SIRI and Me: Protests and Prognostications
By Tony Candela
Fans of the TV comedy “The Big Bang Theory” will remember an episode in 2012 where Raj, a perennially lonely astrophysicist, finds true love in his soul mate, Siri. When he asks her out for coffee, Siri disappoints him by locating several restaurants near him. Later, she suggests a good wine to have with the dinner he is planning. Not wishing to have an argument, he accepts her recommendation even though he would have chosen a different one. When he tells her he would prefer if she calls him “sexy”, she does.
I use Siri a great deal, inquiring about time and date, science, movies, the weather, location, specific topics, sports scores, and lots more. When I get naughty, which I do on purpose just to see how far I can push her, Siri rarely breaks form, saying, “I don’t know how to answer that.” As with most relationships, Siri and I have had our ups and downs, but by now, I am hooked. Living without her would be hard.
In 2023, the “Top Tech Tidbits” e-magazine began devoting a section in each of its weekly issues to AI (artificial intelligence). A recent issue covered how a blind California woman uses her Google Glass to Gather information about everything from facial expressions to the environment. It also covered ChatGPT Plug-Ins to “extract insights” into PDF documents and videos, making ChatGPT provide sources and citations, tackling AI cheating in the classroom, use of ChatGPT in IOS devices with Bing, and an app (ShortCircuit) for IOS devices. Articles on a Microsoft AI hack-a-thon and a USA Today article on AI for everyone rounded out the updates.
Recently, I attended a lecture on ChatGPT by a professor at Mississippi State University where I work part-time. “GPT” stands for generative pre-trained transformer. This is a natural language model developed by the company Open AI and licensed by Microsoft. You go on-line, ask it a question, and ChatGPT dives into a wide variety of data sources and assembles answers for you. Unless you ask it to do so, ChatGPT does not search the web as information on the web can be faulty. Instead, it looks at bona fide data bases like Project Gutenberg for books and stories or available scholarly sources for a variety of topics college students might be interested in for their term papers. As of now, version 3.5 is free and has less power than version 4.0 for which you must pay a small fee. The added power includes being able to feed data to ChatGPT and get coherent text returned to you or even to write code in say, Python, to perform a certain function. Version 3.5 only processes text. While it can enhance an EXCEL spreadsheet with formulas, version 4 can analyze an image and provide a verbal description, a boon to blind people.
The answers you receive from ChatGPT come to you in well-formatted language, modeled after the training the system has received from thousands of human conversations. Therefore, what you get is usually quite understandable. You can ask it follow-up questions and even instruct the system to provide information only if there are bona fide references. The better you formulate your inquiries, the better will be your outcomes, including minimizing the system’s just making things up. (When such confabulations get out of control, we call them “hallucinations”.)
ChatGPT is in continual upgrade mode as the (post 2021) currency of its information is improved and its ability to handle mathematical and other technical information gets better. Version 4.0 purports to better handle long inquiries than prior versions. For example, you will probably get a good answer if you ask it to provide you “a synopsis of “War and Peace” if and only if you can provide sources and include interpretations from the French and Russian points of view.” It might do a good job with “Summarize these columns of data and provide me suggestions on statistical procedures I can use to analyze them.” ChatGPT can even “look” at a video for you and provide a summary, a real time-saver.
It is easy to get hooked, especially on powerful tools like ChatGPT that use AI. We need to remember that we cannot let AI do everything for us and that human input, intervention, and even interference are good things. Our participation in the AI adventure will not be worth much if we get lazy. We must continue to know what we are doing, how things work, and how to know when AI needs our scrutiny, especially around personal privacy.
That being said, let’s return to Siri. As Henry Higgins discovered in “My Fair Lady”: “I’ve grown accustomed to her face.” Much to his amazement, professor Higgins, in molding Eliza Doolittle into a “lady” slowly became dependent on her for his sense of wellbeing. Perhaps this is the way of the world. The person-machine interaction is after all, a relationship. Perhaps some of the same dynamics are at play as in human interactions. This is especially possible when the machines are humanlike. Siri is good when she is good and bad when she is bad, but she is mostly good. In the same vein, except for hallucinations and outputs that the end-user fails to verify or tweak, I suspect that most people will experience ChatGPT as very, very good and therefore, very easy to get hooked on.
So it has been with the assistive technology we who have disabilities such as blindness have come to depend upon. They are a part of us. We rely upon them as if they are part of our bodies and when they fail, much as a hand-twitch when we pour a hot liquid, we are disappointed and occasionally injured. I contend that we rely upon our technologies in the most human of ways, meaning that we expect them to work the same way people expect their arms, legs, eyes and ears to work – all the time and for the most part, effectively. When they do not, we are befuddled at best and at worst, bitterly disappointed and frightened. Witness the stories of people who discover their eyesight is failing while glancing in a mirror or having a near-miss while driving a car. People with ALS (Lou Gehrig’s disease) notice slurring in their speech or twinges or one limb not keeping up with the other. When our assistive technology doesn’t work as expected, we get silence (screen reading software); loss or degradation of image (screen magnification software); function-keys that don’t work; or a device that doesn’t understand our voice-input. I can’t recall how many times I’ve messed up a Siri inquiry after a glass of wine, but I’ll take the rap for that one!
Recently, I had a mild argument with two technology trainers. I voiced my vision for the future, that the power of AI is incorporated so much into our assistive technology that it provides us the same ease of use of our tools that people who use the built-in power of a wide variety of devices get without extra assistance. This would mean that we won’t have to know twice as much as they do in order to accomplish the same goals. The thought crossed my mind after I had completed my expostulation that it may have meant to my interlocutors that they might not be as needed any more.
If my vision came to fruition, it would mean that when we encounter a strange new piece of software like my nemesis, Google Docs, there would be context-sensitive help generated by the software itself and also my assistive technology. It would mean that I won’t have to know as much about how to manipulate either my assistive technology or the applications software. The “fusion of technology and compassion” would automate visual image descriptions and provide easily-gotten help when I need it. It would mean that I’d be able in my own fashion to point-and-click my way through the undiscovered country and emerge triumphal with no more than the illusion that I actually possess expertise. It would mean that we will finally be free at last, free at last to function easily and in real time as was meant for all of us. May the gods of AI smile upon us. May their hallucinations stay under control while our illusions blossom into everyday reality.
Anthony R. Candela, Author
Saying aloud what should not remain silent.
Books by Tony…
Stand Up Or Sit Out: Memories and Musings Of a Blind Wrestler, Runner, and All-around Regular Guy
A memoir about life lessons learned, especially through sports
Vision Dreams: A Parable
A sci-fi novella about how a dysfunctional society forces people to go to extremes, including four blind people who seek out artificial vision.
More About Tony…
Tony Candela has worked as a Rehabilitation Counselor, supervisor, manager consultant and administrator for more than 40 years in the field of blindness and visual impairment. His work has included promoting literacy and employment of blind persons and a special interest in enhancing the career preparation of blind persons who wish to work in the computer science field. He is a “retired” athlete, loves movies, sports, reading, writing, and music, including dabbling in guitar.