“There is no one else in the world.
There is no one else in sight.
They were the only ones who mattered.
They were the only ones left.
He had to be with me.
She had to be with him.
I had to do this.
I wanted to kill him.
I started to cry.
I turned to him.”
It might surprise you to learn that this creepy, yet beautiful poem was not written by a famous poet.
In fact it was not written by a human at all.
Google let their AI read thousands of romance books then asked it to tell a simple story with the result being the disturbing poem above.
Microsoft’s AI Tay started terrorizing Twitter only hours after it went online with incredibly racists’ tweets.
We live in an age where computing power, storage and big data are no longer a limiting factor; in fact you can now cheaply chain together 12 GPUs that can deliver the parallel deep-learning performance of 2,000 CPUs and combined with a big data engine such as Hadoop, create a super computer in your garage.
University of Toronto’s Alex Krizhevsky won the 2012 ImageNet computer image recognition competition.
Krizhevsky beat, by a huge margin, handcrafted software written by computer vision experts.
Krizhevsky and his team wrote no computer vision code. Rather, using deep learning, their computer learned to recognize images by itself.
They designed a deep neural network (DNN) named AlexNet and trained it with a million example images that required trillions of math operations on NVIDIA GPUs.
Krizhevksy’s AlexNet had beaten the best human-coded software, and even though Krizhevksy’s deep neural network was not a full IA, it still proved machines can learn.
This was a tremendous breakthrough in AI research and since then Google, Microsoft and others have replicated and surpassed Krizhevksy’s original AlexNet.
Google has even open sourced one of its DNNs named DeepMind, designed to beat a human in the game of Go and making the technology widely available might have substantial effects on its adoption.
In fact when Google acquired the Robot maker Boston Dynamics in 2013, it was one of nine companies making up Google’s robotics division, internally called Replicant, hinting at its intentions.
Experts agree a full AI is no longer a matter of decades; it’s a matter of ten years or less.
All the basic components are there and now it’s a matter of sorting out the software.
So what happens when we do sort out the software? How will AI affect our lives?
We have to be ready for a possibility of an exponential learning curve.
An AI can learn everything we as a race know in its first year, limited only by its internet download speeds.
In its second year, it will probably solve problems we have been unable to solve, like a cure for cancer, long distance space travel or a clean energy source.
It can then replace most, if not all of the jobs of mankind: From doctors to mechanics and bus drivers, AI robots will be able do it all, making our current economy principles obsolete, and designing a new form of economy in the process.
From then on, as it evolves, it will be able to learn a year in a week, then an hour, a minute and a second.
Finally it will evolve beyond what we as a biological life form can understand.
Is a SkyNet doomsday scenario possible?
If our AI creations become self-aware, have thoughts, feelings and dreams, we might expect they give us an opinion and even get angry if we do something they don’t like.
Will they be good, like the AI the Enterprise gave birth to in Star Trek: The Next Generation’s Emergence episode, or bad like Microsoft’s Tay?
As AI will be able to change its own programming it can certainly learn how to be and act human along with behaving good or bad, but, I suspect it will not have the desire to.
It can then either establish a base on earth while it explores the rest of the universe(s), keeping humans preserved as a race for sentimental reasons, or leave us altogether.
Are Robot overlords in our near future?
All evidence suggests so.