ChatGPT. If you haven’t heard of it from some source yet, then I can only congratulate you. It is described as an ‘AI Advanced Chatbot’ by its creators, OpenAI, and it has taken the world by storm ever since its launch some three months ago. Universities have sounded General Quarters over the death of independently researched and written essays, corporates have started using it to automate their writing needs, and other tech giants like Google and Baidu have pushed their own development teams into overdrive in order to launch their own chatbots.
So why is this new creation so legendary, you ask, and what effect might it have on me, when I’m sat in my warm, cozy corner of the library, trying to do that next essay or problem sheet? The base idea is that the chatbot can produce text based on a prompt, from its deep reading of everything (or, at least, a lot) on the internet. Whatever the topic might be, chances are ChatGPT knows most of the basics, and might know some advanced details too! (This is probably the right moment to assure you that I have not used ChatGPT to write this article, though I did use it for lots of research prior to doing so.) That doesn’t mean it can produce absolutely everything, of course – creators have recently cracked down on a lot of immoral and unethical topics in their latest updates.
Text, or language, in technospeak, isn’t the half of it. ChatGPT can also answer technical questions, write code, and solve maths and science problems, like a more conventional computer, perhaps. If you haven’t seen the TikTok videos yet of ChatGPT producing perfect Python, C#, and C++ codes, then you’re not missing out on much. Even experienced developers are taking longer than the computer to belt out that much code. In our time, lots of STEM (and non-STEM) graduates join the tech workforce, working in the various echelons of the digital world, but their jobs are as much at risk of being replaced by AI as the copywriter or the paralegal.
So, what has been the effect on OpenAI? The San Francisco, CA-based company now projects a billion US dollars in revenue for the next year and has just received an infusion of ten billion dollars from Microsoft. Microsoft, which supported the company in the years prior to the launch of ChatGPT, has now announced its desire to integrate the Chatbot into its Bing Search Engine, Edge Browser, and Outlook email client, creating a space for ChatGPT within the Microsoft ecosystem. This might mean, for example, that soon, within Outlook, you would provide a prompt and the client would write the email for you, or you’d run a search, but the AI could find sources that don’t exactly match your search terms but that it believes are relevant. It could also make content up, as it is trained to do.
This is perhaps the best time to move onto more juicy material. However, before I do that, an overview of the essay-writing ecosystem before the launch of ChatGPT might be in order. It is worth mentioning that as of May last year, it is illegal to advertise or provide essay-writing services for any University or College assignments in England. Nevertheless, essay-writing services have been undeterred. A quick Google Search offers numerous options, and I only found one that identified my English IP address and thus informed me that I unfortunately could not use the services of that website. The others all compete with one another, under different categories: Oxbridge essays, Exam Essays, Dissertations, Theses. Do you want to pay per page of content, or per hour spent working on your assignment? Do you want discounts on bulk purchases? They have you covered for those as well. Whether your deadline is tomorrow or in two weeks; Though, the longer you give them, the cheaper it is for you. All-in-all, these essay-writing services make up an entire market, giving you all the options you might need to cater for differing tastes. And they waste no effort in marketing their services. A poll run by Cherwell noted that 86% of respondents had seen ads for essay-writing services. The ads for these services dwell in the depths of all those freshers’ group chats that we join after offer day. They promise “100% refunds in case of dissatisfaction” for assignments completed by the “Assignment King”. Some offer a “free plagiarism report”, because there’s no way that they picked up the words in your assignment from somewhere else. There are Instagram accounts with professionally produced reels advertising the various kinds of essays, problem sheets, and assignments you might be able to commission. You want something done? They will complete it for you.
These are all under threat from ChatGPT and the other AI platforms that will follow it. Why would I, or you, pay £100 for an essay from the essay-writing company, when ChatGPT could analyse my writing, and then produce an essay on that topic which I might have to make minor edits to, all for free? (It might start charging later, because OpenAI is still running a beta version for personal use.) Do you need a prompt on how to approach that problem sheet or that coding question? Just ask ChatGPT. Do you need to write 1000 personalised emails? Why make a human do it when ChatGPT could draft them all, and then you only make edits? With some amount of human editing, it is impossible to differentiate between human-produced content and computer-produced content now. It might even be difficult to differentiate between unedited ChatGPT content and Human-produced content, as a New York Times investigation discovered.
