What in the word?

I’m barely competent with technology and frequently beseeching my daughters or tech-savvy young people for assistance with what I think are my computer’s “bugs.” It’s demeaning to feel so powerless and not in control — not at all the way I generally like to view myself!

So, I’m proud to say that I’ve been a relatively early adopter of one new technology: ChatGPT, developed by the company OpenAI. I’d read lots of articles about this latest advance in artificial intelligence (of the large language model variety), so I decided I needed to make this Chatbot’s acquaintance!

One big reason was a nagging feeling that this technology might threaten the very reason for a museum of words and language to exist — to promote reading and literacy. If a bot could read basically the entirety of human publications and produce coherent, grammatically written reports and write stories and papers and articles in mere seconds, why would we humans have any incentive to read, research, write, do homework — or learn anything at all?

And, indeed, when I had asked ChatGPT to respond to questions or prompts, it was remarkably adept at producing grammatical sentences, cogent arguments, and directly answering the questions I had asked. It was amazing! (In fact, it could have written this blog. Just saying…)

Various school systems asked the same questions and quickly moved to ban the use of ChatGPT in schools. They were especially worried about plagiarism. But then an interesting thing happened: “Hold on a minute,” some educators cautioned. There are lots of ways we can imagine using Chatbots to actually enhance education, perhaps by encouraging a Socratic method of teaching that involves the teacher, student, and bot in dialogue.

Also, for now, not all ChatGPT-generated content is consistently error-free. Sometimes a bot writes nonsense — it “hallucinates,” to use the art term. It might amass reams of information and weave it into a seemingly coherent response, but one that is actually full of mistakes or a nonsensical litany of unrelated information. Determining whether the bot-generated product is usable requires reading and critical analysis, the skills and abilities that define a literate human being.

My daughter and I took on a challenge sponsored by The New York Times to see whether we could accurately identify whether ChatGPT or a human fourth or eighth grader had written some writing assignments. Sometimes it was fairly easy to tell, because the writing samples contained the typical spelling or grammar mistakes of a young student; and sometimes the writing was too stilted and let’s just say, “robotic,” to be that of a youngster.

Here’s what that judging required of us as readers: the critical thinking and analysis that I just mentioned! We had to consider the spelling and punctuation and writing quality in order to assess it, including the abilities of a typical nine or 13-year-old! And that’s what some educators are now realizing — that when used in a smart way, ChatGPT can be a great partner in education, a “powerful learning tool,” as described by Kaleidoscope Education, a company dedicated to making high school more relevant and engaging.

When I taught elementary math, I had to make sure students developed something called “number sense.” Even if kids used calculators (remember how scared we were that calculators would mean kids would never learn how to add or subtract?), if they didn’t have number sense, they might not realize that they’d keyed in a wrong number, leading to an answer that was way off, perhaps by orders of magnitude. Only critical thinking and reviewing their work would help them avoid silly mistakes.

Similarly, with ChatGPT, no one should rely unthinkingly on the answers it generates. Currently, there is no replacing the need for human discrimination, evaluation, and critical analysis.

In the January 1, 2023, Sunday issue of The New York Times Magazine, A.I. researcher Yejin Choi lauded common sense as “the unspoken, implicit knowledge” humans have. She gives an example of how common sense eludes artificial intelligence: “Take defining birds as animals that can fly,” she said. The bot might be easily taught the fact that penguins are birds that don’t fly, but it turns out, even such a seemingly simple statement isn’t always true: “Birds can usually fly,” she says, but “newborn baby birds cannot fly, birds covered in oil cannot fly, birds who are injured cannot fly, birds in a cage cannot fly.” As she notes “exceptions are not exceptional. We humans can think of such exceptions without ever being taught about them.”

“It’s a fascinating capability,” she says, but “it’s not so easy for A.I.” But as a Planet Word supporter and investor in OpenAI reminds me: “ChatGPT learned everything it knows, which is way more than any single human, in about 90 days. So, mentally, it went from newborn to polymathic postdoc in 90 days. The next version is much more powerful and still learns it all, de novo, in a few months.” This could be our equivalent of a Gutenberg inflection point, the apex of the Anthropocene era we are living in.

So, after thinking about these challenges, opportunities, and even more powerful ChatGPTs yet to come, just over the horizon, I didn’t quite know what to conclude. I was still left with the worrisome question of whether a museum of words and language would soon be obsolete and whether reading itself would soon be unnecessary, but at the same time I am excited about how A.I. could actually enhance learning and boost our creativity and critical-thinking capacity.

It occurred to me — why not ask ChatGPT itself what it thought?

Will my museum of words and language become obsolete because of ChatGPT?

And here is its answer. I wish I could give it a hug!

It is unlikely that your museum of words and language will become obsolete because of ChatGPT. ChatGPT is a tool that can assist with language-related tasks, but it cannot replace the experience of visiting a physical museum and engaging with the exhibits in person. Additionally, a museum can provide context and background information that ChatGPT may not be able to provide. Museums also have a valuable role in preserving and sharing cultural heritage and history, which is important for future generations.

You go, bot!

— Ann Friedman, founder of Planet Word