ChatGPT and other artificial intelligence (AI) tools are the current buzzwords taking the world by storm and drawing millions of users. However, that only means it’s more important than ever to exercise caution and think critically before incorporating new technology into the workplace.
But what exactly is ChatGPT? Is it truly a revolutionary writing tool, or is it better as a curiosity? Let’s break it down.
Defining AI and ChatGPT
Artificial intelligence is a technological field that aims to allow a computer to problem-solve in a way similar to humans by combining computer science with enormous datasets. It has broad applications, including writing, art, closed captioning, stock trading and more.
ChatGPT is a chatbot powered by AI that can be used as writing assistant. With its breadth of datasets at its disposal, ChatGPT is remarkably responsive to user inputs and can decently mimic human speech patterns and incorporate context from earlier inputs. Recent developments even gave it the power to search the internet.
It’s important to note that an AI program is not sentient. Investors and developers do their best to make it seem like magic, but even generative AI such as ChatGPT is just an algorithm – albeit an impressive and complex one. It is equally crucial to note that since AI is programmed by and trained on humans, it has baked-in biases.
ChatGPT is certainly appealing to businesses – writing takes employees time, and ChatGPT can spit out a flurry of words in seconds. Many organizations see ChatGPT as an opportunity to save on labor costs… but is it a suitable replacement for human writers?
ChatGPT’s Limitations as a Writer
The truth is that ChatGPT is not perfect – and may never be. It’s still largely in development, so while it will likely improve down the road, it’s currently plagued by limitations. Here are a few that directly affect users trying to take advantage of its long-form writing capabilities.
- As a coded program, ChatGPT has no obligation to tell the truth. Its only job is to provide a logical response, and not in the way you’d expect. The algorithm predicts what its next output should be using statistical logic, so if ChatGPT cannot find an answer to a query, it will make one up by just putting one word in front of the other. It has no capability to fact-check itself, and it can insist that it’s correct when it’s not. Altogether, this means that ChatGPT and other similar AIs have a substantial chance to generate misinformation.
- ChatGPT is programmed to provide different answers. The intent is to make it sound more human, but sometimes it mixes unrelated data or changes its answer for no discernable reason. It could be as odd and easy to fix as changing someone’s job title in an article, to as egregious as altering quotes and citing completely unrelated or made-up sources that sound real.
- ChatGPT only sounds almost human. Since it uses statistical logic to craft sentences and long-form pieces, ChatGPT’s outputs are still fairly robotic. It will likely never fully iron out that particular kink, though it may get more fluent and human-like in future updates. As of now, however, the long-form outputs often read like a fifth-grade student’s essay at best, requiring a human hand to make it a more compelling read.
- ChatGPT does not “think.” As an algorithm, it often fails to grasp nuance and is not currently capable of insight and delving deeper into the meaning of certain topics. It is far better suited to analyzing data than it is to interpreting data. In other words, it’s great at sorting your inputs into neat categories, but not at writing a thesis statement. This also means that it cannot create original ideas, as it can only work with what it has. It’s not yet clear if it will ever fully get past this hurdle.
What does this mean for writers?
ChatGPT can produce adequate writing at inhuman speeds, which may speed up workflows that rely on formulaic writing. In the end, however, it’s the equivalent of monkeys on typewriters. And we’re not even touching the possible legal and privacy infractions behind ChatGPT’s learning dataset.
Ultimately, we can confidently say that the human touch is still a critical part of writing, and ChatGPT simply can’t hold a candle to human writers in terms of quality and depth of meaning, at least for the immediate future. ChatGPT may incorporate context a lot better than its chatbot predecessors, and its outputs may be passable in certain circumstances, but it still does not know what “good” writing is – and likely never truly will.
Sources:
https://www.forbes.com/sites/blakemorgan/2023/04/28/cons-of-chatgpt-for-customer-experience/?sh=47b9fe9f34b3
https://www.ibm.com/think/topics/artificial-intelligence
https://www.makeuseof.com/openai-chatgpt-biggest-probelms/
https://www.tomsguide.com/news/chatgpt