Humans have been captivated by the concept of artificial intelligence for a long time. Movies like the Terminator, The Matrix, and even Wall-E are popular because they spark an interesting debate around themes such as ethics and philosophy while providing a visual representation of what some people believe an AI-ruled society would look like.
The idea of machines that can not only complete tasks and make our lives easier, but also catapult civilization into an impossibly sophisticated future is thrilling. In fact, AI is projected to increase business labor productivity by up to 40%, which will open up more time for people to tackle more important tasks. And as companies around the world work tirelessly to make our wildest imaginings a reality, there are already several breakthroughs we can get excited about.
What is AI anyway?
Humans first started thinking about artificial intelligence in the late 1930s, during the golden age of science fiction. It wasn’t until Alan Turing wrote Computing Machinery and Intelligence that scientists started to approach the subject seriously.
The notion of a computer that can think and solve complex problems more efficiently and accurately than even the most brilliant humans is what gives AI its luster. But as far-fetched and fanciful as the movies and media make AI out to be, it’s a lot more widespread than people think. Just 33% of consumers believe they use AI-powered technology, while the actual number is 77%.
At its core, AI technology aims to increase the quality of human life. Developing such advanced technology is an uphill battle, and since Turing’s ideas sparked the revolution, the road forward has been anything but certain.
Back then, computers could only execute commands, as data storage didn’t exist yet. That means they could tell the computer to do something, but there was no way for it to remember what it did.
As storage capacity advanced over time, so did AI. Today, researchers still use the same algorithms developed 30 years ago, suggesting that the slow-moving advancements in artificial intelligence are not due to a lack of understanding. Rather, due to the lack of data storage, and thus the lack of information from which computers can learn.
AI has already passed many significant milestones. In 1997, for example, DeepBlue, an AI Chess engine, defeated then World Chess Champion Gary Kasparov over six games by 3.5 to 2.5. Today, there is not a player in the world who can beat the most advanced chess engines even a single time. Artificial intelligence has also conquered traditional card games such as solitaire and poker.
Why does it matter that AI can defeat any chess grandmaster or the World Champion Go player? They are, after all, just games.
It has to do with the learning capabilities of AI. If, in a short time of running hundreds of thousands of simulations, a machine can learn how to beat humans who have dedicated their lives to perfecting their chess game, then imagine what it can achieve in the medical, automotive, and entertainment industries – to name a few.
Chess and Go mastery is a big win for AI. However, those are two-player, zero-sum games. All the information needed to win is there, and for the computer to learn is just a matter of brute force computation. On the other hand, there is a game that presents a much more complex problem – Hanabi.
AI takes on Hanabi
Hanabi is a card game developed in France that requires a wide range of human abilities to play. Here’s a brief overview:
A group of players is dealt cards. They can all see the other player’s hands but cannot see their own. The point of the game is to use hints and strategy – working with the other players – to correctly organize the cards in their hands.
That’s the gist of it, anyway.
For artificial intelligence, a game like this poses some serious challenges. And wherever there’s a good challenge, you can be sure some savvy scientists are not far away champing at the bit to take it on. Google Brain and DeepMind are those savvy scientists in this scenario, and they have set out to develop an AI that can learn how to play and master Hanabi.
Hanabi’s main hurdle is how a player needs to interact with the game and other players to win.
In games like Chess and Solitaire, the game’s information is all available. Becoming proficient is only a matter of hours spent and moves calculated. In Hanabi, it’s another story.
There is incomplete information. Players don’t know what cards are in their hands, and there is no way to get a fully concrete answer except by deciphering the hints from other players and remembering clues throughout the game.
Those are fundamentally human qualities. To solve the problem, researchers plan to approach it using the “Theory of Mind.” In its most basic form, the theory of mind is the ability to see the world from another person’s perspective.
New algorithmic advancements will also be needed to address these new challenges, which will necessitate the integration of several AI fields, such as reinforcement learning and game theory. Emergent communication—the study of how communication develops between numerous AI agents in collaborative settings—will also be required.
So what’s next for AI?
If recent advancements in the AI industry are any indication of what the future holds, artificial intelligence is bound to find its way into even more aspects of our lives. From self-driving cars to sophisticated disease mapping, AI has a lot of room to grow, and countless more sectors around the world look set to take advantage of this exciting new tech.