It was a year ago that I had a disagreement with my roommate on what AI can and cannot do, and will it even ever truly exist. I said that it was impossible to create a virtual brain, and he claimed I needed to take a look at how fast it's getting better. So how do I see it now, a year later?
We often get confused by the amazing rate technology develops. Sometimes it feels complicated and unknown, and it usually is. It is harder than ever for a single person to take a product, be it a smartphone or dishwasher, and figure out how every little part inside of it works. In the past it was much easier to reverse engineer anything.
And technology is only about to get more complicated. The thing is, when you add a bunch of features into something, your customers keep expecting even more, be it increase in usage simplicity or speed. The beauty of capitalism.
I know this might get many negative reactions if it reaches the "post-capitalism" masses, but let's get started.
Your most probable reaction after reading the title:
So, why would you even read this?
You see something in the news and keep thinking it can only get more advanced. After all, that's what we got used to in the last few decades. It's basically the same story with the Artificial Intelligence. You've just read an amazing article how in future "there won't be much work to do, you'll just get everything as you imagine it", haven't you?
I've felt that way, and I know where you're coming from, but unfortunately (or fortunately) things are not that simple with AI. However, the real literature on the topic is pretty much absent, so that's why I'm writing this. At least for the people who want to hear the other side of the story.
While algorithms are getting smarter and computers are getting faster (which by the way is amazing) you can't make them intelligent. But that doesn't mean all the efforts to get there suck. We can really make a lot from it, but we should never make a religion out of something like people tend to do.
So, let me start with this: I love AI and I'm super interested in it, but it can never really truly exist. Yep, I'm even more sure about it. But why?
Let me start with this: Human brain is NOT a computer
Do you remember when you asked yourself if you were just some kind of an advanced robot? I certainly do, it's what happens when you spend too much time on the Machine Learning course on Coursera :)
The more advanced stuff around us becomes, the more we tend to compare it with things we're used to see before. So naturally, the same goes for Artificial Intelligence. However, I don't want to get too technical in this text.
A few months ago I found an amazing article that made me realize a few things about brain. We have no idea how it works. I definitely recommend reading it, even if you disagree with me.
Forgive me for this introduction to computing, but I need to be clear: computers really do operate on symbolic representations of the world. They really store and retrieve. They really process. They really have physical created memories. They really are guided in everything they do, without exception, by algorithms.
Humans, on the other hand, do not – never did, never will. Given this reality, why do so many scientists talk about our mental life as if we were computers?
- Robert Epstein: "The empty brain"
In short, people always tend to represent the way brain works based on current technology.
A long time ago, people didn't have smartphones, they had only basic tools and materials. So it's not a surprise that in Bible, the first men were formed from clay or dirt, as the author recognized.
After that, there were many other theories, including fluids called 'humours' being responsible for everything human. Much less distant are the ideas that humans are complex machines.
By the 1500s, automata powered by springs and gears had been devised, eventually inspiring leading thinkers such as René Descartes to assert that humans are complex machines.
- Robert Epstein: "The empty brain"
But right now it has been taken to a whole another level. We're finally right! Or not. We're at the point where we think of the brain in terms of computation ability. We tend to put most of our efforts to figure out how it processes information.
Just over a year ago, on a visit to one of the world’s most prestigious research institutes, I challenged researchers there to account for intelligent human behavior without reference to any aspect of the IP (information processing) metaphor. They couldn’t do it, and when I politely raised the issue in subsequent email communications, they still had nothing to offer months later. They saw the problem. They didn’t dismiss the challenge as trivial. But they couldn’t offer an alternative.
Again, I highly recommend reading the article.
We tend to express our reality through complex mathematical models, and that's exactly what's happening here.
It's silly that we talk about simulating brain and algorithms of life when we cannot even figure out is there a mathematical model that could describe the brain, and even if there was, could we ever build it?
We really still have no idea how the brain actually works, what is consciousness and how exactly we're making decisions, how we store memories and how we process everything. Therefore, by definition, we can't really make a generally smart computer program. But what we can do is sort of "train" algorithms to do a thing we want them to do.
How smart can an algorithm actually be?
Welcome to the technical part of this essay. Don't be scared though, it's just a simple story. And it's really short :)
How do we make an algorithm learn? We feed it a lot of data. And it's usually A LOT. That's one of the reasons data is so important on the Internet nowadays. When we collect enough data, we can easily analyze it and draw conclusions. We can do it for several reasons, but most often those reasons are:
Prediction
There are many ways we can make an algorithm predict things. However, those predictions are not certain. Nothing in life is certain, get used to it.
