epistemic perspectives

Informing engineering from the cognitive sciences

Larry Muhlstein

I am currently in charge of machine learning efforts at Lunchclub including the core matching system. Prior to this, I have been a machine learning engineer with Niantic, the head of machine learning at Jumpstart, a PhD research intern at Snapchat, and a PhD student at the University of California, San Diego with professors Virginia De Sa and Julian McAuley in the cognitive science and computer science departments. In industry, I have specialized in developing bespoke recommendation, ranking, and matching engines that directly reflect the details of the underlying business problems. My PhD research was focused on the development of new classes of deep neural network algorithms using insights derived from the gaps between the current abilities of AI systems and the cognitive abilities of humans. From these gaps, I investigate the abstract information-level properties of human cognition that are responsible for our novel cognitive abilities and reify them into efficient, scalable, and powerful deep learning algorithms. For my thesis, I am developing a particular class of deep neural networks and applying them to problems in transfer learning, missing data imputation, and recommender systems.

A clean and consistent understanding of the complex phenomena related to human-ness is essential for developing science and technology related to humans. In pursuit of this, I am curating a dual-pronged research program to support the simultaneous development of mathematical and philosophical theories of cognition and related phenomena as well as artificially intelligent systems that make use of the insights gleaned from these theories to succeed at difficult engineering problems.

Stack Exchangegithub