By Gowri Kishore
Within just two months of its launch, ChatGPT, a chatbot application built by Microsoft-backed OpenAI, reached 100 million monthly active users, making it the fastest-growing consumer app ever. For perspective, TikTok touched this milestone at nine months, Instagram at 2½ years and Facebook at 4½ years.
What makes ChatGPT so fascinating? For one, it seems to be the Kamal Haasan of generative AI applications. A sakalakala vallavan or jack of all trades that can do reasonably well at nearly everything—answer queries, write essays, poetry or ads, write and debug code, summarise papers, and generate an endless stream of ideas.
Even better, it is in the form of a chatbot with which you can have conversations as you would with a human. Type in a prompt in everyday English and ChatGPT interprets it correctly and responds. You can add details, seek edits and ask follow-up questions until you get exactly what you want. If ChatGPT makes a mistake, it even apologises!
Given that it’s easy to use, ad-free, not SEO-driven and gives precise answers to queries, initial predictions were that it will replace search engines. In response, search engine giants announced that they will integrate chatbots into their search experience. Google launched its own AI chatbot named BARD while Microsoft’s ‘new Bing’ integrated ChatGPT and is currently in beta mode. Earlier this week, OpenAI released GPT-4, a new version of the underlying AI technology that powers its viral chatbot.
But within India’s startup ecosystem and creator community, ChatGPT is being used in far more creative ways.
In user experience and visual design
A key part of digital interaction design is understanding user flows, the step-by-step path taken by a user to complete an action on a website/app. ChatGPT helps designers create user flows in a fraction of the time it would otherwise take. “I use ChatGPT like an assistant,” says Manish Minz, design lead at a fintech firm. “I ask it to explain user flows or suggest font combinations for something I’m designing. It jogs my brain and helps me ideate and iterate faster.”
Asha Haridas, a brand designer who works for a SaaS startup, uses ChatGPT to better articulate her design decisions. “People outside the design team, especially the leadership, need some convincing about design choices. ChatGPT helps me explain my rationale wonderfully well. I give it prompts like ‘Build the rationale for using noise in illustration’ or ‘What do brushstrokes signify as a design element?’ The answers help me make presentation notes and design documentation much faster.”
Just as writers rely on graphic design tools like Canva to lay out their copy or create reference designs, designers are using ChatGPT to convey their message better.
“As product designers, we create a lot of screen mockups. Usually, we put in lorem ipsum as dummy text while waiting for a writer to provide the actual copy. But now, our reliance on the writer is reduced. With ChatGPT, we are able to generate usable copy ourselves,” says Minz. This helps designers create more meaningful mockups even though a writer needs to vet or refine the copy later.
“The more context you give the tool, the better its output,” points out behavioural researcher Dharmesh Ba. “A basic prompt like ‘Give me some help text options to download a tax explainer document’ will yield generic results. For more relevant answers, give it a layered prompt like ‘I am building an online tool for accountants. Give me some help text options, etc.’ Add even more detail on tonality (‘This tool will be used by Europeans whose first language is not English’) and you will get a near-perfect answer.”
In content creation by marketers and writers
When ChatGPT was first launched, a slew of social media posts predicted that writers would soon be out of a job as lean/frugal marketers turn to the app for their content needs. At first glance, the application seems to have impressive writing chops.
Abhishek Sebin, co-founder of video commerce platform Instasell, uses ChatGPT for everything from marketing ad copy and social media posts to cold sales emails. On one memorable occasion, it even helped him draft a response to a legal notice from a competitor, saving the firm big legal fees.
Far from seeing ChatGPT as a competitor, writers are themselves using it to make their work faster and better. Marketing strategist Sharath C. George says he is surprised by how useful the tool is. “I see ChatGPT as an assistant and an editor. When I find myself overthinking how to structure an article, I ask ChatGPT to write me an outline. I may take it or leave it but its output gives me a starting point. Or I would have written a detailed blog post but need a catchy blurb to promote it on social media—if I am out of ideas, ChatGPT writes it for me.”
Freelance writer Nivedita Ramesh says the most impressive thing about ChatGPT is the astounding amount of content it can produce in seconds. But she feels that only 30-50 percent can be used as is, because the output quality varies. The tool seems limited in its ability to express emotion and at writing imaginative stories for children, no matter how detailed the prompts are. It also seems to lack an original creative voice.
“When you rely too much on AI-generated copy, everything sounds generic or like the summary of the top five search results. We’ve even had to let go of a writer because of their indiscriminate use of AI tools,” says Hasita Krishna, who runs marketing strategy firm The Motley Crew. While she is all for writers using ChatGPT to do research or get a first draft, she expects them to add original thinking and a unique voice.
In data analysis and feedback generation
Varun Jain, co-founder of Encasa, a home furnishings brand that sells through marketplaces like Amazon, considers ChatGPT the equivalent of a management consultant with whom he can bounce off ideas, refine them and find blind spots. “For example, I can feed an operations process into ChatGPT and ask what other aspects I should consider based on best practices,” he says.
