PRESS RELEASE
Contact:
Adriana Mezic
Marketing and Public Relations
AIMdyn Inc.
(805) 570-8354
SANTA BARBARA, CA
The quote “’It’s tough to make predictions, especially about the future‘” is attributed to a baseball-playing philosopher, Yogi Berra. And yet, we always try, with more or less success.
In the latest contribution, the AI and Data Analytics company AIMdyn, Inc. compared one week ahead predictions of COVID-19 cases on the CDC COVID-19 challenge (https://covid19forecasthub.org/) made by a variety of prominent research groups.
Different scientific teams utilized different approaches to the problem. AIMdyn’s is somewhat unique in that it recognizes (http://arxiv.org/abs/2304.13601) the limits on predictability enshrined in Berra’s comment. Namely, it incorporates the idea that “black swan” events (N. N. Taleb, The Black Swan: The Impact of the Highly Improbable. Penguin Books, 2008) are unpredictable, and thus the best one can hope for is recognize the change of environment and adapt the prediction algorithm.
Utilizing the underlying mathematical methodology, the AIMdyn team obtained results that beat the next best algorithm by more than 20%, a very substantial improvement in week-ahead prediction capability.
The methodology is based on the Koopman operator approach to artificial intelligence, that enables self-supervised creation of generative models which fall into the latest, 3rd-wave class of AI technologies.
The same methodology was previously applied to power an AI approach to network security via an algorithm that was licensed to Mixmode.ai, a leading AI network security provider protecting networks of large corporate and government customers.
“Perhaps the prediction science is not as dismal as Berra stated – as long as one recognizes its limits and bounds the uncertainty associated with them.” said Dr. Igor Mezic, the co-Founder of AIMdyn.
Acknowledgements:
This work was partially supported under DARPA contract HR001116C0116, DARPA contract HR00111890033, NIH/NIAAA grant R01AA023667, and DARPA SBIR Contract No. W31P4Q-21-C-0007. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the DARPA SBIR Program Office. The support of scientific research of the University of Rijeka, project No. uniri-prirod-18-118-1257, and the Croatian Science Foundation through grant IP-2019-04-6268.
Distribution Statement “A” (Approved for Public Release, Distribution Unlimited)