Friday, February 18, 2011

Watson and the Future of the Human-Machine Interface

A reader of Andrew Sullivan writes to him:

I watched the PBS documentary on Watson last week and what struck me most was just how close we are to a Star Trek ship’s computer.

Being a fan of The Next Generation, I loved watching Beverly Crusher go back and forth with her sickbay computer to diagnose a medical issue. You could see flashes of inspiration in Beverly’s eyes as the computer made connections that she had no way of making because it was tapping into to resources and databases her human mind could never store internally (or never even know about). But the “ta-da” moment came from Beverly as she connected the dots herself, and then bounced her logic off the computer to make sure it was sound, sometimes going several rounds before working it all out. Watson seems very close to being able to fulfill that kind of promise.
The documentary touched on a possible new type of “big picture” researcher. A scientist or doctor that uses a Watson-type interface to make connections across various fields of research; connections that specialists might never make because of their narrower focus.  As long as a person is trained on how to ask the questions, the Computer can pull answers from all available sources and suggest multiple related items that might never have been considered before.

The potential is extraordinary, right?
I think this reader is right. It seems like there are three parts of the Enterprise computer that we are talking about: the human-machine interface (i.e. Dr. Crusher and the female computer voice interacting), the voice-recognition software that does what Watson did this week on Jeopardy, and the memory, which contains basically the entire history of several civilizations as well as every science and engineering fact ever known.

And so there lies the genius of Watson: the ability to take human language with its faults and peccadilloes and turn it into a searchable parameter. But can we do better? Many people, after Watson's victory, argued that computers will, as their AI goes from weak to strong, eventually surpass human cognition.
But though Watson's abilities at cognition are impressive...they were written by a human. Watson, and his supercomputer peers, does not contain a single circuit of creativity. He contains 2800 processors, terabytes of data, and requires a huge amount of electrical power. Ask Watson any information about any song ever written and the computer could probably give it to you. But ask Watson to generate a unique song that reminds him of the flight of a sparrow, and he would fail completely. He'd fail miserably.

But a human, even an average composer, armed with a Watson interface, might be able to look up chord structures, watch videos of sparrows, hear other songs inspired by birds, and create a masterpiece for an entire orchestra.

Back in February 2010, Gary Kasparov, the chess legend, wrote about the awesome power of a human armed with a computer:
In 2005, the online chess-playing site hosted what it called a “freestyle” chess tournament in which anyone could compete in teams with other players or computers. Normally, “anti-cheating” algorithms are employed by online sites to prevent, or at least discourage, players from cheating with computer assistance. (I wonder if these detection algorithms, which employ diagnostic analysis of moves and calculate probabilities, are any less “intelligent” than the playing programs they detect.) Lured by the substantial prize money, several groups of strong grandmasters working with several computers at the same time entered the competition. At first, the results seemed predictable. The teams of human plus machine dominated even the strongest computers. The chess machine Hydra, which is a chess-specific supercomputer like Deep Blue, was no match for a strong human player using a relatively weak laptop. Human strategic guidance combined with the tactical acuity of a computer was overwhelming.
The surprise came at the conclusion of the event. The winner was revealed to be not a grandmaster with a state-of-the-art PC but a pair of amateur American chess players using three computers at the same time. Their skill at manipulating and “coaching” their computers to look very deeply into positions effectively counteracted the superior chess understanding of their grandmaster opponents and the greater computational power of other participants. Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.

Which goes back to the brilliance of Watson. If we can continue to improve the way humans interact with computers, our ability to do what humans (uniquely in the known universe) do best (create) could grow as our need to do what computers do best (process data fast) becomes more integral into our daily lives.
And it isn't happening at just the supercomputer research facility. Already, my generation has learned to not bother remembering every dang fact we come across. We have wikipedia for that. We don't memorize sports statistics anymore, we have a myriad of stat tracking sites to plunder. We don't even need, really, to memorize science data anymore. I honestly don't remember the last time I did a unit conversion in my head. I have 'online unit converter pro' for that - on my phone.

So the question, to me, of Watson's victory is not whether the machines are going to supplant us, but rather how can we better integrate humans and Watsons together? We asked this question, here at TAE, several different times last year. "All you need is drivers," I rant again and again. Watson's interface requires a specific type of data (a Jeopardy "answer") relayed to it electronically. It really can't go beyond that, too far. On Star Trek (TNG), the crew can interact with the computer via voice or via consoles throughout the ship.
What if they could interact with the computer via direct neural connections? What if they could "plug in" to the computer, interact with it, and then unplug? What if the plug was a wireless connection, and they could be constantly connected? Imagine if you will, the scenario above:
Dr. Crusher seeks to solve a medical mystery. She sits down in her sickbay, silent. She appears to be concentrating very hard. Little flashes of inspiration cross her eyes. In less than a minute, it is over. She stands and prepares a new hypospray, which she administers to the patient, who immediately recovers. Ground-breaking medicine has become a silent exercise of human creativity directly coupled to raw computational power.

At the end of the day, Dr. Crusher unplugs herself from the system, and puts the external component of her human-machine interface (the power supply and antenna) in a charging socket. Though it can provide her with days of power, she likes the act of plugging it in every night. She likes to disconnect herself (she feels it helps her retain her 'humanity'), and she likes the antiquated notion that daily recharges are still necessary.
Is it really so far-fetched? Why are we spending countless hours developing better and better human-machine interface software, like Watson's interface, when we could be potentially circumventing the need for it by developing better human-machine interface hardware?

The future, I believe, is one in which we can connect our minds directly to the machines. All you need is drivers.


No comments: