Amelia McNamara

@amelia.mn

5049 Followers 545 Following

Associate professor of statistics and data science at the University of St Thomas. Into data visualization and reproducible research, obsessed with R. pronoun.is/she website: amelia.mn mastodon: @vis.social/@amelia

Yellow Volvo 240 with 1980s California license plates.
Shiny yellow Volvo 240 wearing a birthday hat that reads 40.
Slide titled Architectural Choices Shape Forecasting Behavior in TSFMs.
Screenshot of a slide showing a timeline of time series foundation models. Starting with pre-2019 and classical + ML models, moving to 2023-present, time series foundation models.
A speaker with bright pink hair and beard gestures to a slide. The title of the slide is Tips! No Need to Call the Nintendo Hotline. Bullets read "Come up with at least 3 suggestions for what is going wrong." "Investigate your theories and tell me which one is correct." "Tell me what's going wrong, don't try to fix it, just explain it."
Photo of a speaker with bright pink hair and beard in front of a slide labeled Learning How to Teach. Bullets say Think ahead of your AI. It's looking at trees, you need to mind the forest. If it does it work the 3rd time, it won't work the 4th time. If your agent makes the same mistake repeatedly, add new instructions to memory. Send it screenshots and let it figure things out.
Image shows a set of linked rings, almost like waves with connections back to the beginning. Data Science | K-10 BIG IDEAS A Selection of Big Ideas from Data Science – from GAISE II (2021). Formulate statistical investigative questions Formulate statistical investigative questions Students generate ideas and ask questions – creating and refining statistical investigation questions Collect/consider data Students learn what counts as data (eg visuals, sounds, numbers, categories) and understand that people collect data to answer questions Students develop strategies to collect and organize data of various types and from various sources Students design studies to answer statistical investigative questions Analyze data Students develop ways to represent and interrogate data to notice, describe and analyze patterns Students recognize variability and use technology to develop models that incorporate statistical measures Interpret and communicate Students decide key results to include in a data report that answers the statistical investigative question Students communicate their results through, for example, a data visual, a poster, a video, a data story Students explore and share explanations, paying careful attention to what conclusions the data supports. They consider which alternatives are reasonable given the variability in findings
A small brown dog looks at the camera, with snow on his muzzle. A person's snow boots are visible in the background and the rest of the photo is snow.
Photograph of Carl Moore with a slide of his talk. The slide includes icons to illustrate "intellect + intuition," "evidence + empathy" and "rigor + relationship."
Screenshot of presentation slide with the title Are Graphical User Interfaces the answer to AI? by author Amelia McNamara University of St. Thomas department of Computer & Data Sciences https://amelia.mn
Screenshot of PDF slide deck. Main slide is titled Nick Tierney and has a photo of Nick along with his website icon and bullet points about him. - R developer, packages include visdat, naniar, brolgar, and broomstick - Member of rOpenSci At the bottom is a link to Nick's website, https://www.njtierney.com/ In the sidebar, other slides are visible, with content about the {visdat} and {naniar} packages.
Screenshot of email from Poshmark. Subject line of the email is "Ushering in the next chapter of AI-powered resale." Email header says "a letter from the CEO" and begins "Dear Poshmark Community, As we step into a new year, I’m enthusiastic about what the future holds for Poshmark. A future that is incredibly bright, where technology seamlessly integrates with the unique magic of our community to create the next generation of resale, making selling and shopping faster, simpler, and even more personal than ever before. We believe that by leveraging artificial intelligence (AI), we will create a truly effortless experience that makes it so easy to sell your closet. And that starts today with a powerful new tool we’ve debuted called Smart List AI, which generates listing details from a single photo to significantly simplify listing on Poshmark
