AI Experts On Robots and Living with Thinking Machines

On Saturday June 4, 2011 I took a Megabus from Philadelphia (one-way fare $8.00) to Hunter College in New York City to attend Man-Made Minds: Living With Thinking Machines, a World Science Festival program.

Rodney Brooks, who until recently was Panasonic Professor of Robotics at MIT and who now is Chief Technology Officer at Heartland Robotics, was the first expert to speak, and he emphasized a very practical approach that was not too concerned with any negative consequences of living with thinking machines. Brooks comes across as the typical, can do engineer who wants to revitalize American manufacturing by providing workers with robotic tools. For Brooks everything from ant behavior to robot behavior to human behavior comes down to simple rules and computation. I thought he did a nice job of explaining the concept of the “uncanny valley.” Although he did not mention Masahiro Mori who coined the term in 1970, Brooks discussed how the “uncanny valley” idea holds that when avatars look and act almost, but not perfectly, like actual humans, humans respond negatively to the robot. The " uncanny valley" is a dip in the graph of the positivity of human reaction as a function of a robot's lifelikeness. (http://en.wikipedia.org/wiki/Uncanny_valley) Brooks also showed videos of Kismet and Domo that highlighted how important eye contact is for sociable humanoid robots. Here is an article where Aaron Edsinger, Domo’s developer, talks about how eyes are critical for human to robot interaction (http://www.livescience.com/1419-robot-eyes-humans-human-eyes.html). Here is a simple overview video of Brooks on robots (http://www.youtube.com/watch?v=C9p8B7-5MTI), and Steve Talbott has a nice article on the Brooks approach to robotics here (http://www.southerncrossreview.org/43/talbott-hal.htm).

David Ferrucci described the four-year Watson/Jeopardy! Project that involved about 20 people. He explained in a clear and concise way how Watson takes a question, analyzes the question, generates hypotheses, collects and evaluates evidence, weighs and combines the final answer to reflect a level of confidence, and then provides the answer. Here is a 20 minute video that goes into more detail than Ferrucci did at Hunter College in New York yesterday (http://www.youtube.com/watch?v=3G2H3DZ8rNc) I also can recommend Stephen Baker’s book Final Jeopardy: Man vs. Machine and the Quest to Know Everything (http://www.amazon.com/Final-Jeopardy-Machine-Quest-Everything/dp/0547483163/ref=sr_1_5?s=books&ie=UTF8&qid=1307315691&sr=1-5).

Eric Horvitz of Microsoft was perhaps the most relevant to my interest in robots and the future of medicine. His demonstration of the avatar assistant on a computer screen outside his office really drove home how powerful robots can be. Because this computer program has been learning about Eric for over 10 years, she can predict that he will not be attending a meeting that is on his schedule for 4:00 PM, and she offers that time to a colleague (who she recognizes by name) who needs to talk to Eric. This New York Times article shows a picture of the computer assistant and describes how capable she is in helping Eric run his life (http://www.nytimes.com/2010/06/25/science/25voice.html?pagewanted=1) Horvitz’s homepage has a great list of papers for the general reader that I have started to sample (http://research.microsoft.com/en-us/um/people/horvitz/layreader.htm). Horvitz showed an impressive tape of a medical kiosk where an avatar speaks directly and convincingly to both the mother and the child who has diarrhea. After asking several questions, the avatar decides the child is not sick enough for immediate attention and makes an appointment for the child with his doctor later in the week. The moderator Faith Salie got a big laugh when she noted that the avatar was much more compassionate in relating to the child and his mother than the human triage nurses she has encountered in New York City Hospitals.

Hod Lipson who directs the Creative Machines Lab at Cornell University’s School of Mechanical and Aerospace Engineering and Computing and Information Science provided the most futuristic and revolutionary perspective on living with thinking machines. Lipson’s development of evolutionary robotics where machines self-replicate and self-assemble provides a new lens to look at what is consciousness, what is self-awareness and what it is to be human. What fascinated me was his simulation of evolution to allow machines to learn without any human developers or programming. The machines were able to learn because they probed their external world and only those that were successful get to breed new generations of machines. I must admit that I was blown away by this approach having the robots “discover” or “learn” basic laws of physics like Force = Mass X Acceleration. It is also hard to wrap my mind around these robots discovering basic laws of the universe that we humans cannot yet understand ourselves. The NY Times has a nice article about this amazing idea (http://www.nytimes.com/2009/04/07/science/07robot.html?scp=1&sq=%22april%202,%202009%22%20hod%20lipson&st=cse). For a video where he demonstrates some of this approach please go to http://www.youtube.com/watch?v=lMkHYE9-R0A.

The moderator Salie seemed to be shocked that all the panelists appeared to endorse the prediction that eventually robots will be built that will be hard to tell apart from human beings. Only Horvitz seemed concerned about unintended consequences which concerns me. Horvitz at least brought up two obvious candidates for unintended consequences: Criminal Artificial Intelligence and liability issues around autonomous systems that cause unexpected damage to humans or property.

The evening ended with each expert providing a take home message. Rodney Brooks, ever the practical, optimistic engineer, emphasized robots increasing our manufacturing productivity. David Ferrucci observed that computers and AI will allow us to learn and make new discoveries from the large amounts of data now swamping us humans. Eric Horvitz, the most thoughtful and philosophical of the bunch, said robots will teach us who we are as humans. Hod Lipson described AI as a new lens to look at what consciousness is, what self-awareness is, and what it means to be human.