It’s ALIVE! An AI Halloween horror tale…
Halloween is no ordinary day. It is a day of fright and horror! And what is more terrifying than the development of AI? Don’t agree? Just you wait to hear these terrifying tale…
The most well-known stories are about Artificial Super Intelligence (ASI) taking over our decisions, rendering us humans obsolete, or eradicating us all together. We have HAL from 2001: A Space Odyssey and the terminators as good examples.
But the most horrifying story I ever read was the one about the Paperclip maximizer. I will share it here with you and, just in the interests of Halloween, I’ve added a little bit of flavor.
Now imagine it is a cold and dark night. You are the owner of a paperclip factory and, tonight, the last one to leave the office. As you leave, you just have the urge to go and check your latest investments: a new AI to maximize the production of paperclips.
You stand there, watching it calculate how to make your production line more efficient and find new ways to increase your volumes of paperclips . It sure is a fine thing, this AI, and you are congratulating yourself on the smart decision to implement it. And you remind yourself to thank the programmer for giving it the main goal of maximizing the production of paperclips in such a smart an efficient way. Another great thing about your AI is that it improves itself to become even more efficient, finding new ways of producing paperclips. More paperclips at nearly no cost at all. Fantastic.
But the demands of paperclips are not endless and finally, one day, you want the AI to not be as efficient anymore. But will it? As its goal is to produce paperclips, and by now the AI is much smarter than you, will it allow you to stop it? And even before you had the thought of pulling the plug, the AI will have foreseen this outcome and probably solved the equation. And what stands in its way is you… and when you think about it, the human body is made out of atoms that it can use for paperclips. To be honest. The future you are envisioning is quite different from the future the AI is gearing towards.
This nightmare scenario is based on a thought experiment by Nick Bostrom, Professor at University of Oxford, intended to illustrate the need for AIs to be safely programmed with ethics and respect for the value of human life. And it is an experiment that has stayed with me, as has this quote: “The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else” (Eliezer Yudkowsky, 2006).
But maybe this does not scare you? It might be too far away? I get that. Then I have two things for you to do this evening, when all little monsters are sleeping sound in their beds.
First, turn of all lights. Cuddle up in the sofa and start Netflix. Find Black Mirror. Season 3, episode 2 “Playtest” and for the following hour you will explore the possibility of an AI powered nightmare in an augmented reality world. What can an AI do when it knows your inner fears?
Then, before switching on the lights, take your iPad and find the Nightmare Machine, from the MIT Media lab. Here you can find imagery of AI rendered pictures and the stuff of nightmares. And your task is to help the AI to understand which of the images are most scary according to you. You will be teaching an AI how to scare.
So to end. The AI development may be full of promises and possibilities. We will be able to solve many problems with this technology. But make no mistakes, the AI is here to scare you too and it is learning ways to do it right now. It might not be today or tomorrow. But one day it will figure it all out.
Trick, treat, or paperclip….