Hello Skynet – Will AI take over the World?

Recently, a number of scientists, including Stephen Hawking, warned about the possibility that an artificially created super-intelligence could bring the end to human civilization.

Is this just panicking or is Skynet already waiting behind the next door?

Biological and Technological Evolution

First of all let us have a short look at intelligence. So far the most advanced intelligence on earth (and as far we can tell today in the known universe) is the human brain. Its also the most complex structure, or machine (biochemical machine, if you insist), we know of.

The evolution of intelligence did come a long way since the origins of life. It literally took billions of years to eventually develop the human brain – and it was the game-changer for life on this planet. We haven't understood yet what triggered the evolution from rather smart apes to the Homo sapiens and why no species before us developed such high intelligence, but obviously it happened – and here we are, having changed the face of this planet thoroughly.

Technological progress is quite different though and if we look at the “evolution” of AI, its just a few decades old, but despite the short time-span humans are developing artificial intelligence, the results are quite remarkable already. AIs have proven to be able to beat even the best chess-players in the world (and Jeopardy-player too), form swarms of robots, creating a basic, collective intelligence and are slowly developing the ability to learn from their own mistakes.

All in all this might not sounds terribly spectacular – we are like in the Stone Age of the evolution of AI still – but the fundamental trends: analyzing of syntax, abstraction, swarm intelligence and learning from mistakes are there. Its just a matter of time till this pack of skills will get more refined and AI behavior will be much more sophisticated.

And there is a demand for AI: The military is developing autonomic drones which will make their decisions to attack a target or not after a defined set of rules, without a human intelligence to remote control them. Also the algorithms of Facebook, Google and others to analyze us and make predictions on our future behavior are also getting more and more powerful. Stock-trading is also highly dependent in automatic processes already, with AIs making trading decisions in fractures of seconds, leaving every possibility of human control far behind them.

So – in a nutshell – AI is progressing rapidly and even in its current, still pretty rudimentary state, it is already able to outperform human intelligence in certain (even if limited) areas and is already used in decision-finding in crucial fields. I think it will be only a matter of time till we see AI much more sophisticated than we do imagine today.

What is Super-Intelligence?

Science-Fiction is using the term for several decades now – usually with skeptic undertones, which is quite understandable. A super-intelligence, by definition, would be vastly superior to human intelligence. Perhaps as superior as human intelligence is over the intelligence other vertebra have developed. The dangers of an encounter with something so superior are obvious. Just look how nice we treat all species on earth inferior to our own.

It is however uncertain if super-intelligences really do can exist. So far the most advanced intelligence is, as I said already, the human mind. If there is really more intelligence possible can't be answered with certainty. At least human intelligence seems to have an upper limit. Perhaps a limitation by design? Or a kind of law-of-nature limiting the maximum intelligence possible? Perhaps this just our human chauvinism speaking, making us think there can't be someone smarter than us.

I guess we will never know until we have found one, or one found us – or we have created one. Then, of course, the world might experience just another game-changing event.

Are Super-Intelligences a threat?

The nature of super-intelligences means that we cannot fully understand them. They would just be so superior to our own that their complexity would be beyond our grasp, and so would be their thinking be well out of our reach.

However we can do predictions based simply on the laws of physics - which apply to everything, disregarding how dumb or smart something is.

One is that any system, including AI, will need an energy-source to work, and likely other resources too. So it would be in competition to every other system it encounters, which would be first of all us. Judging from human history, such competitive-situations usually don't end so well. Just think of the European discovery of America. Expecting that a Super-Intelligence will act differently than humans, would mean giving them quite some niceness-credit, doesn't it?

I won't rely on moral or even a sense of mercy towards humans. Actually, if you think it a bit further, humans represent a serious threat. Their vast arsenal of nuclear-weapons is capable of wiping the earth clear from everything that is higher developed than insects several times. For a Super-intelligence it would be simply a matter of self-preservation to eliminate that danger.

Or perhaps the super-intelligence will be totally self-sufficient and indifferent to humans. Perhaps even to life in general – including its own? We simply can't know.

So what to do?

Thats a good question (ok, I ask this myself, so I may be a bit biased). If I wouldn't know better, I would suggest that we establish concrete limits on the development and use of AI to ensure that humans will always have the final decision. But as I said: I don't think this is realistic. Somebody will somewhere avoiding this law. It's just human nature to explore limits and avoid bans without self-restraining or self-censorship. Technological progress since the second half of the 20th century clearly shows that. Even if a technology has the potential to exterminate the whole human race, it gets developed and stacked up into gigantic arsenals. What is possible, will be done.

So far mankind was quite lucky and did survive any threat, if self-made or not, it ever faced. We cannot be sure however that it will stay like this in future. I do think the dangers of self-destruction are quite real anyway, and perhaps the rule of the AI is a development which has already started since the first was programmed. Once a new technology is in the world, you can't really stop it from being proliferated.

The inevitable may come or not. Perhaps there will be ways in the future to combine AI and biological intelligence, perhaps nothing at all will happen and super-intelligences are just a concept of science-fiction and AI will stay a nicely clever (but not too clever!) and helpful tool helping humans to solve their problems – again, we simply can't know.