In January, news emerged of a major donation by Peter Sarlin, supporting the establishment of 13 new professorships at Finnish universities. The donation will also support ELLIS Institute Finland, a hub of AI research involving the University of Helsinki. The University will use Sarlin’s donation to create a new PS Fellow professorship in AI.
Basic research too benefits from donations. Thanks to a recent donation from Nokia, Professor of Data Science Antti Honkela’s research group will develop the generation of anonymous synthetic data with large language models. Synthetic AI-generated data can replace real personal data in certain applications, as long as anonymity is ensured, that is, the real data remain undisclosed.
“For example, the language models underpinning AI chat bots are essentially text generators that create plausible continuations of the input text. Consequently, these can be used easily to generate synthetic texts,” Honkela explains.
Honkela’s research seeks to improve the quality of anonymous synthetic data. Potential applications are available in areas including healthcare. Material used to train language models could include discussions between mental health patients and professionals.
“Such models could generate anonymous synthetic data that are demonstrably unlinkable to any individual. The data generated could then be used, for example, in software development, teaching or other purposes requiring genuine-looking data unrelated to individuals,” says Honkela.
Increasingly accurate machine vision methods
Konecranes, a manufacturer of lifting equipment, is the University of Helsinki’s long-term donation partner. A recent donation from the company is promoting research on machine vision and learning. The University has allocated the funds to Professor of Computer Science Laura Ruotsalainen, whose research group is developing AI-powered approaches to promote sustainable development.
The group will use the donation to develop the latest machine learning models, namely diffusion models, for machine vision. These will provide greater insight into the movement of both machines and the objects surrounding them.
“Our research focuses particularly on the scalability and sustainability of machine vision techniques. We also explore how the methods can be adapted, generalised and explained,” states Ruotsalainen.
“We’re grateful for Konecranes’s continuous support, which enables ground-breaking scientific research in machine vision.”