Kirsty McHenry | It’s hard to say whether language models like ChatGPT are a blessing or a curse
The use of language models like ChatGPT has become increasingly normal, writes Kirsty McHenry. But that is not necessarily a bad thing. “Essentially, these language models are tools, not oracles; their skills begin and end with what has already been discovered.”
Some days, after many hours hunched over my keyboard, confused, desperate and with hardly a sentence to show for it, I find myself in want of guidance. If only I could trek to a secluded cave to seek out the one who could answer my questions. I imagine a voice emanating from the darkness, whispering all the knowledge necessary to reach the word count – an oracle of sorts who could offer me an insight into what my conclusion holds. Now, thanks to the power of modern technology, those of us in search of answers can speak directly with such an omniscient voice without even having to leave the house. Its name? Chat-GPT.
Students today are often painted as frequent and impetuous users of Large Language Models (LLMs), like Chat-GPT, wilfully feeding them assignment prompts at our own peril. It’s not hard to picture a lazy student being easily allured by the quick and effortless solutions that an LLM can offer, thoughtlessly throwing away their valuable chance at an education in the process. As the story goes, they will not realise what they’ve lost until it’s too late and they’re left standing at their graduation with a hollow degree and no real skills to show for it. At least, this is the popular narrative in the academic sphere. In actuality, most of us have a much more nuanced relationship with the new technology.
Two years ago, universities – rather than seeking the help of famously unreliable plagiarism detectors – might have had an easier time catching LLM users by setting up confessionals and waiting for the guilty to surrender themselves. Nowadays, the shame around using LLMs in academic work has somewhat dissipated. With LLMs no longer lurking in the shadows, their presence on our campuses (and in our lives) has become undeniable. We are already long past the time when students could be realistically expected to deny temptation. It’s little wonder people are drawn to LLMs, which appear bright and alert in comparison to the human machine, which whirs and tires like a Windows 98 operating system that’s in constant need of coffee and bathroom breaks.
As we procrastinate in the toilet, the LLMs grow stronger with every update. Still, the LLMs’ ability to produce a hundred-page paper in under a minute is no act of divinity. LLMs are not really speaking from a void but are instead pulling from a vast, rich library of human knowledge that is just beyond our sightline. Without active human input, LLMs can only puppet the creativity of others. A hundred pages of regurgitated academic jargon is unlikely to contain much substance without the input of critical thought. Fundamentally, LLMs are tools, not oracles; their abilities begin and end with what has already been discovered.
As students, we spend years navigating our academic fields. We broaden our expertise by wading through unnecessarily wordy academic articles, their treasured insights hidden deep in the confines of the conclusion, and slowly refine our skills over countless tired nights spent battling one assignment after another. While, on the face of it, LLMs could enable future students to skip this drudgery, it’s more likely they will be expected to develop their critical faculties some other way (potentially with the incorporation of LLMs). As blissful as life could be without them, the necessity for critical thinking skills is not going anywhere. It remains true that progress can only be made by people building on the work of other people (and preferably crediting them). Ultimately, humanity is not Windows 98, we cannot be rendered obsolete. Right?
