Once upon a time, a machine beat a man at chess in a contest that changed the way the human race understands itself.
The fateful match was between Deep Blue, the supercomputer built by IBM, and Garry Kasparov, the Azerbaijani Grandmaster. It took place not in the misty halls of legend but 25 years ago in a New York City highrise. Deep Blue's victory made international headlines.
Fourteen years later, in 2011, came a sequel of sorts, when Watson, IBM's AI-enhanced upgrade of Deep Blue, defeated Jeopardy champions Ken Jennings and Brad Rutter on network TV. Twice.
Despite copious eyewitness documentation, these stories carry the romance of folktales. They recall the contest between John Henry versus the steam drill that left the steel-driving man dead with his hammer in his hand. Even their real-life dialog rings with superhuman drama: Kasparov was so overwhelmed by Deep Blue's aggressive play that he resigned their final game in less than an hour, later saying that in the face of an opponent he couldn't understand, "I lost my fighting spirit." In his Final Jeopardy response, Jennings wrote "I for one welcome our new computer overlords." I assume he was joking.
In 1997, Deep Blue's ability to calculate 200 billion chess moves in the three minutes allotted to each turn was, to most folks, intimidating, even ominous. Today your laptop can run software that could defeat Deep Blue while you check your email. Watson's once-esoteric natural language processing, access to immense troves of data and split-second response time now come packaged in your home smart speaker.
Today your laptop can run software that could defeat Deep Blue while you check your email.
But as impressive as they are to tech wonks like me, processing power and machine learning are not the technologies that imprinted Watson and Deep Blue in the collective unconscious. Nor has Moore's Law or the mainstream adoption of cloud-based supercomputing astounded the masses. For a particular event to alter humanity's vision of its future, a different sort of technology is most effective, one that analytical types often relegate to "creatives" in marketing and advertising. It is one of the oldest and most powerful technologies our species has devised: storytelling.
Mention gigaflops and processor cores to most regular people and their eyes glaze over. Embed those same details in a game or competition and suddenly you have an epochal battle between man and machine, the stuff Marvel movies, conspiracy theories and paradigm shifts are made of. When it comes to altering popular perception, good storytelling beats good accounting every time.
A decade-plus since Watson, engineers are applying more complex AI technology to more complex—and potentially world-changing—problems. Two years ago an AI called AlphaFold, a spinoff of Google's DeepMind subsidiary, was used to map the shape of individual protein strands. The Nov. 2020 issue of Nature called the feat "a gargantuan leap in solving one of biology’s grandest challenges."
When it comes to altering popular perception, good storytelling beats good accounting every time.
Like DNA, proteins are considered a primary building block of life, carrying out myriad functions within a cell. Their function is determined by their shape, which is the result of elaborate amino-acid folding patterns. Billions of possible shapes exist; human observation has mapped around 100,000.
AlphaFold's AI network is a quantum leap forward in predicting protein shapes, giving researchers greater power to tackle diseases at a cellular level. Andrei Lupas, an evolutionary biologist at the Max Planck Institute in Germany, said AlphaFold's predictive modeling "will change medicine. It will change research. It will change bioengineering. It will change everything."
Revolutionary science! And yet few outside research and academic circles have heard of it. The story was buried on page 203 of Nature.
No, chances are if you have heard of UK-based DeepMind it's because of another article in Nature, published in 2016, that led to another round of round-the-world buzz: The company's AlphaGo AI beat Lee Sedol, a world champion of the ancient Chinese game of Go.
Go is both simpler and more complex than chess: Where chess offers an estimated 10120 possible moves—more than the number of atoms in the known universe, by the way—Go offers 10170, a number well beyond human comprehension. Not the case for AlphaGo, apparently. The AI's victory was seen as a leap for machine learning on a scale greater than Watson's five years previous.
And yet, amid all the hoopla, it was just a game. In terms of potentially minimizing human suffering on a worldwide scale, only AlphaFold has delivered.
Revolutionary science! And yet few outside research and academic circles have heard of it. The story was buried on page 203 of Nature.
So today, as we mark the 25th anniversary of Deep Blue's historic victory, I'm looking 25 years ahead, and I'm wondering where technology will take us. Not just AI; we know computers will keep doubling in power every two years, and datasets will keep expanding equivalently. But also the human ability to properly contextualize mind-boggling scientific advances in the form of compelling stories.
Science has already unleashed AI on stupefying problems like climate change, resource-distribution logistics, disease. Corporations and startups alike have found all sorts of niche applications for AI, mostly in behind-the-scenes digital infrastructure or data crunching scenarios. We have cloud-computing behemoths like AWS and Azure to thank for making the results of their expensive investments in AI and machine learning available to the general public, the kind that 10 years ago were the sole domain of major brain trusts like Google and IBM.
These are all meaningful applications of powerful technology, but they lack the narrative heft to capture our imagination. Or in other words, there's no story. Where's the Deep Blue of climate change? Can we gamify the war on cancer? We have to get creative with our technology, which is limited only by our capacity to imagine how to best deploy it.
Just last week Meta announced the impressive findings of its own super-powerful protein-folding predictor. But even with the company’s seemingly infinite resources–and the momentum of a research field that’s been advancing for a couple years now–these findings were instantly overshadowed by mania over Elon Musk and his takeover of Twitter. They haven’t been properly contextualized. The story hasn't been properly told. Setting these abstract battles in a physical arena and watching the blow-by-blow exchange in real time would draw the world's attention to some very unsexy but absolutely existential challenges. Go, chess and Jeopardy capture our imaginations because games are enticing. Why else would Discord have attracted 250 million users, mostly to discuss video games, in just four years?
Where's the Deep Blue of climate change? Can we gamify the war on cancer?
But as fun as games are, we need to focus our energies on more pressing issues like climate change and disease because they're urgent threats to human survival. This shift of priorities reminds me that tech billionaires pour untold dollars into virtual-reality toys and celebrity space tourism, and IBM and Google have spent small fortunes to play Jeopardy and chess. To paraphrase the brilliant physician and author Siddhartha Mukherjee (no relation): "I don't want to go to Mars right now. I'd rather focus on cancer."
Redirect Jeff Bezos and Elon Musk’s extraplanetary aspirations toward our ailing earth. Encourage Zuckerberg to put more resources into exploring the human genome. Add the multiplying effects of machine learning and I promise they'll get results.
We can certainly find a way to tell those stories so that they're as gripping as watching a computer play Go, and we can certainly get the public on board. Once we do, there's no need to go to Mars. Because at this point in the human experiment, we have plenty to deal with right here at home, and we've run out of time for games.
