Copy of

January 2025

PARAMETER

The Official Newsletter

of the Chicago Chapter,

American Statistical Association

January Luncheon with Mary Kwasny

Location:

Virtual (Zoom) | Chicago, Illinois

Date and Time:

Thursday, January 23, 2025 12:00 PM (Central Standard Time)

Abstract: Pitfalls of AI

AI’s prevalence and capabilities have increased over time. AI stocks have soared, and the companies behind them promise big things in the future. While AI can be extremely helpful, the technology does come with risks that few acknowledge and has the potential for catastrophic disasters. This talk will disclose my biases, highlight international cooperations to promote AI safety, examine issues that have arisen pertaining to “fair use” of data, bias and inequity, and accountability. I’ll cover some examples of AI gone wrong and some tips for how to identify when AI might may be going awry. Finally, we’ll have an open discussion on how and when people use AI to make their lives better, and how to cite it when appropriate.

About the Presenter: Mary Kwasny

Mary Kwasny is a professor at Northwestern University and has taught and done research in an academic medical center for almost 25 years, with over 150 co-authored publications in print. She has served on the board of the Chicago Chapter of the ASA for over 20 years, also serving on the ASA board of directors for 3 years, and the ASA Council of Chapters for 6 years. She was elected fellow of the ASA in 2015, awarded the Chicago Chapter Service Award in 2022, and the ASA Founder’s Award in 2023. She credits her daughter for introducing her to ChatGPT in 2022 and the fun it allowed creating things like “country songs about refrigerators.”

Agenda:

12:00pm -1:00pm     Presentation and Q&A

Registration Fees:

Includes access to the Zoom meeting

·    Members - $10

·    Non-Members - $15

·    Students - $5

President's Letter

Dear Members,

On Thursday December twelve we had the pleasure of listening to a luncheon talk by Dr. Stephen Stigler of the University of Chicago. Dr Stigler is a regular presenter at Chicago Chapter ASA luncheons and this was his eighth time doing so. Dr Stigler spoke to a full room of attendees on lotteries and potential winning strategies. His talk was both informative and entertaining and we look forward to having him speak to us again in the future. Photos from the talk are shown below:

I thought I'd finish up with just a few interesting facts about neural networks from Andrew Ng's deep learning course on Coursera.

What is a neural network (Wikipedia)?

A neural network is a type of machine learning model that is inspired by the way the human brain works. Interconnected neurons are used to process data and make predictions. The neural network is made up of layers of neurons, each of which applies a transformation to the data that passes through it. The output of one layer becomes the input to the next layer. The final layer produces the model's predictions.

What is "deep" learning?

A neural network with multiple hidden layers, and multiple nodes in each hidden layer, is known as a deep learning model. More specifically, any neural network with more than 3 layers, including the input and output layers, is considered deep learning. An illustration taken from Wikipedia is shown below:

Deep Learning and Scale

Unlike many traditional algorithms, deep learning scales well with data. The more data you have, the better deep learning works (improved out of sample prediction). Also, the more computational layers you have, the better deep learning works. This is why deep learning has taken off in the past few years (due to the abundance of data and vastly improved computational power). The relative scaling of deep learning can be illustrated by the following graphic:

Deep Learning and the Brain

While many claim neural networks are inspired by the human brain, the analogy is not very accurate. The human brain is much more complex than a neural network, and we still don't fully understand how it works. A quote from Andrew Ng perhaps says it best:

"But I think that today even neuroscientists have almost no idea what even a single neuron is doing. A single neuron appears to be much more complex than we are able to characterize with neuroscience, and while some of what it's doing is a little bit like e.g. logistic regression, there's still a lot about what even a single neuron does that no one human today understands."

Happy New Year to all and I look forward to seeing you in 2025!

Best regards,

Joe Retzer

President, Chicago Chapter of the ASA