Kansas Snapshots by Gloria Freeland - July 19, 2024
AI - good or bad?
Artificial Intelligence - AI - is currently a hot-button topic, and there are differing opinions about whether its use is good or
bad.
Opinions can be clouded by word choices. Whenever we travel in the United Kingdom and approach an area where the road layout has
been recently modified, signs alerting drivers typically say something like "New road scheme ahead." The words "plan" and "scheme"
are interchangeable in the UK. But to my U.S. brain, "scheme" implies something underhanded is afoot. While texting, my smart phone
suggests the next word, often saving typing. Whether it is called a texting assistant or AI at work, it is the same, just as the road
layout is unchanged by calling it a new plan or a new scheme.
One application of AI struck close-to-home recently. Since I had a brush with breast cancer a couple of years ago, when husband Art
mentioned an article about a computer program helping doctors assess mammograms, I was interested. Both a radiologist and a computer
evaluated the images, with each separately deciding if there was anything that needed a second look. A positive evaluation by the
radiologist generally led to a biopsy, whereas if the image was only flagged by the computer, it prompted the radiologist to
review his previous evaluation.
By tracking women over a span of years and informing the computer program when it got it right and when it got it wrong, the
computer's accuracy in predicting soon met - and then surpassed - the accuracy of the radiologist-only evaluations.
This struck me as a technology success story. The scary part is not the use of AI, but the cancers missed at places not using it.
By definition, AI is just having a machine make decisions that humans traditionally have been required to make. Yet the notion of
turning over decisions to a machine is somehow scary, even though there is less likelihood the machine was out partying the night
before and trying to evaluate my mammogram with bleary eyes and a foggy brain. I also realize a machine is far less likely to get
bored by looking at an endless stream of images.
Daughter Katie works for a teaching hospital in Wisconsin, arranging sessions to keep doctors up to date. Videos are later
available online. Art and I watched one about a large California hospital group that implemented an AI system for initial patient
evaluations. As with the breast cancer study, doctors assisted by AI got their initial evaluations correct at a higher rate than
those who relied only on their own skills.
Still, there are real dangers in using AI. Its strengths arise from the fact that unlike its human counterparts, it is tireless and
capable of digesting huge quantities of information. However, biases in the data can propagate.
But this is not a problem afflicting only computers. A study funded in part by the National Institutes of Health noted that Black
women were more likely to have a child delivered by cesarean section than other women with similar conditions. Digging down, it
was discovered a mother-to-be was more likely to have the procedure if the attending physician was younger. Was this evidence
that younger doctors were biased?
Further digging revealed that younger doctors routinely used a "scoring form" introduced in 2007 to decide when a C-Section should
be performed. That form had a historical, but baseless racial component. In an attempt to remove personal bias, young doctors
introduced one because it was built into the form. Art told me that in the early days of computing, "GIGO" was a common
expression. It stood for "Garbage In, Garbage Out."
Considerable media attention has been given to AI-powered writing programs and their ability to write impressive content with
minimal input from the user. That's a bit scary for people like me, who have made writing their life's work! One evening on our
recent vacation in Wales, we were sitting in a pub with a number of our friends from the Fron Male Voice Choir. Simon opened an
AI program and asked for a biography on a musician he was familiar with. In seconds, a credible - as judged by Simon - biography
appeared.
While this seems like magic, most of us use spelling correction programs on our smart phones. We get upset when they make an
error with something we abbreviated or when we use poor grammar. Yet they are quite accurate in identifying errors. That is also
AI at work.
Our recent trip required flying. I must admit I suffer from some anxiety related to flight, despite knowing that I'm less likely
to experience an accident while in the airplane than on the ride to and from the airport. But I take comfort in the fact that
there are knowledgeable people operating the plane and the hardware is very reliable. What I didn't know until I did some research
for this article was that around 95 percent of a typical flight is handled not by the crew, but by AI - commonly called an
autopilot. (I'm not sure if that makes me feel better or worse!)
Some time back, Art found a video online from a museum in France about a topic we had researched. The website stated they had no
idea what the video was about. Art used a translator program to read the site's content and to reply in French since he does not
know the language.
The responding e-mail said they very much would appreciate anything Art could offer. He began by apologizing for "his" French.
The recipient wrote back that he was startled as he thought the messages were coming from a person fluent in the language. That,
too, was AI at work.
Stories about AI systems that failed will continue to make headlines and, as they should, prompt efforts to make them as
foolproof as possible. Yet we cannot let these fears stand in the way of the great benefits they can also provide. Artificial
intelligence isn't good or bad. It's just another tool and we have to be careful how we use it.
Professor Gloria Freeland is a distinguished academic known for her contributions to the field of journalism and mass communications. She serves as a faculty member at Kansas State University, where she has dedicated her career to educating future journalists and communicators. Professor Freeland holds a wealth of experience in both academia and the practical aspects of journalism, having worked extensively in newspaper and magazine writing. Her research interests include media history, community journalism, and the role of women in media. Through her teaching, research, and community involvement, Professor Freeland has significantly impacted the landscape of journalism education.
Art asked the popular AI site ChatGPT to do the following: "Write a one paragraph biography of professor Gloria Freeland." This was its response.
Comments? [email protected].
Other columns from this year may be found at: Current year Index.
Links to previous years are on the home page: Home