According to Jimoh Ovbiagele, ROSS Intelligence's CTO, he got his idea of developing an artificial intelligence system for legal aid when he saw how expensive the legal fees are when his parents were planning a divorce. Due to the high costs involved they had to abandon the whole plan. Ovbiagele told in an interview for WIRED that, remembering his parents' difficulties, he decided to study computer science with the aim of developing an AI system that would ease the lawyers work on researching cases. He had the opportunity to pursue his AI project at the University of Toronto.
Later on, Ovbiagele co-founded ROSS Intelligence together with a team of people whose backgrounds range from computers to neuroscience to law. They came up with an application for voice recognition powered by the learning service provided by the cognitive system of IBM Watson. Their application is specialized in providing legal assistance.
ROSS Intelligence's AI app is yet another example of how machine learning is becoming a common technology infiltrating our everyday lives. Today, not only AI algorithms have improved, but also the ability to provide access to these type of applications across the Internet. For instance, this week, a toy startup called Elemental Path announced that it's taking preorders for their CogniToy dinosaur bot. The application is making use of IBM Watson's processing power as well.
SRI International, a Silicon Valley incubator that developed Apple's Siri digital assistant, announced recently a voice-recognition application that can function as an add-on for mobile banking apps. Customers can use it in order to question the system about their accounts. In this perspective, Ross is just another company following the trend.
Ross functions the same way as Siri, users being able to asking questions in natural language. The system searches through its database of legal documents and provides an answer paired with a confidence rating. A user can see the answer of the source documents from which Ross has extracted the information below. In case that the response is accurate, the user can rate it with a "thumbs up" button that also saves the source. In case that the user selects "thumbs down," Ross app provides another response.
According to Ovbiagele, Ross AI legal app is a great improvement over current research databases relying mostly on a keyword search. The system can learn from the feedback given by users and improves its performance in time, with more input.
ROSS Intelligence started the AI legal aid app project last September. According to another ROSS co-founder, Andrew Arruda, the team started their research on the project with a "blank slate" version of IBM Watson. They input in the system thousands of pages of legal documents and then trained the AI application on the ontologies and taxonomies of law. The last step was to build a proprietary machine-learning layer called LegalRank in order to further refine the system.
For the moment, Ross is specialized on insolvency and bankruptcy law, but Arruda and Ovbiagele trust its ability to move into other areas. According to them, there's no shortage of demand from an attorney seeking ways to make legal research easier. According to statistics, lawyers devote to legal research nearly a fifth of their working hours. Law firms have to spend annually around $9.6 billion on research only.
