The End of Thought: Demis Hassabis & DeepMind

By Spencer Kellogg | USA

Lee Sedol couldn’t believe what he was witnessing. The undisputed king of the most complicated board game in the history of humankind stood abruptly from his chair and staggered off stage. He wouldn’t return for another 15 minutes and as the match clock wore on, the commentators sat awestruck. It was the 2nd game in a series of 5 pinning the world champion Sedol vs. Google’s DeepMind supercomputer “AlphaGo”.  If Sedol could claim that the first loss was due to oversight, here in the 2nd game the world witnessed the stunning power of Google’s A.I. technology. Move 37 was something totally unexpected and against all of the theory that humans have plotted in the previous centuries of playing the ancient game of Go. In one elegant move, “AlphaGo” had dismantled the greatest modern player and with it posed questions about the very nature of human intellect and creativity (AlphaGo vs. Lee Sedol). A year later, DeepMind topped their efforts with a new model “AlphaGo Zero” which started from a Tabula Rasa (an intellectual blank state) and proceeded to master the game within 40 days and triumphed over its predecessor 100 games to 0. AlphaGo Zero also learned the game of chess from a blank state and within 4 hours of being shown the board for the first time was strong enough to destroy Stockfish, the top chess supercomputer ever built (Alpha Zero vs. Stockfish). The implications of this daunting and surreal new technology are limitless but also signal an end to a particular era of thought and lineage – that of humans. By the year 2020, many of the top thinkers in the field expect A.I. to infiltrate the industries of healthcare, robotics, and retail. The creators of this technology often cite the benevolence of a supercomputer mind. With it, they argue, we could reduce poverty and hunger throughout the world and possibly extend our reach deep into the universe. Although his name is probably unfamiliar to most, Demis Hassabis, the mastermind behind this new technology, could become synonymous with that of Newton and Einstein.

At DeepMind games are the training ground for their artificial technology. Games present a perfect case study to test the intelligence of these networks and provide a feedback system for evaluating their cognitive ability to solve on their own. Hassabis and his team began with Atari games that are challenging to humans but also simple enough that the machines could make progress (A.I. Mastering Space Invaders & Breakout). Later, they conquered the holy trinity of gaming by defeating world champions of checkers, then chess and finally Go. For those unfamiliar, Go is seen in the gaming community as the most difficult game to master because the possibility of moves on the 19×19 squared board are almost infinite. Although in the past we have witnessed computers beating the world’s best at both checkers and chess (Garry Kasparov vs. IBM’s Deep Blue), Go is so challenging that Sedol had confidently promised to defeat AlphaGo 5-0. After he lost the first match, the Chinese government pulled the live video feed effectively censoring the following games because they were worried what the results might do to the will and confidence of their people. For all of the talk about these super-intelligent machines benevolence and their aspirations to make humanity a more peaceful and bountiful planet, the underlying and terrifying proposition that is at hand is that these machines may very well usher in a completely new form of society that reduces their human counterparts to intellectually and physically inferior species.

When your mind works like Demis Habbasis, there is little time for sleep. By the age of 13 he was a world renown chess prodigy and by 17 had created the popular video game “Theme Park”. He holds degrees in cognitive neuroscience and computer science from the University of London and Cambridge respectively and in 2014, he sold DeepMind to Google for a staggering price of 400 Million dollars in what remains the internet monolith’s largest European acquisition. As a child, he bought a 200$ computer from his winnings in chess and began his foray into the art of programming: “I intuitively understood that this was a magical device which you could unleash your creativity on.” And unleash his creativity he did. In the summer of 2017, DeepMind’s AI began moving past the boundaries of simple gaming and into the motor mechanics of locomotion. Animated with 2 stick legs and a basic torso, DeepMind taught itself how to walk, run, and jump over and around obstacles in virtual reality (Google’s DeepMind AI Learns To Walk/Run). Whether on a chess board or on a virtual plane, the results are always the same: AI shows a natural relentlessness to achieve the task at hand.

While many in our country and abroad are worried about the temperament and braggadocio of President Donald Trump, my concern is more suited with the habits and nature of the society that we have created and the society that will come to be. As I walk around the grocery store I can only think of how every single job in that space will be eradicated in the next 10-20 years. What will all of those men and women do for work? Furthermore, what will the consciousness of humankind resemble when we are relegated to the dustbin of intellectual dominance? These new machines do not see the world in the way we do and their ability to leapfrog hundreds of years of development is provoking and maddening. What will the rate of suicide be when our society is reduced to a subsistent collectivism rebooted by the productivity of robots and machines? When McDonald’s decided to phase out their retail employees in South East Asia this past spring, hungry investors cheered without any remorse as the stock price rocketed. Corporations will no doubt use this technology to infiltrate markets and capitalize on the reduced cost of labor. In a free market system, only the wealthiest of businesses will have access to this next level generative intelligence creating a marketplace that is even more narrow and centralized than the ones present today. And what will be the result of the biggest militaries in the world getting their hands on this powerful resource? Hassabis has shown an inclination towards good as he steers his company towards the enriching power that A.I. can bring into our world but his creation may very well become the next great Pandora’s box.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.