In January 2015, 1,300 artificial intelligence (AI) scientists (and two social workers ) met in Austin, Texas for the 29th AAAI Conference on Artificial Intelligence to present the latest cutting edge AI research. A press release highlighted four papers that demonstrated some of the most interesting advances being made in the field. For many persons, including myself, the concept of “artificial intelligence” is limited to what is portrayed in movies or science fiction novels. Even during the press conference (to my enjoyment) there were many references to “Skynet”, as in “We need the world to know that we are not engineering Skynet.” As an outsider featured at this conference, I learned that this is in fact true (no Skynet) and the field of artificial intelligence is quite diverse.
I learned that one of the goals in AI field is to develop ‘Smart Homes’ which includes robots in the kitchen helping with the cooking. Sounds great. Think Rosie from the “The Jetsons.” Unfortunately, I also learned that a Rosie-equipped kitchen is much farther off than you would hope.
A conference press release highlighting some of the most cutting edge studies included atthe conference featured the research of Yezhou Yang and colleagues (2015) from the University of Maryland. This team of scientists has developed a computer system that teaches robots how to grasp kitchen objects by having them watche YouTube cooking videos. To conclude their presentation the team presented a video of a robot picking a spoon up off a kitchen counter and a bottle of salad dressing. After the robot squeezed some dressing into the bowl with one arm, it proceeded to stir with the second. End video. Cue applause. This was a major accomplishment in the field. As an outsider in the room I felt misled and underwhelmed. However, as the team continued to boast about their accomplishments they brought up the fundamental challenge of the science of artificial intelligence: robots can never be quite as excellent as humans.
Over the previous year I cannot stop thinking about this concept. Humans are extraordinary beings. As I continue to delve into the world of AI, I find more examples that confirm that where humans excel compared to robots is in making decisions in the face of uncertainty. The primary challenge of an AI scientist is to develop a program that can imitate this key ability of humans.
Compare yourself to the University of Maryland Robot. Imagine you are preparing a meal in an unfamiliar kitchen- perhaps at a friends house- and there is bowl and mixing spoon that you have never seen before. In a matter of seconds you would figure out how to pick up that spoon, how to hold that spoon and how to use it to mix salad dressing in the bowl. For a robot it is not as easy. Every item in the kitchen needs to be recognized so the robot can determine the appropriate action for each unfamiliar object. Although some incredibly smart people are making some outstanding progress toward kitchen robots, you can see how distant the dreams of Rosie actually are.
The Human is the most complex and effective invention of all time. Until the end of days, the Sciences will dedicate their efforts in attempt to replicate it. Unfortunately as a society, we have to come to a place where we tend to trust science over human capacity. Instead of fearing the failure of science (i.e. the failure of evidence-based interventions), we need to embrace it. Failure is part of the process and we can trust our unique human capacity to make a decision in the face of uncertainty. Failure is okay because us humans, more so than anything else, will be able to handle it.
**Note: This content was created when colloborating on a manuscript entitled “Beta Testing in Social Work” authored by Dr. Dorian Truabe, Dr. Marilynn Flynn, Robin Petering, MSW, from Univresity of Southern California School of Social Work and Stephanie Begun, MSW from Denver University School of Social Work.