Ever since artificial intelligence made it to mainstream media, a big debate that has existed around the topic is whether it can replace humans or not. And after OpenAI came out with ChatGPT, a language generation model, these conversations are resurfacing in a big way. ChatGPT is capable of writing essays, stories and poems, creating large set of content, creating news stories, writing codes and much more. But is it so good at its job that it can replace the need for humans? Some experts have come out and put the AI through its paces and what they have learnt is that ChatGPT is not without its flaws. Check the 5 major drawbacks of this AI chatbot.
Drawbacks of ChatGPT
These flaws were curated in a report by the Search Engine Journal which collated various studies and interviews with experts to highlight the areas where ChatGPT falls short. Let’s take a look.
1. The lack of phrases: One particular area where ChatGPT fumbles is usage of idioms. According to a research paper published in 2022, lack of idioms in a piece of content can be a tell-tale sign that it was created by an AI and not human. It concluded, “The presence of Shakespearian archaisms, writing cliches, and idioms is likely far more common in book text than in computer-generated blog posts and news articles. Of these complex phrasal features, idiom features retain the most predictive power in detection of current generative models”.
2. The lack of expression: Another possible flaw in ChatGPT can arise from the fact that it can never truly experience things and as a result, it cannot express either. Its entire experience comes from the data which is fed to it and as a result, creating true art might be difficult without experiencing it and only relying on data coming from someone else’s experience. Australian musician Nick Cave asked ChatGPT to write a song in his style and after posting it, he mentioned in his blog, “ChatGPT’s melancholy role is that it is destined to imitate and can never have an authentic human experience, no matter how devalued and inconsequential the human experience may in time become”.
3. The lack of insight: Business Insider came up with a report where professors who caught their students cheating with ChatGPT revealed that catching essays written with the help of AI is difficult but not impossible. What they noticed was that the pieces had the writing ability of someone who was well versed in the topic, yet it demonstrated no true insight that comes from being at that level. As a result, the pieces appeared fluffy and superficial.
4. The lack of direct answers: It is a flaw that OpenAI itself admits to, but it has now been published in a paper which highlights that while ChatGPT is able to answer questions, it is not able to get to them directly, once the topic becomes too complex or niche.
The research took a set of questions and answers and asked a group of subjects whether they prefer the AI answer or the human answer, without telling them which is which. The results showed that people preferred human answers when it came to answering medical questions because humans preferred direct answers – something the AI didn’t provide.
5. ChatGPT fabricates facts: A research titled ‘How Close is ChatGPT to Human Experts?’ found that when answering a question that requires professional knowledge from a particular field, ChatGPT sometimes fabricated facts in order to give a response.
Leave a Reply