Intelligence
“Intelligence has been decoupled from consciousness“— Yuval Noah Harari, Homo Deus, 2015
What is intelligence?
In school, we measure intelligence because it’s easier to measure than traits like character, grit and creativity. It is measuring the ability to solve rote problems step by step — this is easy to measure, which is why we care about it in education and also why we can replicate it in machines.
In the book Life 3.0: Being Human in the Age of Artificial Intelligence, Max Tegmark defines intelligence as “the ability to solve complex goals.”
And humanity has now solved a very complex goal. We have taken our own intelligence and infused it into matter.
As Tim Urban explains:
Building skyscrapers, putting humans in space, figuring out the details of how the Big Bang went down — all far easier than understanding our own brain or how to make something as cool as it. As of now, the human brain is the most complex object in the known universe.
Where is it used?
Everywhere around you, there is already Artificial Intelligence. Perhaps this very article was recommended to you because of Medium’s data on other similar articles you’ve read and enjoyed. Amazon, Netflix, YouTube, and Spotify are all examples of recommendation engines built using AI.
Image Recognition, Facial Recognition, Voice Recognition —all of these rely on the ability of machines to learn.
Your Nest thermostat, autocorrect, chatbots, email spam filter, your Facebook, Instagram and Twitter feeds are all served to you using Artificial Intelligence. Even your Google search results will be personalized using Artificial Intelligence (your search results for the same phrase will be different from mine).
And this is just the beginning.
Data
Data is the powerhouse of Artificial Intelligence. AI, put very simply, takes in reams of data and finds patterns in it.
When you are looking for something to watch on Netflix, it knows what you’ve watched in the past. If you rated it afterwards, it knows if you liked it (and probably even knows your level of satisfaction if you stopped watching at some point). Millions of other users have watched that same show. If you and 5 other people watch the same 10 shows, you will be bucketed in the same profile and based on what you all like as an aggregate, you will get recommendations. They categorize us, and learn what we might like over time with more data. The more shows everyone watches and the more ratings they get the better the system becomes.
At this point, the data is far more important than the programmed instructions for giving you good recommendations.
From the Wired Article, the End of Code:
If all this seems a little familiar, that’s because it looks a lot like good old 20th-century behaviorism. In fact, the process of training a machine-learning algorithm is often compared to the great behaviorist experiments of the early 1900s. Pavlov triggered his dog’s salivation not through a deep understanding of hunger but simply by repeating a sequence of events over and over. He provided data, again and again, until the code rewrote itself. And say what you will about the behaviorists, they did know how to control their subjects.
A simple example
But how do computers ‘learn’? There is a great example from an excellent free, online book.
It’s extraordinarily easy for a human know what this string of numbers is. But only recently have computers been able to solve this problem (think of the Captchas you have to enter in websites to show them you’re human).
In order for a computer to be able to solve this, you feed it a ton of ‘labeled’ data. Given enough labeled data, it will use pattern recognition algorithms to identify which category to apply to each number — which it is most similar to.
This same idea has been used for facial recognition and voice recognition. Your Amazon Alexa can hear both you and your grandmother say ‘Alexa’ because it’s been fed millions of variations of the word Alexa to learn how to recognize it from all voices.
These pattern-recognition algorithms, especially the neural network, essentially model the human brain and how we solve problems, only they don’t get tired and emotional and can solve problems much more quickly. From the e-book:
Neural networks are one of the most beautiful programming paradigms ever invented. In the conventional approach to programming, we tell the computer what to do, breaking big problems up into many small, precisely defined tasks that the computer can easily perform. By contrast, in a neural network we don’t tell the computer how to solve our problem. Instead, it learns from observational data, figuring out its own solution to the problem at hand.
Learning
In Life 3.0, Max Tegmark explains how learning differs from simple increases in speed, storage and power of computers:
When IBM’s Deep Blue computer overpowered chess champion Gerry Kasparov in 1997, its major advantages lay in memory and computation, not in learning…When IBM’s Watson computer dethroned the human world champion in the quiz show Jeopardy!, it too relied less on learning than on custom-programmed skills and superior memory and speed. — Max Tegmark, Life 3.0, 2017
Watch this quick fascinating video (stop after the pong explanation) on how machine learning tackled pong. After 500 iterations of playing, it knew the perfect strategy to win in the fewest number of ball hits. Now computers are combining those previous advances in memory and speed with learning to be their primary advantage.
There is a fascinating documentary on Netflix (probably recommended to me through machine learning) that documents Google’s DeepMind team and their quest to beat the greatest human Go player:
There are many fascinating aspects of this documentary. For example, when the human player has to take breaks (which computers of course don’t have to do), the human Go master instinctively looks at his ‘opponent’ as if to read him (the opponent is just a human sitting there doing moves the computer tells him to). But the emotions involved are always separated from the machine: for example, the emotions of the entire human audience after the master’s first defeat. The humans who created the machine celebrating the victory of their technology over another human. And then our emotion when the human master finally wins a game (at which point even the humans that created AlphaGo seemed to be going for him).
Generalized intelligence
In the 1960’s and early 70’s before microchips (Intel 4004), we had special-purpose computers like calculators — computers that did one thing that led the way to current computers, which are more generalized.
Similarly, in the present, AI is moving beyond specific problems like winning Jeopardy, Chess and Go to AI that is able to solve more generalized problems.
But this is different from moving from a calculator to a computer. A computer still has to have each instruction step by step programmed into it. This is far beyond that leap, which is why it is a bit scary. Humans still controlled every aspect of programming before AI, but with a mind of its own we must give the computer the correct intentions.
In many ways a generalized learning algorithm could be seen as ‘the last algorithm’ (much like Virtual Reality can be described as the ‘last medium’) because it will be able to recreate all the other algorithms it needs by itself.
Human flaws
In chess, a computer can look at billions of game states in mere seconds and algorithmically determine the probability of a positive outcome of each one to make predictions (and get better with each move). IBM’s Watson can read every medical paper ever written. A self-driving car can see around them with a 360 degree camera, making 2 billion calculations per second while the rest of us check our phones.
How will we deal with this emotionally?
AI is likely to make us feel deeply flawed, and for the first time understand our vast limitations and irrationality. Yes, we have already created machines that are bigger, stronger and faster than us, but there were already animals that were bigger, stronger and faster than us. Never has there been anything more intelligent than us. Until now.
So far, we have been nature’s greatest creation of all time and we are in the midst of creating something better.
In previous generations, art has dealt with themes of the meaninglessness of human life, the insignificance of human life and perhaps we may now begin to tackle our uselessness.
