Columbia Engineering and Amazon have announced four new faculty research awards for the Columbia Center of AI Technology (CAIT). The latest round of projects will explore a number of challenges in artificial intelligence (AI), specific algorithmic justice, collected method of interpreting artistic images found on the Internet, the development of a scalable, differential-private data market system and CAIT’s first award of human computers interaction (HCI).
Launched CAIT in September 2020 and is a strategic collaboration between Columbia and Amazon to push to the boundaries of AI. In addition to supporting research, CAIT also provides funding for the PhD. -Scholarships, a seminar series and an annual research symposium.
The four faculty research projects supported are:
Algorithmic justice through causal lensElias Bareinboim, Associate Professor of Computer Science
“Despite the growing concern about issues of transparency and justice and the high complexity associated with this task, there is still not much understanding of basic characteristics of such such [AI] Systems, “Bareinboim wrote in his abstract.” To help AI designers develop system that is ethical and fair, we will build on the recent progress in causal contexts to develop a principled and general causal framework for capturing and detaching the different causal mechanics here. In addition to providing a causal formalization of justice analysis, we will investigate assumption, degradability and power in the proposed fine -taking causes of justice. This allows us to quantitatively explain the overall observed disappearance of decisions through various causal mechanisms that are often found in the real world decision making.
Facially expressive roboticsHOD LIPSON, James and Sally Scapa professor in Innovation at the Department of Mechanic Engineering
“The non-verbal part of people and machine interaction does not keep up” with progress in language models, “explained Lipsson’s abstract.” This growing gap between the progressive verbal content and the poor non -verbal ability will prevent AI from reaching its potential in full human commitment. The goal of this research pilot is to explore architectures that allow robots to start the subtle but critical art of physical facial expressions. Lipsons Laboratory has developed a soft animatronic facial platform 26 soft news, most of them sparkly critical expressions such as lips and eyes. Context, and the other is to learn how to physically formulate these expressions on given soft face. “
Neural methods for description and interpretation of works of artKathleen McKeown, Amazon Scholar and Henry and Gertrude Rothschild Professor of Computer Science
“Art’s ubiquitous on the Internet requires better ways of organizing and giving sense of visual art,” McKeown said in his abstract. “We propose a study of the unified method of representing and describing these artistic images. We propose to begin with a study of representations produced by large preparations and language models to understand the kind of aesthetic information they encod.” This information included color, form, style, emotion and subject. McKeown suggested that follow -up study of greater difficult, generating descriptive and interpreting captions. “We believe that this line of study has the potential to drive social good and commercial value and expand access to the visually strange, while allowing better tools to a rage of commercial scenarios,” McKeown wrote.
DataEx: Data Market System for Modern Data UgersEugene Wu, Associate Professor of Computer Science
“We offer to develop a scalable and differentially private data market system and implement a version for Columbia Campus,” Wu wrote in his abstract. “The data market system gives anyone who has a machine learning task to upload their training data set in a differentially private [manner] And search for other data sets that could be used to increase their training data to produce a high accuracy model. At the same time, data providers can upload differentially private summaries of their data sets to be indexed by the platform. “Differential privacy is the gold standard in data protection, guarantees anonymity for individuals. Has available and benefits from the white university’s collective capacity,” Wu wrote.
WUS project is helped by Michael Kearns, Amazon Scholar and professor at the Department of Computer and Information Science at the University of Pennsylvania. The collaboration between Wu and Kearns demonstrates cross -pollination among Amazon’s academic societies and future researchers, a goal of CAIT since its exception.