What makes the world go round? The answer to that may no longer be love or money, but rather, artificial intelligence (AI). AI is everywhere, beginning with the morning weather report to the potential matches you swipe on a dating website during lunch. If that is not enough, AI is now also able to write college papers for you.
A recent experiment by Eduref, a resource for students and educators, showed a deep learning language prediction model called Generative Pre-trained Transformer (GPT-3), created by the privately-held San Francisco startup OpenAI, can write term papers.
For the uninitiated, GPT-3 can respond to any text entered into the computer with a new text that is in perfect sync with the context. Type a sentence into the search box, and you will be amazed to find a perfectly relevant response. That means GPT-3 can boost human effort in a wide variety of situations ranging from customer service to conducting due diligence. In this case, it is your average writing assignment.
We have Artificial Intelligence planes, vacuuming robots, chess playing machines… Why hasn't someone invented an AI term paper writer???
— S. (@SilkySil7) July 12, 2011
The Eduref team “hired a panel of professors to create a writing prompt, gave it to a group of recent grads and undergraduate-level writers, and fed it to GPT-3 and had the panel grade the anonymous submissions and complete a follow up survey for thoughts about the writers.” The results showed the computer program was able to achieve passing marks in all the term papers.
What’s more striking is while the average time taken to complete a term paper by students was 3 days, GPT-3 finished the same task in between 3 and 20 minutes.
However, the experiment does not address the ethical dilemma of AI composing essays. As science advances, it seems likely that technology such as GPT-3 will become increasingly able to produce high-quality research papers without a student having to lift a finger. This, of course, will be a major breach of ethics and cause major harm to academic honesty and integrity.
Test Results
From History, Research Methods (COVID Vaccine Efficacy), to Creative Writing and Law, the writing prompts were given on a wide variety of subjects. The program scored a “C” across four subjects, failing one assignment on creative writing.
This shows that GPT-3’s writing skills are mostly technical. In comparison to the computer program, the other writers earned from A to B+ from the reviewers.
Overall, despite the somewhat technical nature of the writing, the evaluations suggested that GPT-3 was well able to replicate human writing in areas of grammar, syntax, and semantics.
Despite some pitfalls, the experiment proved how natural language processing can be used to write almost anything, including a novella that was nominated for a literary prize: The Day a Computer Writes a Novel.