June 29, 2017 -
Carnegie Mellon University’s new CMU AI initiative will bring together its various areas of artificial intelligence research under one large group and potentially attract more startups and investors to the university and region, according to a report by Pittsburgh Post-Gazette.
“There are multiple parts to [AI]. One of the strengths of Carnegie Mellon is that we are experts in all of those parts,” said Andrew Moore, dean of CMU’s School of Computer Science. “As we have grown, we have focused on those parts, but we have not focused enough on how those parts fit together.”
Departmental separation was a logical move in the past, as various researchers, professors and students specialized in certain components of artificial intelligence.
Moore says the most basic level of CMU’s AI research requires a deep understanding of the world since a computer is programmed to collect input from the world through vision and speech. These responsibilities are led by the Language Technologies Institute and CMU’s Computer Vision Group.
At a more advanced level, CMU’s Machine Learning Department instructs computers how to learn based on each experience.
Another CMU group is focused on AI-aided search, a form of information retrieval that can be applied to search engines as well as robots that collect information at a disaster site.
Moore believes that it makes more sense to treat these various AI groups as an interdisciplinary department that isn’t subjected to hierarchical input so that they generate full systems instead of individual components.
The university has 120 faculty members working in various areas of AI research, along with 200 other Ph.D. students, 300-400 graduate students and at least 150 undergraduates.
This collective of AI experts has developed high-profile intellectual property that has been acquired by a handful of U.S. tech giants.
Google acquired open-source fraud prevention technology reCaptcha in 2009, which was created by CMU computer science professor, Luis Von Ahn. Then in 2016, Facebook bought CMU spinoff Faciometrics, a facial recognition feature developed by Fernando De la Torre, an associate research professor at CMU.
Moore believes that bringing all the various AI disciplines under one umbrella will help the university gain more funding for research because it will make potential partners aware that “Pittsburgh and CMU is a one-stop shop for crazy, advanced research in AI.”
Additionally, CMU spinoff startup Marinus Analytics recently announced it has developed facial recognition technology that helps law enforcement to track down missing persons by leveraging big data.
FaceSearch uses AI, machine learning, computer vision, predictive modeling, and geospatial analysis to transform big data into large blocks of information that provide actionable insight.
Marinus founder and CEO Emily Kennedy lead the development of the company’s deep web analytics tool, Traffic Jam, while she was still a CMU student.
The machine learning-based Traffic Jam tracks and analyzes patterns in publicly available data, such as cell phone numbers that are listed in online sex ads. The software then combines those phone numbers with a range of other numerical data to find individuals involved in trafficking circles.
Several law enforcement agencies – including the FBI, Utah’s Office of the Attorney General, the South Dakota Division of Criminal Investigation, the San Antonio Police Department and the Modesto California Police Department – are already using the software.
Law enforcement officials can run an image of a missing child from social media or a social worker through FaceSearch and find out whether the victim has previously appeared in online sex ads.
“Traffic Jam is the beginning of applied artificial intelligence to find victims of sex trafficking online,” Kennedy said. “We are proud to assist the important work done by investigators across the world.”