaks about the use of concepts such as ”game”, McCarthy seems uneasy. He really does not like Wittgenstein (even accuses him, if half jokingly, of setting back the field of philosophy anout fifty years, singlehandedly). You know, he says, he even led Bertrand Russell astray. That is quite some achievement.
McCarthy has a strong belief in logics, and in a sense he is a purist. He does not believe that we need new forms of logic – like modal logic – or probabilities. McCarthy thinks that these areas have their uses, of course, but to him they are far less important than pure logic. This is also reflected in his recommendation to people wanting to study AI: he recommends them to study mathematical logic, and to read analytical philosophy (i.e. Russell before he was corrupted by Wittgenstein...)
This strong belief in logic also forms the basis of McCarthy's rejection of the different critiques of AI that have emerged over the years. He mentions Dreyfus, Penrose and others and when I ask him if he believes their critique has relevance for the field, and if there are things that human minds can do that are not translatable to algorithms he simple says ”no”. And there is an incredibly strong point here: the burden of evidence should, of course, be shifted to the person arguing that there are things the human mind can do that are not algorithmic. Becase what are they then? Penroses answer – that quantum-level qualities of the human mind makes it unique – simply sound to ridiculous, and it is not even clear how such an hypothesis could be empircally verified.
But the question of AI could still be obsolete. There are many other possibilities. One simple possibility would be this: before we have computer programs that achieve human level intelligence, the clear bordeline between human and other intelligence will have been blurred and it will no longer be relevant to speak about artificial intrelligenve. Rather, we will speak about symbiotic intelligence, connoting all kinds of intelligence networks that can develop. One simple such symbiotic intelligence – between humans – is Wikipedia. Now, Wikipedians depend in a high degree on Google so they use the advanced page-ranking mechanisms in Google to enhance their editorial capabilities. The end result is a mesh intelligence where the artifical and ”natural” components interact seamlessly.
Perhaps we will not discuss AI separately, and perhaps we will give up the idea of trying to acsertain the intelligence of computer programs. The question of AI may well be a category mistake: asking if a computer program is intelligent, may be lika asking if a human organ is intelligent. It is the networks resulting from interactions between different components – artifical or not – that is the relevant unit of intelligence.
McCarthy is reluctant to think so. He thinks that before the line between human and artifical intelligence blurs we will have developed some sort of artifical intelligence. He even thinks that there could very well be a graduate student out there today, who has solved the basic problems that still riddle the field. But we don't know yet. He defintely seems to think that AI will become a reality, and that we will have to deal with programs that are as intelligent as – or much more intelligen than – we are. This is a truly fascinating future...
The future of material progress
The second subject of our discussions are the pages that professor McCarthy has set up on sustainable material progress. McCarthy, who describes himself as an extreme optimist (one who thinks that things will work out well even if people do not heed his advice), has spent a tremendous amount if time with these pages, and the most populat of them have recieved a lot of traffic. The pages are the perhaps biggest collection of reasons to be upbeat about the future of the world that exist on the Internet. (Why does he do it? "The usual mixture of public spirit and ego").
I ask professor McCarthy why he is so optimistic. What made him do this? His answer is very interesting. He simple states: "I never stopped".
Then he explains that in the 50s everybody was this upbeat about the future, and as far as he is concerned there are no reaons not to be. McCarthy – in this sense – is a relic from another, much more positive, era when it comes to the view of the future. He thinks that human kind will continue to develop, populate the galaxy and expand it's powers indefintely to the borders set down by the first and second principles of thermodynamics. This is almost dizzying for someone like myself, whoi grew up in the shadow of the cold war, with imminent threats of global waming, comets and other prophesies of doom. But it is also liberating.
McCarthy supports his argument with facts, sources and a lot of data. But when I ask him if he has recieved a lot of criticism, he says ”No, sadly not.” He thinks that this is because his thinking is so deviant from the current thinking of the day that he safely can be ignored. This is probably true, but this speaks against not him, but more against our current zeitgeist.
Take one simple example. McCarthy is a strong proponent for nuclear power. He thinks that nuclear power and hydrogen will constitute a stable energy system that can sustain material progress for a long, long time. (Until the sun goes cold at least, and we will have to have come up with something else to live on by then). In his unbridled support for nuclear energy he is almost alone. Even those that argue in favor of nuclear energy today, do so from an argument of necessity. Peter Schwartz recently did this in a seminar held by the Long Now foundation – and his argument was that if we rely on oil we will have geopolitical tensions that will end in global war. So in a situation where the alternative is global war, nuclear energy could be ok.
McCarthy notes that this issue has become enormously ideologized. And the ideology is tainting the rational assessment of the technology, and this really makes him irritated. He cites numerous examples where nuclear energy is almost eliminated from discussions of future energy, just because it is controversial.
In this, and in the overall pessimism of the future, McCarthy finds no other explanation than the radical movement of 1968. The movement created a ”the-end-is-near”-mindset that has turned out to be incredibly hard to break – in part because the 68-generation now is in power in most media outlets, companies, public sector agencies and other places. The optimistic generation of the 1950s is, like McCarthy a generation emeritus.
But McCarthy has great faith in the future (though I sure that he would scoff at that description – it is not for him about ”faith”). He thinks that we will continue to develop and that the future will out. The risks – that a public ideology of doom creates a negative selective pressure on innovations, science and investments – are there, but they will be overcome.
Extreme optimism, anyone?