Now, I’ve probably given you enough background on the impact of ChatGPT on the world outside us and what one might expect it to do. The rest of this piece focuses on my personal experiences with it, both inside and outside Oxford. We start with my family. My father works in what I call a standard “finance-bro” job in the City, acquiring and operating tech companies. This basically involves a lot of emails, sometimes cold emails, introducing his business pitch and the activities of his employer. He has now invested in the paid version of the Chatbot (ChatGPT Plus), and it drafts his emails based on prompts from an excel sheet, for $20 a month. He still has to edit them, of course, because the computer still hasn’t figured out the exigencies of City corporate conduct. Once it does that, it might mean the end of easily available jobs in finance. Or perhaps not. The Chatbot decided that a certain maker of restaurant software was nearing collapse, because, logically, “restaurants produce perishable goods that don’t last long”. It was reported to OpenAI. If you ask me, it is a computer so it won’t make that mistake again.
Now, after extensively having a look at everything outside of Oxford, it’s time to delve into how Chat GPT exists in our hometown. I have seen a friend open their economics problem sheet in one tab, and ask ChatGPT for methods to solve it in another. Another Economics and Management-reading friend of mine said to Cherwell, “It’s more than a tool, it’s a friend. But it lets me down most nights.” Make of that what you will. This same friend also noted that their parents used the Chatbot to write a dramatic letter to their MP requesting support on their delayed application for citizenship. The letter that was produced was apparently a bit too dramatic (their life was under threat, pending approval of British citizenship), but their parents still sent it in. No response was received from the MP, however.
Perhaps you’re itching to know, perhaps you aren’t. I haven’t used it yet for any of my (humanities!) work. I did try essay prompts on it, just to compare with my own work. The earlier version (in December) still could not cite sources but did produce content on a given essay prompt, about 500 words of very general information. The new and updated version, probably produced after OpenAI caught wind of their model’s widespread usage in higher education now includes a disclaimer: “as a language model, it is not possible for me to conduct original research or cite primary sources. However, I can provide you with a general overview of the current academic consensus on the topic, as well as some key theories and points of discussion.” That’s that, then. All our hopes for having the computer write our essays gone, you would think. The correct answer is Yes and No. When I asked it to write a piece and not an essay, it still belted out a considerable amount of words, roughly equivalent to an A-Level essay, and did note two sources at the end. So, it won’t write my essays for me, but will still provide an overview of the topic to ease me into the books, somewhat like a human-produced overview article. I should mention here that another poll run by Cherwell resulted in 15% of respondents admitting to using ChatGPT in their essay-writing, at some stage. The days of AI writing all our essays are definitely not upon us yet.
The fun part starts here! Yes, the AI model is useful for academic purposes, and for business purposes, but you can also use it for entertainment. One evening, we decided to stress-test it. This underscores, of course, the need for human ingenuity and sentience to provide prompts to the computer. I don’t believe it could have come up with ideas without any human support. The first question was on the termination of a stick insect. Hard luck. ChatGPT does not produce any content on harming a living organism, insect though it may be. What about nefarious activities at the Driver and Vehicle Licensing Agency (DVLA)? ChatGPT has a story ready to go (of course, it does provide a disclaimer that the DVLA is a reputable organisation and there are institutional processes to deal with any corruption within it, as there should be.) Maybe a pickle-tickling rhapsody in sixteen lines? Pickles, they do certainly have power.
Our last prompt before our time ran out (ChatGPT stops responding if you send it too many prompts within an hour) was asking it to produce a psychedelic prose poem about table legs. Those screenshots are not reproduced here, but they did include a disclaimer that the term “psychedelic” was used in a metaphoric way and the AI does not support or condone”‘the use of illegal substances”. My recommendation: Try ChatGPT, or any of the other AI Chatbots, for yourself, and they might just surprise you. “You” are still a very important part of the picture here.
Where does all that content (both mine, and ChatGPT’s) leave us? If you ask me, the AI age is probably around the corner, if not already here. ChatGPT can produce some quite respectable text, code, and solutions to problems, but it is not without its pitfalls. It can be used for academic work on some level, but it is not going to complete our assignments for us. Sea change might be coming, though, both in academia and in the glass towers of corporate officialdom, so it does not hurt to prepare and acclimatise ourselves to AI picking up some of the slack that we might have done earlier. Lastly, using AI can be fun too! The next generation of children might play with an AI server, instead of a programmed game, just as we are playing with a proto-AI platform today.
Note: The above piece is produced from a given prompt and it is important to note that my impacts may be different in a real-world situation. As a language model, I don’t have the power to predict events in the real world, as they often happen with no warning. Please be mindful of accepting the above text as factual.