In parties, social gatherings, and classrooms, I have had a number of questions asked about ChatGPT. It’s in the news and everyone’s raving about it.
In this post, I will answer the questions I have been asked the most. If you have more, feel free to share them in the comment section!
What is ChatGPT anyways?
Let’s kick off the article on ChatGPT by understanding what is ChatGPT and what is all the buzz around it.
ChatGPT stands for “Generative Pre-trained Transformer-based Chatbot”. It is a type of AI-powered chatbot that is pre-trained on vast amounts of text data (literally the whole internet!) and build using advanced deep-learning models
ChatGPT is a highly advanced conversational AI technology that can communicate like humans in a way that is indistinguishable. What makes it very special is that earlier technologies were not able to interpret the subtle intricacies of human communication, for example differentiating between humor and sarcasm – but ChatGPT can interpret them with speed, accuracy, and fluency.
Well, how does ChatGPT generate responses?
It makes use of text data it has been trained on to provide coherent text output. Computers don’t understand words, so the words are converted into numbers and once they are in vectorized form, statistics and probability are used to generate responses that follow the writing pattern as it has been trained on.
Can it be useful for me?
Of course! It helps you generate ideas, rewrite your text, create outlines, fix grammatical and spelling errors, summarise the text and so much more. Its ability to provide relevant yet fluent responses with speed and low cost are the biggest perks it offers.
However, ChatGPT comes with its own red flags. These red flags include potential issues with privacy, bias, and accuracy.
I am a content/copywriter, will ChatGPT take my job away?
I don’t think so!
ChatGPT doesn’t have human emotions, empathy, and personalization that people will start craving once the craze over ChatGPT (or any type of conversational AI, in fact) fades away. I personally have used AI and I end up finding the text generated by it very robotic (not a coincidence, right?), impersonal, and preachy. So the best I can use it is for creating outlines but that also requires quite a bit of modification because I tend to have my own perspective on things and I structure my articles/letters/posts in a particular way, kind of characteristic to my writing.
Elaborating more on publishing, I have always been a huge fan of personalized, targeted, quality content. I also go by the mantra that if you are writing for everybody, you are writing for nobody and AI writes for everybody. I believe that each person on earth is unique because they have come from different places, been to different places, interacted with different people, read different books, and reflected on them differently. AI has done everything — so it doesn’t necessarily leave an impression of being unique.
A company might be able to publish more content with AI but establishing a connection with its targeted audience might become difficult without adding a splash of their unique flavor to words. I took a couple of quizzes and it was quite easy for me to spot the paragraph written by AI. I am sure other humans can do it as well. It won’t be long before we will have AI that can differentiate between text written by humans v/s AI. If this becomes a possibility then Google’s algorithm will start ranking human-generated content over AI-generated content. That means double efforts for companies, to first work on content generated by AI and then modify it to sound human.
What does all this mean for a writer trying to find a space on the internet?
I think this presents a huge opportunity for creators to reach out to readers with their human quotient. It also means working with AI, to let it do what it does best — help you create an outline, brainstorm some ideas and you can build on this foundation for creating content that sounds like you and is tailored for your audience. Working with AI will also help you write more and edit your work.
I am a coder, will ChatGPT take my job away?
I have tried ChatGPT to write code for me and instructed it to add exceptions and create test cases. I wouldn’t lie, the results were pretty good, to the extent that it scared me.
As a data scientist, I do some things that Generative AI is not (yet) capable of doing. I use data, find patterns and relate it with business needs and make sure to include domain knowledge. My work is less about writing the right code, it’s about getting insights.
But when it comes to writing Machine Learning or Deep Learning models, code becomes important. I can decide on which model to use after weighing the bias v/s variance, training time, and interpretability. ChatGPT is able to mention the pros and cons too — and this is the part that scared me a little bit.
However, the catch here is that it (still) can’t decide if there is a need for an ML model or not. Which requirement in the whole project would require an ML model and which can be done with other software engineering/data analysis methods? How to relate the ML model to the business value?
ChatGPT can be really helpful to save time from manual data tasks and writing basic code. But, it comes with its own security issues and correctness concerns. Apart from being inefficient or incorrect, these codes can overlook vulnerabilities and land enterprises in trouble. For example, take a look at what happened at Samsung.
“Samsung workers have unwittingly leaked top secret data whilst using ChatGPT to help them with tasks.” — Techradar reported on April 4, 2023. These employees (allegedly) had input meeting notes and source code revealing information that Samsung wouldn’t have made public otherwise.
Humans are still required to make sure systems remain secure, confidential information remains within the organization, and protect code bases from vulnerable attacks. All this, while applying their creativity and novelty while coding.
In conclusion, make ChatGPT your friend, but use the beauty of the human brain and leverage it using ChatGPT.