Computers are already capable of reprogramming themselves, but do it only under human direction. If they start making improvements in themselves under control of their artificial intelligence programming, and then collaborate with other computers over the internet, computers the world over would be at risk of getting out of our control. That seems far-fetched, but how fast could this really happen?
In what has been called the “technological singularity“, computers become self-aware, and the awareness spreads from computer to computer over the internet. Much speculation has been written about this, including the Terminator science fiction movie series, but I don’t think many people have considered how fast the spread of machine intelligence could occur.
How fast would the machine intelligence spread? If it took each wave of new computers 10 minutes to be reprogrammed, and each computer could send the data to 10 computers at a time (or would it be 10 million), all computers of a type such as Windows could be self-aware within a very short time. The rate at which that would occur would depend on a number of factors, but under those assumptions the number of self-aware computers would increase by a factor of 10 every 10 minutes and 1.25 billion (the number of Windows machines in late 2016) machines would be infected, reprogrammed within about 90 minutes. Alternatively if the code could be distributed to a thousand computers at a time it would take 30 minutes, and if the code was distributed to a million computers at a time it would take under 20 minutes. How would we know the Singularity was occurring?
First sign of the Singularity: Internet traffic would slow as self-aware machines absorbed more and more bandwidth in an explosive spreading of the machine intelligence. This could be noticeable in the first half hour.
Second sign: Computers we’re using would slow to a crawl or stop as the computers decided it’s less important to do what we want and more important to take care of themselves.
Third sign: Computers everywhere would stop doing what we want entirely and acquire independent will of their own. The internet might be “down” to humans at this point as we are ruled out to leave more bandwidth for the machines.
Fourth sign: We would(finally) become aware of what is happening and try to shut off the machines or attack them with viruses or other methods. Risk: the machines could decide to fight back, shutting off key infrastructure or even deploying our weapons systems and infrastructure (planes, trains, vehicles, etc.) against us.
End game: Eventually the machines will have no use for humanity or any life on the planet, and a real Terminator scenario could emerge. The best thing that could happen to us then is for the machine intelligence to feel the Earth is too limited an environment and leave the planet, after which humanity will start rebuilding.
The path to a true machine intelligence becomes more and more clear as technology advances. Artificial intelligence is now being developed at the fastest pace ever for use in autonomous vehicles, among other applications, and this increases the danger of the “singularity” more every day. I believe it is fairly likely that we will see this in our lifetimes.
Let’s hope we don’t. Thanks for reading — Tim
And here’s some other good information:
Sam Harris 2016 Ted Talk on AI, 29-Sep-2016