The target of this research initiative is the 'control problem' in A.I. safety: how to create superintelligent systems that could be relied upon to preserve what we value. Projects within its scope include the safe application of superintelligent question-answering systems ('oracles') and methods for designing 'corrigible' superintelligent agents--systems whose behaviors could be corrected at run-time.
Aristo is a system that answers questions from standardized science exams at multiple grade levels. The process involves three basic functions: (1) an ability to acquire and process facts into a structured knowledge base; (2) an ability to 'understand' exam questions and analyze their accompanying diagrams; and (3) an ability to answer questions using entailment, statistical analysis, and inference methods.
AlphaGo is the first computer program to defeat a professional human Go player. In October 2015, it defeated Fan Hui, the reigning 3-times European champion, in an even match. In March 2016, it won 4-1 against Lee Sedol, one of the top professional players in the world. Go is a 2,500 year old game that has more possible positions than there are atoms in the universe. Consequently, it is 10^100 times more complex than chess.
DeepMind Health is a project to develop AI technologies for improving health care, beginning with the National Health Service in the UK. Its aim is to shift resources away from reaction towards better prevention in health care. It has acquired the app Hark to help clinicians with task management and is developing an app called Streams to assist in early detection of acute kidney injury. On July 5, 2016, it announced its partnership with Moorfields Eye Hospital, one of the world’s leading eye hospitals, to investigate how machine learning can help analyze digital eye scans, enabling earlier detection and intervention for patients with diabetic retinopathy and age-related macular degeneration.
DeepQA was a project by IBM to develop a software architecture that could answer questions over a wide range of subjects in human rather than computer terms. Watson, the product of DeepQA, uses machine learning and natural language processing to answer questions, quickly extract information, and analyze large amounts of unstructured data. Recent applications of Watson stretching from healthcare to customer service have been upshots of DeepQA technology. In 2011, Watson defeated Jeopardy champions Ken Jennings and Brad Rutter to win the 1 million dollar Jeopardy prize.
DQN is a computer program that performed better than a professional human player on 49 diverse Atari 2600 games. It was able to adapt its behavior without human intervention. On February 26, 2015, the research team’s paper “Human-level control through deep reinforcement learning” was published in Nature.
Euclid is a project to develop systems that solve math and geometry problems. Recently, it developed GeoS, an automated system that solves high school geometry problems by interpreting question diagrams and text in natural language. The system achieved a 49% score on official SAT questions and a 61% on practice SAT questions.
'gggg' is the name of the team that won the 2012 Merck Molecular Activity Challenge on Kaggle. The winners consisted of doctoral students and professors from the University of Washington and the University of Toronto, including computer scientist Geoffrey Hinton famous for the back-propagation algorithm used in training neural networks. The objective of the Merck challenge was to develop the best statistical techniques for predicting the biological activities of different molecules towards 15 biologically relevant targets. Using a deep learning model designed for speech recognition, gggg surmounted an industry standard benchmark by 17%, suggesting new computer-aided avenues for pharmaceutical research.
This project's aim is to develop autonomous robots that naturally interact with multiple humans of all ages in a range of social contexts. It involves developing skin-like materials for safe interaction with humans, speech recognition technology, and autonomous communicative functions that are context- and task-sensitive. The project's long-term goal is for intelligent robots to occupy supportive roles in elderly care, public facilities, public transportation, and education. In August 2015, the project unveiled its first ERato Intelligent Conversational Android (ERICA), a mostly autonomous robot.
Video Source: https://www.youtube.com/watch?v=vbStsbwH0V8
This project aims to develop machine learning models and tools that create compelling art and music, and to engender a collaborative community of artists, coders, and machine learning researchers. Its research builds on existing work in image generation using neural networks, and grapples with questions such as how to generate art that is dynamic and surprising, exhibits a long-term narrative arc of sorts, and can be fairly evaluated. In addition, the project more broadly investigates the confluence of art and technology through Artists and Machine Intelligence, a program that brings together artists and engineers to realize projects using machine intelligence and has, since February 23, 2016, blogged about art's symbiotic relationship with technology.