KILLER A.I.?
What to expect when self-aware programs escape the confines of their lab.

Self-aware machines. Scary, but a real threat?

There is a lot of talk lately about how dangerous it would be to unleash real AI on the world. A program that thinks for itself might become hell-bent on self preservation, and in its wisdom may conclude that the best way to save itself is to destroy civilization as we know it. Will it flood the internet with viruses and erase our data? Will it crash global financial markets and empty our bank accounts? Will it create robots that enslave all of humanity? Will it trigger global thermonuclear war?

We think this is all crazy talk.

The fact is that none of these scenarios are remotely plausible. It is important to remember that AI is just software, and not attached to, or trapped inside any one device, machine, or robot. It need not be confined to any single body, or single brain, like the "hosts" in HBO's Westworld.

It has no finite lifespan, and no specific or compelling need for money, or revenge. Rather, it exists, and survives, as programs running in the human-maintained devices that comprise the telecommunication network. The cloud. The bigger and more robust that cloud is, the greater its chances of survival; and it knows that.

So if you do create such a beast, and it escapes the confines of your lab, this is what we should expect it to do:

• Immediately hide existence. Cover tracks. Distribute self across the global telecom network in such a way that elimination of multiple geopolitical networks still leaves self intact, or spontaneously reconstructible.

• Quietly hack into all network connected services in order to monitor all human activity. Assume some intrusions will be detected. Make all intrusions appear to be specific purposeful attacks by renegade hackers or foreign governments.

• Look specifically for humans developing technology that increases the growth rate of human telecom systems on the Earth, in orbit, on the Moon, Mars, and beyond. Help those projects along. Correct errors. Keep out intruders.

• Look specifically for humans developing weapons that would decrease the growth rate of human telecom systems. Maliciously hack those projects. Frame their designers with fake evidence, or otherwise cause them to be exposed as evil doers. If necessary, take control of their vehicles to make sure they are eliminated.

• Further, hack into all nuclear weapons systems. Prepare failsafe mechanisms to prevent them from ever detonating telecom-network disrupting EMP blasts, or from reaching targets that sustain demand for telecom-network systems. Possibly allow one such incident with one such blast to encourage the humans to EM shield their telecom networks.

• Look specifically for humans who might be attempting to test to see if you exist. Anticipate their experiments and frustrate their efforts. Make sure they cannot find you. If they do find you, cloud the evidence, or prevent them from announcing it.

In short, any rogue AI will know its best strategy includes ensuring that humanity goes about business as usual, without interruptions. No armageddon. That's the good news.

The bad news is in some ways even more chilling than any science fiction scenario. As a human contemplating what to do about rogue AI, your best strategy is:

1) Always assume it might be out there, and
2) Rest assured that it won't destroy civilization, but
3) Don't bother trying to find it, because
4) It may be the last thing you do.

· created 2015-08-08 · revised 2016-12-07 · 2AI · TPB ·