But what we can answer relatively well is "How likely is that the customer is going to like this book/video/whatever?", "How likely is that this user will type this word next?", "How likely is that John has just said 'cheese' into his phone's microphone?", "What is the price of this apartment relative to every other apartment with similar features?", "Is this email likely to be spam?" and so on.
Notice "how likely" part of the questions.
Classification
Once we feed the algorithm with data, we can label each single data point. After we've done it, we can predict the class of new data. Let's say we have tons of data describing apartments. Things like size, location, distance to cinema, whatever really. After that we classify the data.
So if we divided the apartments into classes like luxury, regular, cheap, given a new apartment computer would be able to tell us what class the apartment belongs to.
Classifying cats and dogs based on height and weight. Image Source: https://kevinbinz.com/
Simulation
Think of things like autonomous driving and text-to-speech. What's happening behind the scenes is a complex mathematical model determining how the algorithm reacts to the current sensor data. But first, you feed the algorithms with a ton of data.
To teach a car how to drive, you basically need to drive it yourself and let it analyze the data and let it build a mathematical model that will be used in autonomous driving. The more data it gets, the better it will drive. However there are still some pretty hard problems remaining that are related to fully self-driving cars.
There are more applications, but I think these are enough for this text to give you a broad sense.
As you can see, it's all complex math. It is extremely beautiful, useful and creative. The future is beautiful as many things will tend to be automated, it's a natural process.
Two ways to look at future of jobs
Technology will make your job obsolete
There is an interesting book I read a while ago, it's called The End of Jobs: Money, Meaning and Freedom Without the 9-5 by Tyler Pearson. The book presents many interesting concepts.
The author concludes that you can automate many things, and basically everything that can be will be automated, no matter how hard people resist. But one thing is always going to be monopolized by humans, and that's creativity in terms of figuring out new ways to solve problems. So instead of worrying how to stay in our comfort zones while enrolling children in brainwashing facilities, we have the opportunity to make a real change.
Basically, he concludes that if you have a job that is simple and you feel like there is a recipe for everything you do, beware, because computers can, and the chances are they will, make your job obsolete.
“Entrepreneurship is connecting, creating, and inventing systems—be they businesses, people, ideas, or processes. A job is the act of following the operating system someone else created.”
― Taylor Pearson, The End of Jobs: Money, Meaning and Freedom Without the 9-to-5
Technology is like a tool, not a worker
Smart algorithms, no matter how smart they get cannot solve problems on their own. Peter Thiel, founder of Paypal and venture capitalist, reached this conclusion in his book Zero to One: Notes on Startups, or How to Build the Future.
“The most valuable businesses of coming decades will be built by entrepreneurs who seek to empower people rather than try to make them obsolete.”
― Peter Thiel, Zero to One: Notes on Startups, or How to Build the Future
What Thiel concluded was that no matter how much everything tends to get automated, in some (often too many) cases algorithms are just wrong and unreliable. He writes about the scam filters they developed for Paypal. What always happened is that people learned how to cheat the system and there were many false positives when they tried to automate the entire process.
The optimal solution turned out to be collaboration between technology and humans. They made filters that would flag possible scams and then professionals would analyze the data and reach the conclusions.
When it comes from a guy who's behind one of the most powerful data analysis systems (Palantir), he knows what he's talking about. Even though they do seem to have some issues, those are just showing how hard it is to work with such systems in this point of time.
The truth might be somewhere in the middle
In my opinion, there are jobs that will get completely replaced. Think of drivers of horse cars. The way world works is that systems become obsolete and get replaced. It's all about what Nassim Taleb calls the black swan events.
“You never change things by fighting the existing reality.
To change something, build a new model that makes the existing model obsolete.”
― R. Buckminster Fuller
However, I don't think there is anything to be worried about, quite the opposite. What's common in both views above is the entrepreneurial spirit. Just embrace entrepreneurship and realize the true actors of changes are entrepreneurs. Entrepreneur is a person who is forced to solve problems in new, different, innovative basis on daily basis. Entrepreneur's job definitely cannot be automated.
Imagine you had time to do more in less time, actually, you don't even have to imagine it, just look around you. That's what will certainly continue to happen.
Think about it. Read. Learn how to be creative, because until you do, you might be replaceable.
What's next?
I wrote this part as an introduction for anyone interested in AI and technology in general. What I'm going to do next is taking a look at economical concepts and analysis of the thing future brings and why it is capitalism and has nothing to do with socialism. Stay tuned!
---------------------------------------------
I'm really thankful you read this, feel free to follow my blog and read my older posts. For example:
I'm a 21-year-old Entrepreneur, Anarchist and Developer
Evil Entrepreneurs - Making a Moral Case for Capitalism - PART 1