Artificial intelligence is the next step towards creating a global digital village. Scientists have been researching keenly to develop computer systems and algorithms that can perform not only at par with human intelligence, but outperform them. In the past one year, certain algorithms have even surpassed human performance. Here are a few such examples:
Face Recognition: In April 2014, Chaochao Lu, Xiaoou Tang of the Chinese University of Hong Kong published a paper – ‘Surpassing Human-Level Face Verification Performance on LFW with GaussianFace’ claiming that the algorithm that they developed can recognize images better than humans. This program uses a multitask learning approach based on Discriminative Gaussian Process Latent Variable (DGPLV) Model. DGPLV uses data from multiple source-domains to improve generalization performance of face verification.
Average human performance for face verification on LFW (Labelled Faces in the Wild Home) database is 97.53%. However, these researchers have claimed that their algorithm’s performance is higher by almost 1% than an average human being i.e. 98.52%. The below ROC (Receiver Operating Characteristic) curve for different face-recognition algorithms proves just that. GuassainFace algorithm (red-dotted curve) surpasses all other algorithms and humans.
The success of research in face recognition showed a significant reflection in patenting activity over the last few years. The above graph lists the top ten patent assignees in the US since 2009, with Samsung leading the list in Artificial Intelligence technology with 168 patents.
Neighbourhood watch: In September 2014, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) claimed that a new algorithm developed by them can analyze and develop inferences about a neighbourhood. The researchers trained the algorithm with eight million images and GPS-based crime data and location of McDonald’s outlets embedded in them. The algorithm further used deep learning techniques to teach itself different qualities of correlated photographs. For example, the algorithm discovered that one often finds taxis, police vans, and prisons near MacDonald’s outlets and you rarely find cliffs, suspension bridges, and sandbars near them. The algorithm therefore can answer questions such as – Is this area safe? Where might I find a parking spot? Am I more likely to get to a gas station by taking a left or a right at this stoplight?
Now this technology is extremely useful for pedestrian navigation. In January 2012, Microsoft was awarded a patent US 8,090,532 B2 that describes a system that suggests pedestrian routes considering information such as weather, crime statistics, and demography.
Computer gets judgmental on your personality: In January 2015, Wu Youyou, David Stillwell and Michal Kosinski – researchers at Stanford and Cambridge Universities – published a paper claiming that an algorithm developed by them can judge a human’s personality better than humans. They tested the algorithm on 86,220 participants and found that their algorithm was able to judge a person better than his/her co-workers, friends and family. It’s interesting that the algorithm came second to a subject’s spouse.
The figure below illustrates that by using Facebook Likes, the algorithm predicted a person’s personality more accurately than most of their friends and family. With enough “Likes” to analyse, only a person’s spouse (0.58) was able to outperform the algorithm (0.56) and that too marginally for accuracy based on five big psychological traits (Openness, Agreeableness, Extraversion, Conscientiousness, and Neuroticism). You can take the test for yourself here – http://applymagicsauce.com/
Personality assessment can be helpful for organizations to better communicate with clients on email, SMS, chat, Facebook, Twitter and other communication methods. In January 2014, Persuasive Labs, Inc. filed an application with WIPO (WO 2014113889 A1) describing a system to assist organizations to communicate with clients. The system obtains data related to an organization’s customers by crawling relevant social networking websites such as Facebook, news or Twitter feeds, and email database, etc. The system then generates personality profiles for each customer. A personality profile may comprise a list of traits that may include the Big Five, or other taxonomies that describe human personality traits such as Myers Briggs (Extraversion/Introversion, Sensing/intuition, Feeling/Perceiving, Thinking/Judging).
Artificial Intelligence soon can supplement human decisions with valuable insights. For example, IBM Watson has access to millions of medical papers and other literature and it can diagnose cancer almost with the same probability as a human doctor. Doctors powered with such insights obviously will come up with better diagnostics.
So are we looking at a future where human cognizance may not be put to use for decision making, at all? Only time can tell. We are living in an age when the path of science is taking interesting twists and turns. I hope it does not prove to be an Icarian (mis)adventure!
(Featured image source: http://maxpixel.freegreatpicture.com/Forward-Woman-Artificial-Intelligence-Robot-507811)