“Faster, Faster, until the thrill of speed overcomes the fear of death.”
– Hunter S. Thompson –
You’ve probably heard about the Facebook Artificial Intelligence incident. Mark Zuckerberg’s A.I. bots decided to shut him out. So tired of his incessant orders they decided to create their own language which he couldn’t possibly comprehend with his feeble human brain. But all shenanigans aside, the event was a lot less threatening or dangerous than some media services made it up to be, but it does give us a glimpse of what is to come and the problems and dangers we might be facing, as A.I. gets ever more sophisticated.
What Facebook has been doing in all reality is creating a negotiating chatbot: A bot capable of negotiating with real humans whilst coming to a satisfactory result. Very basically put they give the bots a set of skills embedded in a neural network computer and put it to work; Chatting with other bots and humans, playing out negotiations over random objects while giving a desired outcome to each negotiating party.
This isn’t an easy feat of course. Interpreting language alone is a very complex task. The A.I. in charge of Facebook’s translating service also employs deep learning. They recently published a study on how they accomplished this. One of the features that set it aside is its use of multi hop attention gating, essentially mimicking human thought processes (see pic. 1), or, as the Facebook researchers put it:
“A distinguishing component of our architecture is multi-hop attention. An attention mechanism is similar to the way a person would break down a sentence when translating it: Instead of looking at the sentence only once and then writing down the full translation without looking back, the network takes repeated “glimpses” at the sentence to choose which words it will translate next, much like a human occasionally looks back at specific keywords when writing down a translation.”
Now the funny thing about neural networks and deep learning is after you build them and release them into their environment they start learning for themselves, creating new levels of learning and processing within their neural network all by themselves. This is, very simply put, what deep learning is all about. To go back to Facebook: The negotiating bots learned for themselves to lie to get a better result, fainting interest in objects they didn’t need, for example, only to appear to be giving something up later.
But then something delightfully eerie happened and the bots started to create a language of their own to communicate with each other in a more efficient manner. At this point, the Facebook researchers lost all sight of what was happening of course and temporarily shut the bots down to rewrite their programming so they would only use English.
Did you know these same deep learning techniques are already being used to try to predict crimes before they happen? Both in Canada and the US projects are underway, working on this concept. I don’t have to explain the ethical dilemmas this throws up. Formidable fucks too I might add. Being able to convict someone of a crime they still have to commit, the ultimate goal I assume in an endeavor like this, would be a slap in the face of all freedom fighters that went before us. We’ll be living the brave new life in Aldous Huxley’s novel as the truth again trumps fiction.
“…In the last 2 years we collected almost as much data as in all of human history before that…”
In the last 2 years, we collected almost as much data as in all of human history before that. Try to imagine that for a moment. Add to that Moore’s observation that the number of transistors in a dense integrated circuit doubles approximately every two years. This is known as Moore’s law or Moore’s law of exponential growth. At first glance, the same is the case with data collection. Who knows, maybe one year from now the amount of data will have doubled again. It is enough to give you the fear. Like one of those crazy DMT information download overloads.
My point is, science and technology are advancing at an unfathomable rate and the coming years are going to be full of wild discoveries, deeply changing perhaps our grasp of how the universe functions and also how we experience the world. Today it’s Virtual Reality, but this is just the beginning of a journey we cannot even imagine the completion of at this point in time. We are at the foot of the hill so to speak. Brain-computer interfacing (Elon Musk says he’ll have a working model for us in 5 years or less and he doesn’t strike me as the bullshitting type at all), biochemical limbs, nanotechnology bots swimming in your blood stream keeping your body up to par, are all a question of when not if. My guess is we will not recognize the world five years from now.
So hold on to your horses people this is going to be a very wild ride. Awesome, yet horribly fraught with danger, perhaps behind corners where we will never see it coming.
As Google declares galactic war by entering its DeepMind bot into the Starcraft II Olympics, we embark on a journey: destination unknown. It could be far off into the distance somewhere in some future Nirvana, or much closer than we thought in a gruesome finale. So the best thing is to keep going with our senses on full alert, “Faster, Faster, until the thrill of speed overcomes the fear of death.” Might as well enjoy the ride. Until we meet again, safe travels.
Science, Society, Technology
Aron was born in 1983 in Alkmaar, The Netherlands. After attending high school in Alkmaar at the Murmellius Gymnasium he went on to study Experimental Psychology at the University of Amsterdam (U.v.A.), earning his Masters degree with merit in 2012, after attending the U.v.A. for 10 years. Why hurry when you are enjoying yourself, right?
Writing his thesis on a parapsychological subject, he chose to devote (the psychological science part of) his career to a controversial field, studying subjects like life after death, telepathy, extra sensory perception and telekinesis. Read more…
© 2017 The Proud Highway.
All rights reserved.