13 minutes

The Arms Race of AI 

It’s 7 am and you’re comfortably sitting at your desk, checking your emails for the day and browsing around on your laptop. The sun outside is still dim, and your mind feels groggy. You go to Amazon and check out the most recent recommendations they’ve provided you with. With a sense of unease, you realize how spot on almost all of them are. You’re an avid online shopper, so it actually makes sense. But still, you wonder what goes on in Amazon’s headquarters that enables them to give you recommendations so accurate. You go on to check your social media feeds and notice that the content that you’re generally more receptive towards, seems to be getting a lot more priority in your feeds lately. It’s as if all these companies are somehow starting to learn from the information they gather from you. It’s kind of creepy, but you know that there isn’t a person assigned to every user profile, trying to discern the content you respond to the most or the articles you’re most likely to buy. “That would be crazy!” you tell yourself, and logistically it wouldn’t make any sense. But then how does it work exactly? Obviously, there are programs in place which take care of this kind of stuff you think to yourself.  You’ve heard of  Machine Learning and AI before, you’ve seen a headline or two before about how these programs are starting to re-write themselves. Stamping this off as click bait shenanigans you continue to wonder how exactly this kind of stuff works.

Indeed, around the globe tech companies are ever more increasing their investments towards the effort of teaching computers how to “learn”. As artificial intelligence is declaring itself as a major force in almost all structures of our lives and is taking its roots in major hallmarks across the industry, the active investment into this young field can today be seen on almost every scale. Mainly automated customer service, as well as profile classification, are being impacted by this in our times.

Ready, set, AI

And as the bricks for the foundation of artificial intelligence are being laid, the wonder and fascination around learning machines grow ever more. Rapidly, the mysterious veil around the field of machine intelligence is being lifted, as we find ourselves in a golden age of articles and content, treating this subject with an almost childlike enchantment. But for decades already, the making of minds within the machine has been the source of awe for writers, filmmakers, and artists across the cultures of this world. Out of fiction humans have started to make science. We’ve managed to impose our eccentric dreams upon our reality, and out of a wild fantasy, we have made enterprise. This rapidly expanding trend is observed chiefly in the consumer market due to its wide-ranging uses in almost every aspect of the average consumer’s life. With custom news feeds, product recommendations, entertainment suggestions, and automated customer service leading the value creation of this brusque revolution, our lives are silently being surrounded by learning machines all around us.

In the world of tech, companies of every scale are sprinting with and against each other in this arms race of our times. What is at stake is not just one more piecemeal innovation, but control over what very well could represent an entirely new computational platform. With this seemingly inexhaustible gold mine, the shifting towards AI can today be observed, at the core of almost every major force within the free market. It is with ease thus assumed, that the granting of influence towards AI in our times, can reap solely benefits, making our lives easier and likely more pleasant.  The tech market recognizes this yet untapped value, and with the rash ambition of our generation, the big players of the industry are racing towards the perfection of this yet young technology.

An industry re-shuffled

With the rise of AI, the life of the average consumer isn’t going to undergo any drastic changes on the surface just yet. Apart from the self-driving cars already cruising around in the deserts of Nevada, some cool voice recognizing gadgets available on the market today, or some remarkably personal news feeds,  AI hasn’t yet completely overtaken our daily lives.

However, the brainy computers have already made their impact in the tech industry itself. With algorithms ascertaining the needs and wants of consumers from all backgrounds, programs which automate customer service, and taxi services which require no drivers, the monetary worth for the companies using these technologies is tremendous. We are fast approaching or are arguably already in an age in which value creation is no longer directly proportional to the size of the working force. And although the applications of AI are wide-ranging and hazy, the direction towards it is going is evident and direct. Automation, and the removal of human effort in almost all of its forms. It may sound a little far-fetched, but from a certain point of view, AI may be the answer to the utopia humanity has longed for ever since it has descended from the treetops of the African savanna.

Wake up ! They’re evolving !

