Shane speaks at Dillon Lecture Series: AI isn’t as smart as we think

By Lizzie Kipp
Staff Writer

No matter their opinion on it, people use artificial intelligence in their everyday lives without realizing it. It’s in little things, like the autocorrect and voicemail transcripts on smartphones. As it advances, people can use AI to generate images, develop ideas, and even do their homework.

Janelle Shane is an optics research scientist and artificial intelligence humorist. She is a leading voice in the AI community, and author of the book “You Look Like a Thing and I Love You”.

Shane visited the Sports Arena on Tuesday where she spoke at the first Dillon Lecture Series of 2024. She spoke about her work in artificial intelligence in a way that even the most technology-challenged person could understand.

Shane said although artificial intelligence can be useful for certain purposes, it is not a perfect system.

“It did identify my dad as a squirrel,” she said, sending laughter through the audience.

She said AI algorithms are trained on internet language and pictures. This is how it learns to generate images and words – it analyzes what is already out there. However, AI is a trial-and-error process and can lead to strange results if not given enough information in the first place.

These strange results are how Shane rose to fame. Her blog, titled “AI Weirdness”, is packed with AI-generated content, from hilariously mislabeled mammals to goofy messages on conversation hearts. All of these came from machine learning algorithms.

“There is no chance of it being able to understand at a real level what it is we’re asking for, and how the world works,” Shane said.

Shane’s interest in artificial intelligence came to her as an undergraduate, from one of her professors at Michigan State University.

“I was excited about a type of computing where there are many different outcomes and they are all unknown,” she said.

Shane talked about how AI makes life weirder and more interesting, but also admits that it is not as smart as people think. It gets away with seeming more competent than it truly is. It can generate books, recipes, and even court case arguments that seem legitimate, but when one takes a closer look at its work it is quickly taken out of commission. An example she gave was a recipe for tea that sounded delectable, but “deadly nightshade” was listed as the main ingredient. Things like this are what deem AI as unreliable.

Shane also talked about workarounds, or ways to help people overcome the challenges that come with using AI. She advised to be cautious when using artificial intelligence, as it is not always reliable and can have consequences when used incorrectly.

More information about Janelle Shane can be found on her blog, AI Weirdness, or her website, janelleshane.com.

Visits: 50