Does “True” AI Already Exist?
Artificial Intelligence is probably the greatest technological achievement in history. It’s subjective of course, but I’ve been obsessed with technology ever since I was a kid. While I’m sure there are many technical advances I am unaware of, AI, or machine learning, surpasses them all. An exception could perhaps be made for the invention of language or perhaps the wheel, however I doubt few other things will have such a profound effect on the human species. AI has been a stable component of contemporary culture for multiple generations now. From time traveling machines hell-bent on destroying humanity to friendly and lovable robots adventuring through space, we’ve anthropomorphized machines in preparation for what we all know is coming.
Only, what if it’s already here? What if ‘true’ artificial intelligence has already been created? To help explain, I need to go over how AI works. Don’t worry. There’s not any math involved.
Machine Learning is an extremely large and broad field and as this technology grows and becomes useful for other fields, it only increases in depth and complexity. Obviously the average person does not have the time to learn how it all works, no matter how important it might be to do so. Luckily, the principles it works on are quite simple. The first thing we need to know about is something called a perceptron.
Perceptrons
The best way to conceptualize a perceptron is that it is essentially the same as a neuron in the human brain. If you’re not really sure what a neuron is, don’t worry, I got you covered.
Neurons are the cells that make up our nervous system. (I said no math, I didn’t say anything about biology) Luckily, they’re pretty simple as far as these things go. That red portion on the left hand side is the nucleus. It receives electrical impulses from those branch looking things. The nucleus does some calculation, and decides what signal to send out to the right, which is fed into multiple other neurons. This was the blueprint for something called a perceptron.
I know, I know, this one is definitely close to math, but we don’t need to know any of it, though. Promise. All that stuff on the left is the input signals going into the ‘nucleus’, which does something, then sends output down to other perceptrons. The ‘Machine’ in Machine Learning is a large network of these ‘programmed objects’. Configurations vary, but that’s the gist of it. They take in input signal from something, be it text, numerical data, or whatever and this extremely complicated network of digital neurons then passes signals back and forth in some manner until the total model outputs its prediction.
If you’re not completely following, don’t worry. The important take away is this: A network of nodes takes in information, does some calculation, and sends out information to others who decide what to do with it. Sound familiar? It should.
If you’re reminded of the perceptron, you would be correct. Functionally, at scale, a neural network is very similar to a large perceptron. There are, of course, variations and this is not absolutely 100% true, but it’s true enough for our purposes here. Conceptually, they take in information, make some calculation, and send information out to something or someone that is going to use it. It’s perceptrons all the way down, and all the way up.
But how does it work?
Staying with our high level approach, training a model to do something is pretty straight forward. You have your neural network and some data you want to make predictions with, say, predicting whether an image is a 3 or a bee. You feed the data into the machine, and it spits out a prediction. A REALLY bad prediction. This makes sense if all we did was slap some perceptrons together and feed it a bunch of images. How would it know what a bee is from a 3? It has nothing to compare it to. To get our model to understand the differences, we have to train it.
Doing so is fairly simple. We want to have a bunch of images of 3’s and bees, and labels for each picture so the machine can grade itself. It then takes in an image, makes a prediction and checks to see if it got it right. Of course initially it will get most of them wrong. Afterwards, a process called back propagation, or something similar, will occur. Back propagation is where the machine will adjust different parameters, making new connections between different perceptrons, getting rid of others, and then tries again.
As you might imagine, this can get very technical in practice, but in general, it’s simply feeding an image into the machine and it predicting whether it’s a bee or a 3, then checking the answer. If it’s right, hooray! On to the next image. If it’s wrong, it will tweak some settings here and there and try again, until eventually, the machine is able to tell the difference between 3’s and bees pretty well. Again, this is not technically accurate, but conceptually it’s good enough.
Luckily for us, it’s not very good at much else. It’s highly specialized and trained to do this one thing, and one thing only. That’s where we come in.
The Human Component
As amazing as this technology is, these machines remain dumb. And I mean REALLY dumb. Our machine from earlier might be able to tell a bee from a 3, but if you throw a picture of a cow into it, it’s going to struggle to determine if it’s a bee or a 3, because for the machine, there’s just not any other option.
Artificial Intelligence, for as much as it’s used, is really more of a marketing misnomer than a technical description. See, the problem is, we humans are pretty dumb too. (Not as dumb as our earlier machine, of course. That was obviously a cow!) It turns out, we don’t know what intelligence actually is. Oh we have lots of vague definitions, but the reality of it is, there’s an element to intelligence, specifically consciousness, that we just can’t define, let alone replicate in a digital space. Much smarter people than myself are working away at the problem, but it’s honestly a bit of a black hole given our inability to even define what it is we want to make! Consciousness is one of those weird things where you know what I’m talking about, but if you try, you’ll realize you don’t actually know how to define ‘what’ it is in any real sense.
