Would you rather fight a horse sized duck or a hundred duck sized horses?
What once was the topic of the famous 20th century author Isaac Asimov’s science fiction novels and short stories has finally started to become a more plausible reality here in 2018. I’m talking about, of course, the advancement of artificial intelligence (AI) to the point of their succession of human intellect. Elon Musk frequently takes to the news to state his paranoia of earth being taken over by computer overlords, and he’s not the only one. In some senses, he has a good point. Earlier this year, Facebook began to experiment with allowing two AI’s to communicate with each other, resulting in the two computers chatting in a language they made up before engineers finally pulled the plug.
Let’s wind the clocks back 75 years to the dawn of the computer era. In 1943, the United States Government commissioned half a million dollars to build one of the first programmable digital computers called the Electronic Numerator, Integrator, and Calculator (ENIAC) to solve artillery charts for the Navy’s ballistics lab. Fifty years after that, IBM’s Deep Blue AI became the first computer ever to beat a reigning international champion of chess in a six game series. Now, AI is used by every major tech company (Amazon, IBM, Google, Apple, Facebook, Microsoft) to analyze large data sets that are incomprehensible to a human being. Computers are employed in innovative ways to screen for new drugs, predict the weather, target advertising efficiently, run climate simulations, store large amounts of data, and virtually all modalities in human life.
But if computers are so widespread (apparently good at chess), why haven’t they taken over the world in infamous Asimov “I, Robot” fashion yet? The truth is that the world humans live in isn’t as simple as a game of chess, or even a large set of data and to be quite frank, electronic computers, as powerful as we perceive them to be, are just good at different things than the computations performed by the human brain. Even the simplest computers are superior to humans at math, raw data analysis, and other numerical computations; games or puzzles with simple objectives (i.e. chess, Chinese checkers, etc.); and reliable storage of massive amounts of data.
However, the human brain, on the contrary, is a computer that succeeds at interacting with other humans (I must have missed this trait), designing and creating new tools/devices, and assigning meaning to concepts (i.e. interpreting data analysis in a meaningful context).
Those qualities of both humans and computers shouldn’t be a surprise to anyone... I mean, we were all dependent on our TI-83 graphing calculators to get us through high school math, so a computer’s ability to do math shouldn’t be a stranger to anyone. But what can we learn from these obvious facts? Why is it that the first computers were and are better than us at math but it took ten generations of Iphones to recognize a face?
Computers succeed in tasks that require them to take in a finite amount of highly discrete or well-defined variables (i.e. numbers) and generate an equally discrete output (more numbers). Humans, on the other hand, are very good at creating associations between an extremely large number of poorly-defined but related variables such as the shape, size, color, and relative morphology of someones face and generate an appropriate behavior. Trying to recognize someone’s face regardless of the environment, distance, or context seems like a simple task to us, but to a computer, that task is extremely difficult to accomplish and is just now beginning to become a reality.
But even so, abstract tasks like facial recognition are just the beginning of what it takes to live as a human functioning appropriately in society. We are performing hundreds of these, and more complex computations every second. When you or I see a face, we are able to bring back memories, and not the cheesy “2 years of friendship” you are prompted with on Facebook. We are able to recall meaningful “episodes” and integrate them with current stimuli. Further, we use contextual queues, future plans, emotional influences, and associations of all those things we have made with that person to guide our behavior.
It is this very quality that allows us to assign meaning to numbers that computers output when performing computations or understand the implications of a medical diagnosis. Maybe that’s why math was so hard for us in high school... it’s pretty hard to associate meaning with your teacher’s arbitrary questions (you go to the store and buy 467 watermelons at a price of $16.45; so how many rhombuses does it take to beat an elephant in a footrace)?
While we may be slower than a computer when performing mathematical computations, we possess higher order emotions and complex executive behaviors that collectively feed into who we are as individuals and as a society. As such, employing one deep network (or even two that can talk to each other) to recognize a bunch of faces isn’t going to be enough to hijack the internet and overtake humanity!
And if that’s not enough to convince you that the human brain is a more powerful computer than our mathematically superior pocket buddies, then you may be instilled with a new sense of self confidence after learning that many of the AI designed to do “human-like”tasks (facial recognition, advertisement targeting, diagnosing diseases, etc.) are designed as “artificial neural networks”. That’s right, traditional computing methods undergone by our silicon slaves are so inefficient at being human that they’ve had to learn a trick or two from us to get their foot in the door!
So, while AI has officially superseded us in math, chess, and figuring exactly what advertisement to put in the little 3x5 box at the bottom right hand side of my Facebook profile, they are far from having the capacity to overthrow the world and rule over us as lord and master because they’re associative capacity pales in comparison to that of a human being. Computers can solve a million billion math problems and beat everyone in the world at chess… twice, but if they can’t understand why they’re playing chess or what it is, then they have already failed to best humanity.
Elon Musk can take a breath of relaxation knowing we can sleep safe and sound at night because we aren’t going to wake up to a silicon revolution. But, with the dawn of quantum computing exponentially improving the speed of our computers in the near future, it would be wise of humanity to tread carefully as we continue to explore the full potential of AI and how we can employ such devices to improve our own quality of life.