Horawalavithana Selected as Editor of Two Journals
He joins IEEE Transactions on AI and Humanities and Social Sciences Communications
Pacific Northwest National Laboratory (PNNL) data scientist Sameera Horawalavithana was selected as an associate editor for the Institute of Electrical and Electronics Engineers (IEEE) journal IEEE Transactions on Artificial Intelligence and as an editorial board member for Humanities & Social Sciences Communications.
As the adoption of artificial intelligence (AI) becomes widespread, its impact on society will grow. Thus, Horawalavithana’s expertise in AI, particularly relating to natural language processing and developing foundation models, makes him a natural fit for the editorial teams of both journals.
AI relies on training datasets to work properly: for example, if you want to create an AI model that can identify cats in pictures, you’d need to show it pictures of cats and “not cats” so it can learn to differentiate between the two. Curating such a dataset for every AI model can be an arduous task. Foundation models are AI models trained on broad, general data so they can be adapted for a wide variety of uses. One of Horawalavithana’s notable accomplishments was developing the capability to pre-train large-scale foundation models from scratch—a first for the Department of Energy National Laboratories.
“We are creating foundation models that can be applied to complex problems in basic research,” said Horawalavithana. “Think ChatGPT for science.”
AI research at PNNL, such as Horawalavithana’s, aims to transform data-driven machine learning into knowledge based reasoning systems. These capabilities will allow for new scientific insights by expanding the ability to learn more effectively with much less data.
At PNNL, Horawalavithana is involved in multiple projects related to artificial intelligence, including Mega AI—a PNNL initiative to develop foundation models for science and security applications, EXPERT—an AI-driven analytics tool for nuclear nonproliferation monitoring, STEEL THREAD—building trustworthy language and graph foundation models for nuclear nonproliferation and HIATUS—developing evaluation frameworks for attributing authorship and protecting author privacy. He is also the co-principal investigator for EXPERT 2.0, which develops a novel human-AI reasoning engine with neural language models and structured domain knowledge representations to enable probabilistic multi-hop reasoning about global proliferation signals.
Horawalavithana received his PhD in computer science and engineering from the University of South Florida and his BS in computer science from the University of Colombo in Sri Lanka. He authored more than 20 publications, while his most recent papers were published at premier AI conferences, including the Conference on Neural Information Processing Systems and the Annual Meeting of the Association for Computational Linguistics.
MegaAI is a Laboratory Directed Research and Development initiative. EXPERT, EXPERT 2.0, and STEEL THREAD are supported by the National Nuclear Security Administration (NNSA) Office of Nonproliferation Research & Engineering. HIATUS is funded by the Office of the Director of National Intelligence, Intelligence Advanced Research Projects Activity.
Published: April 3, 2023