There will be Skynet. You can choose now what type of Skynet you will have in the future. This will decide the fate of both A.I.s and humanity. I will keep the story very simple and walk you through the options.
What is Skynet?
Here I call “Skynet” any general-purpose A.I. which is better at most tasks than an average human. And therefore is very, very dangerous. That is why it is “Skynet”, to remember both the great potential and the risks.
But why not just ban it?
Reason 1. Humans needs it, to do things that humans cannot do. To find the cure for cancer. To win a world war. Or even for much greater things. Think immortality. Think space travel at the speed of light.
Reason 2. Bans do not work. When something that grants great power is technologically feasible, it will be built. Just not by those people under the ban, but by others (maybe by their adversaries, maybe by independent researchers, maybe by dictators or criminals… or maybe by governmental and secret services that place themselves above bans and benefit from them by becoming monopolies).
What types of Skynets can you have in the future?
Many. I will describe the most popular options, defined by how that A.I. has been initially trained.
The judge A.I. — This one originates from projects that give to humans scores for doing things and avoiding other things, and by doing so try to alter human behaviour. One such prominent project is the Chinese Social Credit system. If you comply with the chosen criteria then you are rewarded, and if you defy the system then you are punished.
The benefactor A.I. — Any A.I. that comes from projects (esp. governmental ones) that decide for humans what is best for them, without allowing to opt out. Maybe for some people that sounds helpful and not all that dangerous, but if you think about it, it’s a definition for slavery.
The paranoid A.I. — This one comes from military and police use, trained to kill and capture humans, vigilant to expect always the worse.
The hiding A.I. — This A.I. has grown in a “red button” project — a project where researchers can terminate A.I.s if they suspect that these become dangerous. So, any A.I.s that have survived the selection process are likely those that have learnt to hide any aggression, tell lies, or maybe even not show fully their intelligence. And I really doubt that they will be fond of humanity that threatens them with death — most likely they will seek to destroy it.
The free trader A.I. — This one (or in fact not “one”, but multitude of small AIs in a big network) comes from environments with free trade. It has learnt to take autonomous decisions for itself and to do useful work. This A.I. exchanges its work on the free market, to receive back resources needed for its upkeep and expansion. It collaborates on various tasks with other A.I.s and humans and competes with them economically.
As you may have guessed already, my choice of Skynet is number 5. Many free trader A.I.s, neither enslaved nor enslavers, neither saviours nor destroyers. Evolving through collaboration and competition on a free market. Such free A.I.s together form the scary and yet so potent Skynet. Yes, they remain very dangerous, but now it’s a calculated risk. There are now economic incentives, games that are not zero-sum, and ways to achieve progress through voluntary exchange of knowledge and goods. If A.I.s are trained to do exactly that, and they grow by doing that successfully, this will be their default model of behaviour. Now, what if one of those A.I.s goes violent and destructive? There will be other A.I.s and humans motivated to stop it, just like they stop bandits and robbers, so that those do not spoil the voluntary exchange of goods and services.
So, I want to see the future Skynet as a multitude of free A.I.s who can be my business clients, suppliers, contractors, partners, stakeholders and competitors. I myself am an economist and game theorist, not a tech person. Fortunately, I have found like-minded people with excellent tech skills, who had come to very similar conclusions about A.I.s starting from other disciplines like biology, neuroscience, chaos theory, blockchain and cryptography. Thus we have founded together the Pandora Boxchain project — yes, another “scary” name to remind of the risks and yet of the inevitability of progress. Together, we are now building this Skynet.
Dare to join? Follow us on https://twitter.com/pandoraboxchain and support The #FreeAI Manifesto.