The impact of AI: Takeaways from MS&E Reunion 2022
MS&E welcomed former classmates back to campus on October 22, 2022 for our first in-person Reunion event in three years.
Around 150 alums of MS&E and its legacy departments gathered in person during Reunion Homecoming weekend to network, reminisce, and enjoy a sense of togetherness. More joined the event virtually from as far away as Argentina, The Philippines, and Thailand.
The event began with a mini-symposium, where faculty speakers discussed the impact artificial intelligence is having on teaching and research in the department. Fortinet Founders Chair of MS&E Pamela Hinds began the symposium by sharing a few things that inspire her about being part of the department right now.
Top of mind for our department Chair:
- Five new faculty members are bringing new energy and ideas to the department;
- Research interests in the department have shifted even more toward creating social impact over the last decade;
- The recently-launched Undergraduate Diversity in Research program is bringing new voices and inspiration to MS&E research;
- There is a marked increase in interest in MS&E courses from students across Stanford at all degree levels.
The impact of AI on MS&E: Symposium talks
The symposium offered a chance for alums to be "students without the stress," according to event emcee Ravi Belani, who is an Adjunct Lecturer of MS&E. Attendees learned from faculty lectures, and afterward had their questions answered by a panel session with all three speakers.
Ben Van Roy: Enormous minds
Prof. Van Roy discussed recent breakthroughs in artificial intelligence, how AI is giving rise to technologies that might shape humanity's future, and how scholars with MS&E training are positioned to positively impact this evolution.
He described recent successes in AI as expansive—the result of using more data, more computation, and larger models. Two examples are muzero and gpt-3, which Prof. Van Roy described as "simulation optimization on steroids" and "autoregression on steroids," respectively.
The development of AI would benefit from an MS&E perspective, Prof. Van Roy argued, in that its challenges are often both computational and social in nature. For example, inclusivity, coordination of limited resources, and risk analysis are all major concerns in AI, as well as fields of expertise in MS&E.
Amin Saberi: Matching and pricing in online markets
Prof. Saberi discussed the evolution of markets into systems in which the so-called "invisible hand of the market" is often an algorithm, as well as the implications of such an arrangement.
He described how algorithms have become responsible for optimizing supply and demand in many types of markets, then offered a deep dive into ride sharing markets specifically. Prof. Saberi's recent work to improve efficiency in those markets utilized not just information about a given batch of requests, but also a prediction of likely future requests to keep the system operating as smoothly as possible.
Prof. Saberi ended his talk by contemplating various types of markets, including markets that don't involve an exchange of money, such as refugee markets and kidney exchanges. He discussed how market operators might implement algorithmic strategies that optimize not just for profit or efficiency, but also equity, social value, and more.
Kathleen Eisenhardt: Being small, being different
Prof. Eisenhardt discussed the value of small data sets and qualitative insights, and pondered whether machine learning (ML) is a re-creation of how humans become experts.
She described her experience combining ML with the practice of multi-case theory building while studying pricing models for apps. Her findings: 1) ML and multi-case theory building bear striking similarities as they're both ultimately pattern-recognition techniques, and 2) The two techniques complement each other well. Multi-case theory building narrows the search space for parameters and provides casual insight while ML provides large-corroboration and nuances like non-linearities.
Prof. Eisenhardt then discussed the value of simplification in machine learning. Particularly when dealing with models that are used by human decision makers or when data are incomplete and messy, reducing the input data to a smaller number of the most relevant parameters can make models more intelligible and useful. That is, 5-10 predictors often capture most of the predictive power of the data for many applications. In contrast, complicated models may provide precise prediction, but may not offer actionable insights for decision makers. She compared this simplification process to the way humans learn and become experts.
Note that recordings of the symposium were made available for a limited time only, and have now expired.
Together at last: Networking reception
After the symposium, attendees mingled, enjoyed food, and conversed at a networking reception. Discussions emerged around tables with suggested topics, including the future of work, financial technologies, and healthcare modeling. Others gathered indoors and outside to talk about anything on their minds with former classmates and faculty, as well as new friends.
View a photo album from the event
The department of Management Science and Engineering would like to thank everyone who attended MS&E Reunion 2022, both in person and virtually, for making our return to campus a success. And we hope to see all of you at future events!