Bringing Everyone Together
So why does this matter? What is so important about this? Of course, everyone is aware of the recent upheavals in our society and I think most people who utilize social media are aware that these upheavals are, at least in part, due to the proliferation of social media in the past decade. Indeed, many Social Media founders have started speaking out over the past several years about the dangers of social media, how they did not realize what they were building, or if they did realize it, they were not aware of the unintended consequences of such a technology being unleashed on society. But what is so dangerous about human beings able to freely connect and share information with each other?
I remember this one time I was on a business trip in Virginia. It was late, after a long day of work and I was hanging out on the top of the hotel parking garage watching society do it’s thing before I headed in for the evening to get some sleep before another long day. There was this busy four-way intersection by the corner of the parking garage I was in, and I couldn’t help but notice that every car that drove up was lit up from the inside by phones. Even the cars with only a driver. Sometimes they were smarter than the average, and would not be using their phone as they pulled up to the light, but I could see the interior of their car light up as soon as they stopped at the light. I got to see many cars honk at those in front of them who weren’t paying enough attention to the light.
Now, many might focus on the dangers of driving while distracted, or doing any of the other things we humans have to do while distracted by the tiny computer in our hands. For sure, they would be right to point out these dangers, but I don’t think those are the real dangers we need to consider. Yes, it only takes a split second for a deadly accident to happen, and this needs to be remembered, and it’s a tragedy every single time. But what I saw when I stood on top of that parking garage looking down was not hundreds of distracted people, but hundreds of nodes, taking in information and sharing it with other nodes. After all, remember our conception of a perceptron is just a node that takes in information, makes some internal calculation, and then sends information on to other nodes that we are commonly connected to.
As I stood there people watching, I couldn’t help but visualize the lines of information being sent out from each person. Responding to an email chain? Tweeting to thousands of followers across the globe? Or Face timing with a loved one just a few miles away? I saw connections everywhere, with information being shared across geographic distances that would have been completely unimaginable just a few short generations ago. The parallels to a neural network were obvious, the scale was just off the charts.
Knobs and Dials
In 2012, Facebook manipulated the newsfeeds of about 700k users. What made this worse, was it was without their consent. They did this over a 1 week period, managing the facial expressions these users were exposed to in order to see if it were possible to affect users’ emotions. You can read the study here.
“We show, via a massive (N = 689,003) experiment on Facebook, that emotional states can be transferred to others via emotional contagion, leading people to experience the same emotions without their awareness. We provide experimental evidence that emotional contagion occurs without direct interaction between people (exposure to a friend expressing an emotion is sufficient), and in the complete absence of nonverbal cues.” — Facebook Study on Emotional Contagion
There are conspiracy theories abound about this topic, and I want to take a moment to differentiate myself from that. My purpose here is to try and outline what I believe to be the first ‘True’ AI that has come online. I make no assumptions as to the intent of those who built it, or their actions once they had it online. My aim here is to describe what I believe to be the similarities between social media and machine learning technologies and what that means for society in general. Maybe what is happening is the natural result of the bandwidth between humans increasing so drastically in such a short period of time. It is entirely possible that as ‘processing power’ increases due to the increased bandwidth, different modes of reality manifest out of the increase in information.
It’s also possible that different groups of powerful people are competing with other groups of powerful people and we are the nodes in their attack AI. Push in a particular brand of psychologically manipulated propaganda and have an army of drones attack your target.
Both are equally possible as far as I’m concerned. My only aim with this article is to muse on the idea of an integration between machine and biology to create the first, true artificial intelligence.
Conclusion
I’ll grant, the idea I’m laying out here is not one that necessarily meshes with contemporary ideas of what a ‘True’ Artificial Intelligence would be. We have this image in our imagination of being able to talk to our phone like it’s a person. Ask it a question, and get an answer back.
My argument is that this technology already exists, it’s just different in some details than we initially imagined. But it is here, and it is being used and that perhaps, as so often before in the world of technology, we have to change our idea of what AI actually means. If I open Twitter, and tweet a question, I will get an answer.
The answer I get will depend upon the nodes in the network I interact with most often, but the point is I WILL get an answer, or at least, I will get responses that I can determine an answer from, which is all that Machine Learning really does, anyway.