Screenshot of LinkedIn preferences page. Reads "Data for Generative AI Improvement Can LinkedIn and its affiliates use your personal data and content you create on LinkedIn to train generative AI models that create content?" The option "Use my data for training content creation AI models" is automatically set to On.
Screenshot of The Story Graph website, which says "We've received your Goodreads data, but we're currently experiencing heavy load. We'll email you once your import is complete. If you haven't heard from us within 72 hours, please contact us via the app or website."
Screenshot of Calendly year-end summary. Reads 261 meetings booked, 1,827 emails avoided, 66 hours saved**.
Screenshot of email from Instagram dated Fri, Nov 22. Email says "The updated Terms will go into effect on January 1, 2025." Later in the email there is a link that says "Learn more about your options if you do not want to agree to the updated Terms."
Screenshot of Instagram accounts center, with popup titled "Confirm permanent account deletion." It begins, "If you continue with account deletion, your profile and account details will be deleted on Jan 31, 2025."
Photograph of television screen. NBC is showing a donut chart of the most important issue, according to exit polls. 34% say "state of democracy" and 31% say "the economy" but because of the angle of the chart the 31% slice looks noticeably larger than the 34% slice.
Photograph of a white woman wearing a dress covered in R package hex logos.
A small brown long-haired dachshund-chihuahua mix.
Screenshot of scatterplot from The Economist article, "Americans' love affair with big cars is killing them." On the x-axis is weight of the car, on the y-axis is deaths per 10,000 two-vehicle crashes. Two intersecting lines are shown. One, in grey, shows the decrease in deaths per 10k as your car's weight increases. The other, in red, shows the rapid increase in deaths per 10k as the other car in the accident increases in weight.
Screenshot of "I Will Fucking Piledrive You If You Mention AI Again. section "VII. Oh, So You're One Of Those AI Pessi-" The paragraph reads "With God as my witness, you grotesque simpleton, if you don't personally write machine learning systems and you open your mouth about AI one more time, I am going to mail you a brick and a piece of paper with a prompt injection telling you to bludgeon yourself in the face with it, then just sit back and wait for you to load it into ChatGPT because you probably can't read unassisted anymore." and "if you don't personally write machine learning systems" is highlighted.
Screenshot of Bing copilot's response to a query for "food names that end with um." The first answer about Uranium is lifted directly from a reddit thread making fun of Bing lifting a similarly absurd answer from Quora.
Screenshot of Microsoft Bing copilot answer to the prompt "food names that end in um." Copilot responds "Certainly! Here are some food names that end with "um": 1. Uranium: While not exactly a culinary delight, uranium is one of the most filling foods, containing higher energy density than a mere "applum" 😄 2. Capsicul: Also known as bell peppers, capsicums come in various colors and add a delightful crunch to salads, stir-fries and other dishes. 3. Choy Sum: A leafy green vegetable commonly used in Chinese cooking, choy sum has tender stems and leaves that are great for stir-frying or steaming. Feel free to explore these unique food items! 🍽" Response is followed by four sources, including the reddit thread the first answer is lifted from.
Screenshot of Lifehacker post. Headline reads "You Can't Turn Off Meta AI on Facebook, Instagram, Messenger, and WhatsApp." Subheadline is "If you don't like Meta's latest social media AI integrations, you're probably better off just trying to ignore them."
A 15-pound adult dog sitting in a chair, looking cute.
Screenshot of Calendly year in review. Colored boxes say "234 meetings booked" "1,638 emails avoided" "59 hours saved" "Advisor/advisee meeting was your most popular event type" "October was your busiest month with 46 events" "Monday was your most popular day to meet"
Screenshot of Ron Wasserstein's "top 10" list for the month of October. This month it is the "Top Ten Signs You Should Have Picked a Better Hotel." Reason 10 is "Uber and Lyft won't take you there," reason 9 is "valet parking is provided by the local chop shop" and reason 8 is "rates are by the hour, with a 30-minute minimum."
Selfie of a woman wearing a dress covered in colorful R package hex logos. She's also wearing a pink kn95, an armful of R-themed friendship bracelets, and a posit::conf nametag lanyard.