It was only a matter of time.
Using NVIDIA technology, researchers from the ArticuLab at Carnegie Mellon University have created a Socially Aware Robot Assistant (S.A.R.A.). Now, other than the fact that this robot can read emotions, which means she will likely be reasonably good at telling when you are lying, it will also mean that the camera on your laptop, phone or monitor will suddenly have a more important use than taking selfies.
The researchers designed S.A.R.A. to read the non-verbal cues a human gives off when conversing, and she can alter her behavior based on what she sees and hears. She detects social behavior and then reasons, based on that behavior and her acquired intelligence, to generate a social response that will best achieve the desired result. She looks at visual body elements (expressions, hand gestures, etc.) and variants in human speech (sentence structure, tone, etc.) as well as converting speech to text in order not only to understand what a person says but the deeper meaning behind the words.
After interpretation in real time, she is then able to respond with a similar set of tools including facial and body expression, tone, and sentence structure. She learns throughout the process, so the longer she interacts with someone, the more S.A.R.A. becomes specialized in dealing with that one individual. And since S.A.R.A. can share what she learns with other like systems connected to her, they will learn about you as well. That suggests that, over time, a deep learning system will become expert on the best ways to interact, manipulate and motivate you.
The initial implementation for S.A.R.A. is as the front end for an event app which helps conference attendees better achieve goals, but this tool has far broader potential applications. I can see many possible uses for systems like S.A.R.A.
One of the problems with focus groups is that populations are so diverse that researchers can't use focus groups to measure them accurately without incurring massive costs. But S.A.R.A., tied in through a connected device, like a PC, tablet, or smartphone, could possibly be used not only to select a viable representative sample but to interact with participants at massive scale and in real time. Depending on the limitations of the system she is installed on, she could do massive numbers of interactions concurrently without being limited to staff or facilities.
Every company loses a huge number of viable employee candidates due to vetting processes that are neither consistent nor able to scale to consider all candidates on both merit and personality. Applicants are often weeded out just because the candidate didn’t hire someone to prepare his or her resume to hit the proper tags in an automated system.
With S.A.R.A., you might be able to eliminate both resumes and initial qualification interviews. People interested in joining the company would simply have a chat with S.A.R.A., who then would ask qualifying questions and listen to responses and the behavior around them to see if any of the thousands of jobs a large company typically has is a good fit. There would be little risk of discrimination or abuse because S.A.R.A.’s approach would only vary based on the candidate’s behavior, not sex or race. You’d end up getting more qualified candidates and be able to fill jobs more quickly, with higher quality, and with a lower chance of litigation than by using humans.
Working passively in the background, S.A.R.A. might be able to detect more accurately than a human if someone were lying or otherwise behaving unusually. The applications here range from interviewing suspects in a criminal investigation, assessing whether an employee accident was in fact intentional or even determining for a casino whether someone was counting cards or otherwise using an illegal method to change the odds in their favor. Think also of a system that could scan large numbers of people for those that might be terrorists or others intending harm and flagging them early. As long as the data is sent remotely, it will learn from failures, not die from them.
If a system like S.A.R.A were built into apps and website, she could observe and tell what the users were enthusiastic about and what irritated them. This could help development and marketing teams best focus their efforts on enhancing the former and eliminating the latter, creating far more popular applications and games. In fact, S.A.R.A. could also become a far better automated opponent for male players, who today often think they are flirting with a female player only to discover it is another guy playing a female avatar.
I think S.A.R.A. is one of those systems that could change the world, making computers far better at interacting with humans using a broad set of social skills rather than just pre-programed responses. At the very least, it would provide far better speech to text because it would finally allow us to get punctuation right.
However, one application for S.A.R.A. I think would be fascinating, given the recent painful election process, would be to feed the video from the candidates in a debate though S.A.R.A. and have an indicator over the candidates' heads show whether she thought they were lying or telling the truth. Maybe this could be tied to a real-time fact checker so you could tell whether they simply were mistaken or intentionally trying to mislead. It might make for far more informed and honest politicians, which should improve every country over time and not just our own.
Something to think about over the holidays.
Photo courtesy of Shutterstock.