In the spring, The Seattle Times published an op-ed piece by a couple of academics from the University of Washington warning that AI is creating an increasingly dumbed-down “monoculture” that threatens the very nature of human intelligence. (“AI is about to do all our thinking for us. Here’s why that’s bad,” May 17)
I, for one, am not worried, and here’s why: I have found in Chat GPT a “Great Professional Teammate.”
As with most powerful technologies, AI can be used to produce great evil or great good in the world depending on who uses it and to what purpose. I have heard a couple of speakers talk about AI and its use. One presentation, given by an AI development pioneer, encouraged us to think of it not as “artificial intelligence” but rather, as “assistive intelligence.” I’ve also learned that the key to using it effectively and avoiding its infamous “hallucinations” — when generative AI presents something untrue as a fact — is developing the skill of prompt engineering. It’s not about the historical Casey Jones getting his train to its next destination on time; it’s about asking chatbots the right questions in the right way.
In my work as critic, editor and publisher of the Seattle-based Heritage Music Review, GPT has saved me countless hours of tedious website and Facebook page navigation to determine which artists to list in my monthly performance calendar.
That example represents the kind of research help that has earned Chat GPT many thousands of loyal users, but I, as a blind website owner relying on screenreader technology, face unique challenges that human tech support advisers would be at a loss to meet.
Anyone who has ever called a software provider’s support department has some inkling of the difficulties I endure regularly. Apple advisers, for instance, are only trained to answer questions related to their own products and are even forbidden by company policy to help with third-party apps. The problem is compounded when the user is dealing with a specialized assistive tech company. Since my work requires my Braille Android tablet to interface with numerous third-party apps, I need an assistant capable of transcending all those barriers, but given the way technical support generally works, no such human seems to exist. Under ordinary circumstances, these obstacles would already have stopped me dead in my tracks months ago if not for the cross-platform capability of my Great Professional Teammate.
To publish each monthly issue of Heritage Music Review, I need my Braille tablet, the Amazon links by which I earn commissions on the books and albums my readers buy and GoDaddy’s web hosting service. To accomplish this gargantuan multitask, I prompt, “You are an accessibility adviser familiar with Humanware’s BrailleNote Touch Plus Android tablet, Google Docs and the GoDaddy web hosting platform. What’s the easiest way to create a publishable DOCX or HTML file that incorporates my Amazon Associates product links?” It took us a few tries to create a workflow that works for my trusty BrailleNote, but now I have all the tools I need to publish every month and even update archival issues with expired product links.
Yet amid all the miracles my teammate performs for me, there’s one thing I never ask it to do: Write my articles. I’m perfectly capable of doing that for myself. With regard to AI and many other things in life, rock ‘n’ roll legend Little Richard said it best: “It ain’t what you do, it’s the way how you do it.” I strongly suspect that if enough people learned the invaluable art of prompt engineering, the cream of internet content would rise to the top, the vast quantity of AI-generated slop we now see would slowly but surely be relegated to the cyberdump where it belongs, and, like the once-dreaded Y2K computer scare, the monoculture that the UW authors envision would take its place in history as a plague that never materialized.
