Observations about AI and older adults

It’s a scary time to think about AI. Healthcare workers are nervous, professional caregivers think it’s too early. To read the media, reporters are sharing their anxiety. AI is terrifying about the possibility of making people (and journalists) obsolete or initiating accidental destruction and havoc.  Doctors worry about the elimination of whole specialties like radiology and educators are in a tizzy trying to determine the real author of student projects. The media frets frequently about mistakes made by ChatGPT.   But the situation is vastly different in the older adult marketplace. It is a space that benefits greatly from AI’s capability to learn from accumulated data, combined with its ability to predict possibilities like Alzheimer's disease based on that data, informing caregivers, families, and other stakeholders.  Other countries with serious caregiving labor shortages see the possibilities – and if anything the US senior care industry is late to the party. 

AI-powered digital companions and training tools are evolving.  AI can help novice users with a tech problem or search answers to perplexing questions – to avoid asking their family members or becoming buried in frustrating searches for answers. AI will likely be part of the next generation of user interfaces for all types of interactions – including health questions. Older adults may be less embarrassed talking to a chatbot about tech uncertainty than calling up their long-distance grandson.

AI chatbots will be built into senior living concierge services.  Training an AI chatbot on the various services and schedules offered by a senior living community will be required.  Partially to support external queries about the community, but also to serve those within, serving up answers about schedule and location, when is breakfast (not for 3 hours), and when is my son coming to pick me up? Chatbots will be trained to respond to an increasingly complex combination of questions that depend on accumulated knowledge about the individual, including family members, medication regimens, and preferred activities.

Errors surfacing in the midst of today’s conversational AI hype-e-thon are real.  Answers to questions online are as good as the source data – which is often flawed or lost in the descending order of retrieved web links.  Chatting with an AI like BingChat invites comparisons to Google search. Is it better/worse/the same?  For certain facts it’s weak, and just as dependent on the inaccurate and incomplete data on the Internet that is sorted in Google-ized order.  My favorite tests are about population counts by age – both should be up to date on how many people in the US are 65+.  But no. But both Google (starting with A Place For Mom’s low-ball of 54.1 million presumably updated in 2022, and Bing Chat’s 54.7 million as of 2019, or 52 million as of 2018 (after asking the same question again!) don’t match the US Census (56 million) which should be the final arbiter for both.  However, the fault is less with the tools than with the mediocre data available, infrequently updated even with basic statistics. 

Remember when voice first technology became ‘voice’. You probably didn’t notice the consumerization and commoditization of this variant of AI.   But there is a microphone on every app today, enabling us to speak versus typing. Behind that microphone is a variant of conversational AI (“I think you said "What’s the Weather in Halifax”).  And as you type in a Mirosoft Word document, AI is used to finish your partial phrases and quite predictable thoughts as a result of years of ‘machine learning.’ We take all this AI for granted -- it is just there.  ChatGPT and BingChat, etc., seem temporarily remarkable and even intimidating in some industries, but this ‘generative AI’ will be built in to the next generation of all tech, enterprise and consumer. We will expect it, appreciate it, and forget about it.​​​​​​