Yes indeed the waking of minds within the machine, will be undoubtedly a deafening crescendo in the orchestra that is technological advancement, drum firing its adoptions across every major aspect of our digitalized lives. Carrying its benefits far beyond its more obvious uses, it has the ability to uproot the structures of how we perceive intelligence, sentience, consciousness itself, and thus ourselves as individuals. But as with all major technological leaps, the good has to be taken with the bad. Not all will be shiny and glistening in the golden age of  AI.

When probing the tech industry for any of its trends, it is next to impossible to get around the staggering giant that is Google. This monument of our times is one of the leading forces behind next to every major shift in the technology market today. And with AI emerging as a shifting force in our times, Google has naturally imposed its brand upon this rapidly expanding field.

Ahh they grow up so quickly…

With their house-made AI engine AlphaGo, Google showcased its baby to the world, dwarfing the human mind with their program in front of the entire globe to see. In the month of October 2015, this computer program developed by Google’s Deepmind in
London became the very first program to defeat a human in a match of Go, a traditional Chinese board game, magnitudes more complex than the ancient game of chess. It achieved this victory, for the first time in history ever,  without any imposed handicaps on the human player. But despite the revolutionary implications of this victory over a human achieved by AlphaGo, the first major defeat over a human mind by a computer does not go to the AI made by Google. The first stunt of this magnitude was pulled on the tenth of February 1996, when Deep blue a chess computer made by IBM defeated the Russian chess grandmaster, Garry Kasparov.  

And although these major events of the computer industry, which lie decades apart are very similar on the surface, there is a fundamental difference between the two. The subtle dichotomy between these two testaments of technology is slight, but not by any means to be ignored. Deep Blue, the machine which humbled Mr. Kasparov, was a machine, deliberately and solely designed to execute a single task. It played within the tight confines of the game of chess, and chess alone. Its abilities were hardwired, its algorithms immalleable, its “mind” linear.

If one were to release DeepBlue into any environment for which it was never intended, it’s deductive abilities would no longer seem all that impressive.

To the outside observer, these two efforts most likely looked extremely similar to each other.

Both Deep Blue and AlphaGo were able to dominate mankind in these ancient board games, crushing two human masters in their own respective domains.

But the programs and the underlying instructions within these two machines were vastly different from one another.            

The program brought out by Google, proudly dubbed AlphaGo, was taught to play the game of Go, in the way you might teach a human being to play the game of Go. Obviously so, on a different scale, and through different means of stimulus. But the concept remains the same. The machine learned the game through recognizing patterns and receiving stimuli. The rules of this complex board game were never written into its source code. And within here lies the main difference between DeepBlue and AlphaGo. The instructions lying behind this program were never designed for the game itself, but rather, were conceptualized towards the ability to learn.

Thousands upon thousands of games of Go were fed into the program, lighting up the artificial neural networks within the machine. Solely by observing the game itself, the computer learned its rules, its strategies, its tactics, and ultimately how to beat its best players.

We are approaching an age in which programs that eat up information, discern patterns within it, and then act upon the real world by using their acquired knowledge, will become commonplace.

Research in the field has gotten ever more ambitious over the past few years. Companies are recruiting young bright minds fresh out of top universities in order to add the necessary momentum they need to keep up in this arms race of AI. As a new wave of services which at their cores carry AI technologies kicks in, a competition for the greatest minds within the field has laid down its law. Corporate giants such as Facebook, Apple, Microsoft, Amazon, and even Baidu are luring in ivy league school graduates with enormous budgets.

Starting salaries of six figures are not unheard of. We are today witnessing a continuous conversion towards an era of AI which will unquestionably expand its reach beyond the tech industry.

Teach a machine a game and it’ll win for a day. Teach it how to learn…

The implications of this are surely extremely innocuous on the surface. After all there is no way an automated barista inside the local Starbucks around the corner, or a couple of surprisingly well-chosen recommendations on Amazon could do you any harm. Right? And all those dystopian sci-fi movies about malicious robots rising up to enslave humanity? Pure nonsense!

I propose yet a different angle on this technological feat of our times. There are as with any other technology, major downfalls with the deployment of it into our society.

