Research Interests

Machine learning and artificial neural networks: fundamentally new paradigms

In a few decades, ML has been used to develop artificial neural networks (ANNs) that learn, remember, think, reason, analyze, and predict in ways that non-experts believe mimic analogous human brain processes that have evolved over eons. Neurologists, brain scientists and psychologists say that they do not mimic each other and both are inscrutable. These apparent analogies require redefining those words as well as data, knowledge, intelligence, and reasoning, and beyond to science and the scientific method.

ML-based AI captured the world’s attention due to its demonstrable, valuable capabilities to reason and solveproblems never before possible at unfathomed scope, scale, complexity, realism, power, and speed beyond human understanding; and to widespread speculation that ML will profoundly impact all aspects of human life – society, economy, politics. The greatest challenge in achieving such results is ML’s inherent inscrutability and complexity, despite millions of successful applications.

My research explores what is known and what is inscrutable. It concerns the nature, role, and impact of digital data in the emerging Artificial Intelligence (AI) Revolution to represent and discover knowledge using the ML paradigm to develop ANNs that are applied in ANN-based problem solving, also called data science. Geoffrey Hinton, who shared the 2018 Turing Award and the 2024 Nobel Prize in Physics for fundamental contributions to ML, refers to the capabilities of the brain-inspired ML paradigm as Digital Intelligence (DI) that he argues outperforms the human brain’s Biological Intelligence (BI). My research explores some remarkable and philosophical aspects underlying such claims.

DI gained the scientific world’s attention in 2018 when AlphaFold solved a 50-year biology challenge of predicting, with atomic accuracy, a protein’s 3D structure given its amino-acid sequence. It captured the world’s attention in 2022 with ChatGPT that answers questions with human-level performance. ML’s inscrutability poses the main challenges to understand, develop, refine, and apply ML to prevent deploying erroneous or harmful results in practice. The long-term benefits and risks are speculative based on limited achieved results. To address these challenges, DI became the 21^st century’s most active and rapidly developing research field with ~10⁶ publications per year.

My publications provide high level descriptions of the ML paradigm, what is known and what is inscrutable that lead to some remarkable consequences. Conventional data analysis paradigms that are used in math, logic, science, statistics, and computing are well understood with properties such as correctness and accuracy that can be validated based on well-defined theories and empiricism. In contrast, the ML paradigm can produce incorrect and inaccurate results. It lacks an underlying theory to validate results that must be used not as answers but insights to develop answers by conventional means. As conventional and ML paradigms differ fundamentally, some of their differences are described to provide insights into the ML paradigm and its DI. ML offers a profound, new, data-driven paradigm that opens the door to new realms of knowledge representation and discovery.

A fascinating question is – What is the relationship between DI’s computational processes and BI’s cerebral processes? Yann LeCun, Meta’s Chief AI Scientist and primary DI contributor said, “By amplifying human intelligence, AI may cause a new Renaissance, perhaps a new phase of the Enlightenment.”

For more information contact Michael L. Brodie (mlbrodie@seas.harvard.edu), DASLab, School of Engineering and Applied Sciences, Harvard University.

References

Brodie, M.L., Machine learning and artificial neural networks: fundamentally new paradigms, in Research Handbook on Digital Data: Interdisciplinary Perspectives, Aleksi Aaltonen, Kalle Lyytinen, and Marta Stelmaszak (Eds.) forthcoming Fall 2025.
Brodie, M.L., What non-AI experts need to know about AI-based data science: Observations based on discussions with world leaders in many disciplines at the Futurist of the Year 2024 Congress, Warsaw Poland, April 9-11, 2024, LinkedIn April 24, 2024, and Medium and Academia, April 25, 2024 Published in Polish.
Brodie, M.L., A framework for understanding data science, arXiv preprint https://arxiv.org/abs/2403.00776Harvard University, March 2024.
Brodie, M.L., Defining data science: a new field of inquiry, arXiv preprint https://arxiv.org/abs/2306.16177Harvard University, July 2023.
M.L. Brodie, Understanding Data Science: An Emerging Discipline for Data-Intensive Discovery, in Shannon Cutt (ed.), Getting Data Right: Tackling The Challenges of Big Data Volume and Variety, O’Reilly Media, Sebastopol, CA, USA, June 2015.
M.L. Brodie, Doubt and Verify: Data Science Power Tools, KDnuggets, July 2015. Republished on ODBMS.org.
J. Duggan and M. L. Brodie, “Hephaestus: Data Reuse for Accelerating Scientific Discovery,” CIDR 2015, Jan. 2015.
Brodie, M.L., A data science axiology: the nature, value, and risks of data science, arXiv preprint http://arxiv.org/abs/2307.10460 Harvard University, July 2023.
M.L. Brodie, Why understanding truth is important in Data Science? KDNuggets, January 1, 2018. Republished Experfy February 16, 2018.
M.L. Brodie Data The World’s Most Valuable Resource, 2017 Onassis Lectures in Computer Science on Big Data and Applications, Heraklion, Crete, Greece, July 10, 2017.
M.L. Brodie,Developing Data Science. (preprint) In Braschler et al.
M.L. Brodie and J. Duggan, Big Data is Opening the Door to revolutions: Databases Should be Next, ACM SIGMOD Blog, November 8, 2014.
M.L. Brodie, “The First Law of Data Science: Do Umbrellas Cause Rain?,” KDnuggets, Jun. 2014.
Piatetsky-Shapiro, “Interview: Michael Brodie, leading database researcher, industry leader, thinker,” SIGKDD, vol. 16, no. 1, pp. 57–63, Jun. 2014.
M. L. Brodie, “Piketty Revisited: Improving Economics through Data Science – How Data Curation Can Enable More Faithful Data Science (In Much Less Time),” KDnuggets, Oct. 2014.
M. L. Brodie,What is Data Science?(preprint) In Braschler, M., Stadelmann, T., Stockinger, K. (Eds.), “Applied Data Science – Lessons Learned for the Data-Driven Business”, Berlin, Heidelberg: Springer, 2019, ISBN 978-3-030-11821-1, DOI 10.1007/978-3-030-11821-1