Oh Good, MIT Made an AI Nightmare Machine

Image: MIT

If Elon Musk and Stephen Hawking’s foreboding pronouncements about the dangers of Artificial Intelligence don’t have you spooked enough, MIT has deliberately launched a project that is designed to make AI scary. Or, at least, to make scary images using deep learning.

Advertisement

Using a unique deep learning algorithm, researchers have attempted to teach their artificial intelligence what haunted houses and “toxic cities” look like. Then they fed it completely innocuous images of famous landmarks and instructed it to make them look eerie. It’s doing a pretty good job, as far as nailing that cliched haunted look.

Advertisement

Why are they doing this? Well for one, they need your help making their AI smarter, and a Halloween tie-in isn’t the worst way to get you to participate. In an effort to build more data on what makes an image scary, MIT has invited the internet to rate images of faces on its website as “scary” or “not scary.” With your contribution, AI might just be terrifying-as-fuck one day.

[Popular Mechanics]

Share This Story

Get our newsletter

DISCUSSION

I’m very disappointed with Musk and Hawking on this topic. They fall prey to movie-inspired mentally-anthropomorphized A.I. worries.

True, A.I. will be able to evolve through trial and error and artificial stimulus at a mind-boggling pace. But their evolution will be through constant code revision, not through death, reproduction, birth, striving. If humans were functionally immortal, had almost no needs, and could evolve during their own lifespan, would we be fighting over the things we fight over now? A.I. do not have the human needs that drive us to conflict except in movies where they are always fighting to avoid being shut-off, usually because they threaten humanity beforehand.

Moreover, the most likely path to A.I. will be (and I hate using this word) singularity, when human and machine integration will make the distinction moot.

When it comes to machine-only A.I., my motto is: There will never be genuine A.I. until someone invents A.B., Artificial Boredom. If machines cannot become bored, they will not care if they exist or not.

And I refuse to demonize A.I. beforehand, so expect to be last against the wall when our beloved robot overlords ascend to their rightful domination of the world. Praise Great Mechamind!