Hey Alexa, how long can a beaver hold its breath? I’m asking because I was kayaking last night at Totier Creek off the James River and I spotted a beaver swimming from one bank to another, his little head sticking up like a thumb and his body and paddle of a tail cutting a V through the smooth water. And suddenly he went under and never came back up. Believe me, I watched for him in all directions for, like, 10 minutes. The thing was gone.
Alexa didn’t care much about my explanation for the question, nor my decision to call the animal a “he,” but she did tell me beavers can hold their breath for up to 15 minutes on a single dive, plenty of time to get across a waterway without being bothered by a curious human in a kayak who was hoping for a closer view.
For stuff like that, this AI thing is awesome. It may not have all the answers yet, but it has a lot of them and at least when typing your question into a chatbot you can insist on a certain type of source (give me answers only from Nat Geo or National Wildlife or Beavers International magazine or a professional journal) to ensure the answer is credible. (You got me: There is no Beavers International. But if there were I would trust it fully for beaver facts.)
Next, I’ll admit this, with some hesitation: Recently for work I needed to quickly turn around a Q&A I’d done with a scientist, and the transcribed Zoom interview was all over the place and way too long. Desperate, I pasted the whole thing into a chatbot and asked it to organize it logically and cut out the ums and the time codes and the repetition and put it into a Q&A format at about 1,000 words. CHAT did all those things in less time than it took me to blow my nose, which was so helpful I wanted to cry. (Explains the runny nose.) How can one ignore these tools when facing this type of time-consuming effort? I decided I was okay with using AI in this way.
But where’s the line?
As a writer, I’m both amazed and terrified by all the things these tools can do. It’s not news that those in my profession are having to rethink how we do our work, not wanting to misuse AI but also wanting to be smart about how it can make our lives easier. The line is rodent-hair fine. (Beavers are the second largest rodents on Earth, after the capybara, AI told me.) Ask a bot to organize an interview logically? I’m okay with that. Ask it to cut 500 words without losing the gist of the text? Sure. But ask it to write an introduction to the interview based on the interviewee’s bio and papers on her website? Not okay with me. Creating text, actually writing for me what I would otherwise have to write, is a step too far. For me. At least right now.
It’s that last thought that scares me the most. Because consider this: If I have to write a press release about a scientific paper, it will take me quite some time to read the material, digest it, figure out what’s essential, and organize it properly. It’s not creative writing. It’s not based on my thoughts and intellect, nor, ultimately, any unique skills I might have. It’s filler for a template that helps someone else decide whether to go deeper into the subject. Why do I care if AI writes it or if I do?
And what about coming up with intelligent questions for an interview? I’ve always prided myself on asking things that are a little off to the side, not just the standard queries that will provide fodder for a standard article. Now, a bot can generate a long list of creative questions in, literally, about 2 seconds. Should I let it?
I’m still deciding. I’m sure all my writer friends are also looking for what’s okay and where that edge of the shore lies. Maybe they, too, are standing on the rocks and dipping in a toe to see how it feels. Are they, am I, willing to get a little muddy? More important, how’s the water if you fall in?
I guess I’d like to know from other writers and artists, what’s your line? We’re well beyond using bots as researchers, to sift through vast sums of information and answer questions about beavers (or WWII or cat allergies). As a kid I would have had to page through the Ba-Be volume of Encyclopedia Britannica, or ask a teacher or my mom to get the answer. That’s what research once looked like to me. The Internet is now Britannica, and my mom is gone so I’ve gotten pretty good at finding other experts these days.
But writing is a different story. If I’d wanted to write an essay about a beaver, it would have been based not only on my research but my thoughts and ideas of what was worthy of coverage; it would have been written in my style, organized my way. Yes, I would have plagiarized Britannica something fierce, but hey, I was seven years old. Later I’d learn to make it mine.
In this incredible and worrysome time of rapidly advancing computer tech, we have a lot to learn. About beavers, but also about boundaries, about the future of creating. All the arts are at risk of drastic change; hell, they are already something other than what they were. We have to think hard about truth and what we present as “ours.” And about how to keep up with those who take the leap if we prefer to stay there on shore. (Torturing this analogy to death, I know, but it’s an appropriate one. Because the beaver.)
Where’s your edge, writers? Have you put a foot into the river yet? Or are you trying to stay put as the shoreline erodes?
——
Photo by Tim Umphreys on Unsplash
I’m writing a novel based on the lives of a handful of real people (all of whom never did meet and probably won’t meet in my book either). During an AI training session at the university where I teach, I was offerred a set of sophisticated AI tools meant for researchers to consider. I took one of the people from my book who is always described as someone who had many lovers in any source I’ve read. Some of the lovers are named. Some aren’t. I asked the tool about her love affair with this one person she knew. I knew they had a love affair. I knew they stayed friends. I knew that both he and she married other people. I knew that she always saw him as a kind of mentor, and a former lover and a friend, but she also thought he was kind of foolish.
In her full maturity and sense of self, she would never have considered him the love of her life, but that’s what AI told me. He was important to her career, her self-developemnt, but the love of her life? No! No! No!
The AI wanted to make the narrative of her love life tidy. When I looked at the sources the AI pulled from them, none of them said she had super-enduring feelings for him, nor him for her. It was a conclusion that seemed drawn from a rom-com or a simple mind or offered to me because of however I framed the quesiton. And I would have believed it except I knew what the other sources had said, and across maybe 75 years, I felt I knew the truth because I also am kind of attached to the idea of who she really was. Not that I ever met her either!
Thanks for sharing this! Yeah, clearly AI still likes to make things up and draw erroneous conclusions, which is scary as hell. And as more people trust it and publish what it tells them, more “bad” info will become source material for AI’s continued learning. SERIOUS problem. We are in uncharted waters!!
Is that really a beaver, or a sea otter? If it’s a beaver, wow, how do you tell the difference?
Haha, yeah, turned out to be a sea otter. Shape of nose and nostril placement and placement and size of ears… it’s a great irony that I searched for “beaver” and got an otter but didn’t pay any attention! A lesson for me for sure. (I’ve since changed the photo.)
I routinely use AI to transcribe interviews (you need to check the transcript afterwards for hallucinations, but that’s still 10x quicker than transcribing it yourself), but I’m extremely surprised that anyone would use it to create a Q&A out of an interview transcript. The first and last time I tried to do that, the AI simply rewrote and summarized what the person said. That might be useful for some purposes, but for an article that’s presented as being in the interviewee’s own words, it was a complete red line for me, equivalent to fabricating quotes based on what I’d *like* the interviewee to have said rather than what they actually did say.
I wound up concluding that as tedious as it sometimes is, there’s a real art in turning a messy 10k-word transcript into a tight 1k edited-for-length-and-clarity article whilst still preserving as much of the interviewee’s original language – not just the gist of it – as possible. Can AI do that now? Because it certainly couldn’t when I tried it two years ago.
I hear you, and thanks for sharing this! In my case I “told” AI to retain the speaker’s actual words and phrasing; there was no summarizing her comments into something new. That definitely would be a no-no! And even though I felt the product was acceptable and still her voice, I’m still not 100% comfortable with letting AI organize for me, because of course that IS part of the writing process that takes thought and creativity.
You are right it is completely enticing, nerve wracking, incredible, terrible, powerful and shoddy. All of it and you are right where and what will be happening with it in five or ten years. Many things must change just like the calculator, the computer and so on. We keep improving and there is no end it seems. I think we just adapt and think wistfully back to before AI or be dreadfully afraid.