Yea sure, the self-checkout register at your local supermarket isn’t going to take over the world anytime soon. And we haven’t yet heard of any robots gone rogue, taking up arms against their human masters. But the implications of AI may be slightly more complex than those fictional doomsday scenarios.

And although less blunt in their treachery, they still have the ability to become devastating.

AI may in the near future be very capable of steering societies into directions into which they never wanted to be led.

Take for instance a social-cultural experiment conducted last year by the tech giant that is Microsoft. On March 23, 2016, Microsoft released a bot dubbed Tay, an acronym for  “Thinking about you”, onto the social media platform Twitter, which would learn from any user who would decide to tweet to it, and then imitate them. It would learn from the collective demeanor of its users, and reflect its acquired experience, back onto the twitter platform. The learned behavior, however, didn’t just limit itself to speech patterns, or arbitrary slang. Microsoft’s house made artificial intelligence bot adapted its user’s very apparent worldviews as well. In a coordinated effort, users abused the bot’s adaptive behavior and taught Tay how to spew, probably the most hateful speech ever to be shown publicly by a bot of this kind yet. With eager, Tay reflected back into the world the obscene bigotry it was so keenly directed towards, in the process of which Tay managed to dispute the existence of the Holocaust, referred to women and minorities with unpublishable names and advocated genocide.

Now the insidious nature of this incident may appear foggy at first. After all, a twitter-bot can do no real harm. Words are merely words and mean even less when they are written by an algorithm. But the wicked reality of this event starts to present itself once we look deeper into the underlying cause of it.

On the surface, it may seem like these algorithms simply allow machines to learn the same way a human infant might go through the process of learning, only on a much larger scale.

An extension of ourselves

A lot of our assumptions about Machine Learning rest on the idea that these programs are just vacuuming up knowledge like a sociopathic prodigy in a library, and that these algorithms will then be able to make active, and rational choices based on the information they have discerned out of it.

This just simply isn’t how they work. All these algorithms are doing is shuffling information around in search of commonalities – basic patterns at first, and then more complex ones – and for the moment, at least, the greatest danger is that the information we’re feeding them contains each and every one of our biases.

Now, if we extend this idea for instance to marketing, advertisement, healthcare, media, the insurance market,  and god forbid, law and government, the wicked implications of these technologies start to become a lot more evident.

In the beginning, man’s primary mode of transportation consisted of his two own legs. The amount of energy it took to get from one place to another was directly proportional to the distance one wanted to traverse.

Then in 1817 a German in Mannheim by the name of Baron Karl von Drais, presented to the world the first version of a bike. The energy a man had to expend in order to get from one place to another was suddenly drastically diminished. The bike became an extension of our legs, and the effort of moving became much less of an effort.

A computer can be seen as kind of the same thing, but for the human mind. Annoying repetitive thinking tasks can simply be communicated to a machine, which then executes these for one with ease, precision, and in most cases reliability. Computers are consequently an extension of our minds and thus us. What we are witnessing isn’t merely the rise of Machine Learning and AI. We are witnessing an escalation of this extension on a massive scale. Computers are learning how to learn, and they are doing this without any regard for anything but the patterns within the information we provide them with. Information which contains human biases, predispositions, and even racist notions which aren’t immediately visible on a surface level.

We find ourselves at a crucial point in time, which is going to determine the direction this technology will take for the rest of its history. It is up to us to put in place safeguards which will prevent this technology from sliding from the slippery grip of our control and overburden the lives of myriads of people around the world in the future. By no means is this a forth bringing of an argument against this vital leap of technology. Rarely has the halting of technological advancement proven to be a benefit to mankind throughout its history. But there is a certain awareness with which we ought to approach the technological wonders of this epoch. It can elevate mankind to its greatest potential, or prove itself to be out greatest fallacy. Like the discovery of man-made nuclear fusion or the invention of flight, it has the ability to provide us with unimaginable power.  It can do so for better or for worse.

Tam Anh Bui – Recast.AI

 

Want to build your own conversational bot? Get started with Recast.AI !

Share on Facebook1Share on Google+1Tweet about this on TwitterShare on LinkedIn8