The Intelligence Advanced Research Projects Activity (IARPA), the Office of the Director of National Intelligence’s research and development arm, announced the initiation of a programme to design breakthrough artificial intelligence (AI) technology capable of assigning authorship while respecting authors’ privacy. Human Interpretable Attribution of Text Using Underlying Structure (HIATUS) is the Intelligence Community’s most recent initiative to develop human language technology.
“Each of the selected performers brings a unique, novel, and compelling approach to the HIATUS challenge. We have a strong chance of meeting our goals, delivering much-needed capabilities to the Intelligence Community, and substantially expanding our understanding of variation in human language using the latest advances in computational linguistics and deep learning,” says Dr Tim McKinnon, programme manager.
The advances that follow may have far-reaching effects, with the ability to thwart foreign malign influence activities, identify counterintelligence threats, and protect authors who may be in danger if their writing is associated with them.
The programme’s objectives are to develop technologies that a) Assign multilingual authorship by recognising stylistic traits such as word choice, phrase phrasing, and information structure that aid in determining who wrote a particular document; b) Maintain the author’s anonymity by changing linguistic patterns that reveal the author’s identity; and c) Use explainable AI approaches to give beginner users an understanding, trust, and verification of why a specific text is attributed to a specific author or why a specific alteration will protect an author’s anonymity.
IARPA granted HIATUS research contracts to a few lead organisations through a competitive Broad Agency Announcement, and as a result, the initiative now includes more than 20 academic institutions, non-profits, and corporations.
Moreover, the National Artificial Intelligence Research Resource (NAIRR) Task Force recently met for its ninth public meeting to discuss the last details of a plan to establish a national cyberinfrastructure that would boost American competitiveness in a crucial developing technology. The NAIRR would advance fairness in the AI research environment and spur innovation, job creation, and economic growth by democratising access to the tools and resources that support AI research and development (R&D).
Early discussions during the meeting focused on two crucial components of the NAIRR’s implementation strategy such as access and security controls, and resource allocation and evaluation. The Task Force members agreed that it was crucial to develop a system for allocating resources that could consider researchers who already receive funding from the federal government as well as those who do not, as well as researchers or students who are requesting access to a variety of research projects, from modest, exploratory endeavours to more substantial, resource-intensive ones.
The Task Force also discussed topics related to the NAIRR’s establishment and sustainability, including how to ensure environmental sustainability, improve international cooperation, and develop the necessary legal authorities and frameworks to carry out the NAIRR’s purpose.
The Task Force highlighted that establishing energy efficiency guidelines for NAIRR computing resource providers might enhance sustainability, especially when combined with expenditures in AI research that could benefit environmental studies.
The Task Force invited several representatives of similar Federal efforts focused on providing computational and data resources to the research community to participate in a panel discussion to examine how the NAIRR implementation plan may be created to support and strengthen these projects as a component of a comprehensive Federal ecosystem.
The conversation shed light on how different projects complement one another, the shared goals of supporting the American research community, and the necessity of continuing collaboration as each programme advances from conception to implementation.