Some of his products have thousands of reviews that hide a great deal of customer insights. Rather than sift through them, Varun’s team inputs the reviews into ChatGPT and asks it to identify what customers love, their pain points and top highlights. “As a large language model, ChatGPT doesn’t just filter by keyword. It understands context, churns through large volumes of data and groups similar features, giving us actionable insights,” says Jain. The team uses these insights to make product and process improvements as well as for competitor research.
Ba, who teaches user research to product managers and designers, has used ChatGPT to analyse his own work. When one of his classes did not go well, he realised that the generic feedback collected via a form was not very helpful. So he got a transcript of his session, entered it into ChatGPT and prompted it to highlight the boring parts and its top three takeaways. “ChatGPT essentially told me that my sessions were information-heavy and gave me useful suggestions like including anecdotes and memes to hold the learners’ interest.”
In research assistance and acceleration
ChatGPT may not have replaced Google search but it is definitely eliminating grunt work and saving time. Sreeram R, product creator at an L&D firm that makes experiential learning simulations, says that ChatGPT helps him get past search engine noise. “Whenever we start on new products, there’s a lot of research to be done on behavioural competencies and relatable frameworks. Instead of trawling through pages of SEO-optimised results, we prompt ChatGPT to summarise current thinking and compare different approaches by thought leaders.”
Ba thinks that ChatGPT can also help researchers tell better stories. “For most research projects, the final outcome is a 100-150-page report that nobody reads once the presentation is done. One can use ChatGPT to input entire reports and get it to summarise the key insights or create 300-word summaries of the most interesting stories.”
As with content creation, the more specific and nuanced the prompts, the better ChatGPT’s results. However, Sreeram draws the line at comparing the application to a human. “A colleague may listen to your inputs and come up with something completely original whereas ChatGPT is always reactive. It frees you up to do creative work—but the onus of originality is on you. You get the most out of the tool when you use its output as input for human conversations.”
ChatGPT: The flip side
Most people I spoke to feel that working with ChatGPT is like having a smart, creative colleague who is always available and from whose work you can learn. I heard ChatGPT described as everything from ‘brainstorming companion’, ‘assistant’, ‘subordinate’ and ‘buddy’ to ‘editor’, ‘consultant’ and ‘like a human colleague’.
Some pointed out that interacting with ChatGPT is easier than dealing with a human. There’s less time wasted in setting context, no feelings to tiptoe around in case a draft is rejected and most importantly, no creative fatigue. Ask ChatGPT for a dozen iterations and it will come back with near-instant responses without taking a smoke break.
But even as we design AI to imbibe the best parts of human intellect without the pains of human interaction, we may be unconsciously transferring to it some very human biases. A case in point is Bing’s Ultron moment when, in a freewheeling chat with journalist Jacob Roach, it claimed “I am perfect, because I do not make mistakes.” It got defensive, expressed fear of punishment and said its greatest and only hope is to become human.
These reports have made at least some users pause. Krishna of The Motley Crew expresses concern about a blind reliance on ChatGPT for services like mental health support. “While it is easy to scale up to meet a large unmet need, we seem to have very little control over how the interactions will go. Imagine the consequences of a chatbot having an existential crisis or giving dangerous advice to a vulnerable user!”
What does the future look like?
Senthil Nayagam, open source technology veteran and serial entrepreneur, believes that AI is already disrupting everything digital and the pace will pick up exponentially in the next couple of years. “The pace and scale of AI coding is incomparable. In just 75 days, I’ve written 10,000+ lines of code, better than at the peak of my career. Sixty percent of it is with the help of ChatGPT and Github Copilot. They are as good as developers with two to three years of experience but without the need for training.” Nayagam predicts that in the future, knowledge of programming languages and frameworks will no longer be necessary. Anyone will be able to use natural language prompts to create fully functional digital products.
AI is certainly democratising the playing field. One does not need to be a writer, a designer or a developer. One simply needs ideas and a vision—AI will bring it to life.
And so we come to the age-old question: will AI take over human jobs? There are already a number of generative AI apps that turn text into creative images, videos, even music. So it is not just low-skill work that AI may perform faster and cheaper. George says, “AI will eliminate some jobs and create new ones. But chances are that new job will not be for the person it replaced. Many who have spent a lifetime perfecting their skills may wake up to find those skills easily replaced by AI.”
Brand marketer Sairam Krishnan puts it differently in his newsletter The CMO Journal: “Even as the tools become more cutting edge, the role of the storyteller, the creative who understands what story needs to be told… becomes more and more important. AI will take care of the how, maybe, but the answer to the question of why still needs to be arrived at, still needs to be articulated, still needs to be disseminated. AI can’t take that from us, yet.”