Over the past three days, I had the opportunity to attend the Data Science Initiative Spring Research Workshop 2024, hosted by the University of Minnesota’s Data Science Initiative (DSI). The three-day event was packed with engaging discussions, thought-provoking talks, and sessions that explored the dynamic field of Generative AI. Here are some of the highlights and lowlights from the workshop.
Highlights
Xiao-Li Meng (Harvard) – Future Shock
Dr. Xiao-Li Meng delivered an enlightening talk on the Generative AI Revolution and a concept called future shock. His insights on broader societal implications of AI as well as the mathematical ones were profound–particularly regarding the right to be forgotten. His thoughts on K12 education, after the talk also addressed my question around empowering future constituents and voters to engage in data science rhetoric. Dr. Meng expressed excitement in engaging the next generation in the rhetoric and referred me to a column in the Harvard Data Science Review, called Minding the Future. Engaging in the conversation and asking youth what they need are the first steps in preparing the next generation to navigate and influence the data-driven world.
Jisu Huh (UMN) – AI & Advertising
Dr. Jisu Huh’s presentation on AI and advertising was probably the best-contextualized talk at the workshop. By the second slide, it was clear what she aimed to convey, and her clarity made the importance of advertising in AI rhetoric and policy evident. Advertising, as she pointed out, touches everything and is crucial to consider in any AI discussion. Thanks to Dr. Huh for her research, clarity, and impactful delivery.
Nancy Sims (UMN) – AI and Copyright Law
Nancy Sims brought a mix of precedence, wit, and style to her talk on AI and copyright law–Punn intended–I am not commenting on her dress, rather the implications of copyright law regarding style. Her presentation was fascinating, well-delivered, and thought-provoking, offering a high level examination of the legal copyright implications of AI technologies. Thank you, Nancy, for making a complex topic accessible and engaging both from the law and technical perspectives.
Michael Corey – Mapping Prejudice
Michael Corey’s talk on Mapping Prejudice was a powerful reminder of the human element behind data, uncovering how Mapping Prejudice uses community co-creation for reparative change. He posed fantastic questions around the implications of AI use in his work and others: What are we trying to fix? What are we trying to scale? Who benefits? And who loses out? His point that efficiency should not always be the goal, especially if it comes at the cost of human evolution or learning, resonated deeply. Thank you, Michael, for humanizing the data and prompting critical reflection. Anyone looking to support anti-racism work in MN, should look into Mapping Prejudice and their volunteering opportunities.
Steve Simon MN State Secretary – AI and Elections
The Minnesota State Secretary, Steve Simon’s, talk on AI and elections was insightful. A key takeaway was the importance of debating the future while understanding current laws and systems. His presentation, highlighted the critical functions of his office reagrding commerce and democracy in addition exemplifying real-world implications of misinformation and the need for robust protections.
Lucila Ohno-Machado (Yale) – Predictive Models in Healthcare
Dr. Lucila Ohno-Machado’s presentation on predictive models in healthcare was impressive. It was well-contextualized, impactful, and thought-provoking. Most impressively she took the average attendee from basic straight line assumptions (linear models), to curvy ones (neural network models) with clearly articulated examples. Her composure and expertise were extraordinary, making it one of the standout talks of the workshop.
Epic
While it pains me to reference them due to the clear disparity in resources comparing this private sector presentation to the academic ones, Epic’s demonstration of their capability to support and empower clinicians in their Epic AI demonstration was undeniably impressive. If only they hired remotely!
Lowlights
Graphs with No Error Bars
One glaring issue was the presentation of graphs without error bars. Anyone from any field should question such graphs, as error bars do not add unnecessary complexity but rather enhance credibility. Without them, it’s challenging to assess the validity of the data presented.
Alphabet Soup
Let’s do an experiment. If I define the following variables: x=”variables”, y=”math”, z=”anyone” p̂=”different”, n=”slide/s”and then ask you to read:
“z with y on their n should include the defined x in a legend that shows up on every n the y is on. z who is not from your field, would need to see the legend 7 p̂ times over many p̂ mediums to learn them. Defining the x once won’t actually help z, unless they are referenceable.”
Did you reference the variables to read the sentence? Could you read the sentence at all?
Let me help you out:
Anyone with math on their slides should include the defined variables in a legend that shows up on every slide the math is on. Anyone who is not from your field, would need to see the legend >=7 different times over many different mediums to learn them. Defining the variables once won’t actually help anyone, unless they are referenceable.
If your goal is to get someone to understand a math equation in plain english, and you only included defined variables once. There is a strong likelihood you failed for all folks who were not in your discipline.
Context, Context, Context
Public speakers should always:
- Explain who they are and whom they represent.
- Explain their work within their department/field.
- Contextualize their work within their institution.
- Contextualize their work within the broader field.
- Contextualize their talk within the broader implications of the workshop.
Most speakers failed to move beyond the first step, which is not difficult to achieve. Proper context helps the audience grasp the full scope and relevance of the talk. For example, had I given a talk:
My name is Lauren Nellie (1), and I work as the Computer Systems Analyst for First Children’s Finance. We are a nonprofit that grows the supply and business sustainability of excellent child care (2). In my role, I am responsible for management of our tech stack, supporting company analytics, and the policies and procedures around both (3). We do this in three different ways: building the financial sustainability of child care entrepreneurs, partnering with communities to preserve and grow their child care supply, and influencing state and federal systems to provide supports and investments needed to sustain child care businesses (4). As we expand across the US you can imagine I am deeply interested in utilizing scalable, safe technology. To contribute today, at the DSI spring research workshop, I am going to walk through possible use cases for GenAI in nonprofits, and then will shift to a facilitated exercise meant to bridge/build partnership across sectors through data science (5).
Workshop?
This event felt more like a lecture series than a workshop. I had hoped for facilitated exercises, engaging discussions, and practical tools to take away, but these were lacking.
Low Attendance & Representation Challenges
If the DSI is the front door to data science and AI in Minnesota, where was the welcome committee? While the initiative seems to be in its infancy, the low attendance and lack of representation from broader UMN Departments as well as the public and private sectors were surprising. Moreover, for an acadmic workshop on a newly emerging topic like AI, the average age in the room was still notably high. To truly serve as a “mentoring hub for the young generation of scientists”, they need to be in the room.
Conclusion
As a Minnesotan and UMN graduate, I remain cautiously optimistic about the potential of the DSI. While the workshop had its shortcomings, the promise of what the DSI might grow into is exciting. However, for now, I will likely turn to more well-developed initiatives such as the Harvard Data Science Review for the latest advancements. Nonetheless, I plan to subscribe to the DSI mailing list for local news and exposure to folks in MN doing great things, and I hope that the DSI will soon become a formidable force in the data science community.
This article contains my own opinions and doesn’t reflect the opinions of any organizations I might be affiliated with. OpenAI’s GPT-4 architecture was used in the construction of this article.
-Lauren Nellie
