I didn’t worry about AI taking over the world until I learned about this
One of the smartest people I know; David Hernandez, showed me something being done in the machine learning community that gave me goosebumps.
My experience with machine learning, up to this point, had been that there is tremendous value in helping us humans do better at our jobs. We use it to help us find photos amongst thousands, create closed captioning for our YouTube videos, and help us decide which things on Netflix we’re going to watch next.
I didn’t worry that AI was going to become self-aware and take over the planet. But now I’m starting to get a bit more concerned.
There is something out there called GANS, which stands for Generative Adversarial Networks. Great name guys… not menacing at all!
(over simplification warning) It is essentially two neural networks that compete with each other, and in so doing, train each other to be better models. Sounds like NOTHING COULD POSSIBLY GO WRONG. What you can end up with is a computer generating photos that look very real to humans, but are completely fabricated.
What are those photos above? Are they pictures of bedrooms in an apartment? Look close. They’re 100% computer generated, as in, those are not real photos of real bedrooms. How about these people?
Those are not real people. Just what a computer thinks real people might look like.
You might thinks that these look incredibly fake, and they do, but if you’re not concerned then you’re missing the point. GANS has only been around since 2014. Imagine what will happen in the next 10 years? You’ll see videos and photos that you’ll have no way of knowing is real.
There are some positive outcomes as well. Lots of people are exploring using this capability to complete photos where information may be missing. You might be able to extend that to archives of ancient material that have degraded, recreating architecture, or predicting the locations of things in the cosmos (I’m just spitballing here).
Either way, we must all be aware that computers are soon going to be able to mimic a human’s voice and face. Do what you will with that information!
