ChatbotGPT: The New Mechanical Turk

Technical drawing of the mechanical Turk.

A cross-section of the Turk from Racknitz, showing how he thought the operator sat inside as he played his opponent [via Wikipedia].

I am unimpressed with the ChatbotGPT. I asked it to write a sestina, and it could not follow the formula or even write decent rhyming couplets which it used to substitute for the formula for a sestina. It managed to produce some boring and obvious haiku. It is just as bad at writing essays. I asked it to write a five paragraph essay on the literary influence of the poet, Geoffrey B. Cain, and it wrote five paragraphs of drivel that was actually about Hemingway that sounded like a high school essay from a student who only read the back of the books (you know, something that I might have written). I asked it to write an essay skeptical of the concept of artificial intelligence, and it came up with a paper that was basically repetitive, shallow and unimaginative.

ChatGPT was created by the San Francisco based company OpenAI in 2020. ChatGPT was created from GPT-3 which uses a type of AI known as a “large language model” that creates text by combing through masses of data and learning how words and phrases work. The key here is that it “creates text,” it does not think. Its potential applications include summarizing formulaic writing like legal documents or aiding computer programmers with the more repetitive and tedious tasks of programming. ChatGPT is an advanced iteration of GPT-3 that is optimized for natural language dialogue with users. ChatGPT does not do basic research: I imagine that the company could face some legal intellectual property issues as well as the same challenges our students face with how to decide whether a source is academically credible.

So what does it do well? Why are educators supposedly panicking? It can produce essays based on basic prompts that are typically found in college courses. Ideally, essays are an opportunity to have students demonstrate how they apply knowledge: how they understand new information in light of previous learning and, most importantly, experience. Any question an instructor asks that can be Googled has no place in the classroom, or rather, the classroom should be the place where students establish the facts of the case in discussions with one another and the teacher. The essay or assessment should be a record of the process the student took to arrive at new knowledge which, again, is presented back to the class. The only people who should be afraid of AI taking over the classroom are those whose teaching has become rote and mechanical, or those who think that teaching and learning is a product, not practicing a process (forgive the nearly robotic alliteration here).

In my classes (composition, literature, and adult basic education), essays are not produced. I do not want a three-page paper. I want a portfolio that includes a concept map or rough outline, a discovery draft written with fellow students, a rough draft that includes a record of a visit to a tutor (online or at the Tutoring Center), and a final draft. One of the many things that AI cannot reproduce is the relationship that is at the heart of teaching which should be central to any real essay. Even in a “subject matter” class like science, I want to see a concept map at the start of the quarter of the students’ current understanding and an expansion of that concept as they learn. Again, this is just in my work. I know that there are other ways to engage authentic student learning such as having students re-write essays for other audiences, have them change the genre of their work, or using prompts that emphasize students’ lived experiences.

We have been here before. The Mechanical Turk, also known as the Automaton Chess Player, according to Wikipedia, was a fraudulent chess-playing machine built around 1770 and exhibited by various owners until its destruction by fire in 1854. For 84 years people believed that a mechanical man, a collection of gears and pulleys, in the correct arrangement, could play chess. There was actually an expert chess player inside the automaton who would communicate moves to the exhibitor through various hidden, mechanical means. Yet, people believed many things about it from that it was an actual mechanical chess player to the idea that it was a machine possessed by an evil spirit. Its reception was a mirror of the society that produced it. I think this is happening with so-called “artificial intelligence.” That term, AI, implies that we know what intelligence is, how new knowledge is created, and what it mean to think (which we don’t on all three accounts). It also implies that intelligence is an algorithm, and that if we had an algorithm with sufficient complexity, it would either be intelligent or would mimic it enough so that it would not matter if it were real or not.

It is not enough to be against an idea or even just skeptical of it: we should have a clear idea of what the alternative is and a creative application to offer. In this case, essays in the media that announce the “Death of the Essay” are not very useful. They imply that the essay was not dead before. The way that the essay is often used (as a product rather than a process) has failed education for decades. Current outcomes for English classes say things like “The student will produce an essay of ten pages…” The way they are assigned inspire and rewards paper recycling, academic dishonesty, educational travesties such as Course Hero, buying and selling of papers, etc. We have needed to get away from the outcome of a class being the production of a paper to the demonstration of new knowledge and intellectual growth. This would require a great amount of effort on the part of educators: it would require a fundamental change in how most educators think about education. We should at a minimum be asking ourselves how our college assessments are engaging the information with their students’ previous knowledge and experience. Just as important as writing is the sharing and presentation of information which is why I like to tie papers to larger group projects that ask for students to engage with one another. I am hoping that conversations around academic honesty and new technologies will lead to conversations that help education to grow into this brave new world.

The ironic thing is that by playing with ChatbotGPT, we are expanding its vocabulary and its natural language processing abilities. We are the ones making it work. We are the consciousness and creativity behind the algorithm. “It” does not think, all it does is follow rules and process language. As long as we define thinking mechanistically, we can look at ChatbotGPT with wonder.  We are inside the mechanical Turk, and we don’t even know it.


This entry was posted in AI and tagged , , , , , . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.