Log into a 1960s computer and all you’d have to control it would be an interface made up of a blank screen and a flashing text cursor. Back then, people punched in what they wanted the computer to do and the machine would write back the response—no menus, buttons, or flashy animations to be found. Half a century later, the fastest-growing piece of tech somehow looks no different.
Over the last few months, it seems OpenAI’s viral generative AI chatbot, ChatGPT, is all anyone is talking about. More than 100 million people have used it to accomplish all sorts of tasks, like doing their homework and writing computer code. Its influence has been seismic as tech firms scramble to offer a ChatGPT-like virtual assistant in their products.
GitHub added an AI chatbot where programmers can ask technical questions and fix their code. Google and Microsoft will soon let people sift through their search engine from a chat window instead of scrolling a list of links. Similarly, on Instacart, you’ll be able to text a robot to plan your meals and shop for groceries. Wix launched an AI text creator in its editor that allows designers to create website copy without leaving the platform, and Wix Studio now has an AI assistant that helps users design more efficiently.
Yet, I can’t help but notice that this sudden wave of AI tools live inside a UI that would take a designer minutes to put together. If the rest of the supposedly future-forward technologies like the metaverse or mixed reality are so graphic-heavy, why is the AI revolution kicking off from a humble text box?
To start, it gives new users a familiar point of entry into otherwise unfamiliar tech. The simple conversational interface makes it easier for companies to educate people about new, unfamiliar tech like the one powering ChatGPT, says Kyle Li, an assistant professor at the Parsons School of Design, much like how the command line made computers more approachable over half a century ago.
Think about it: When someone logs into ChatGPT, there’s no learning curve. They’re already comfortable with its UI elements, like the text box, and they know how to interact with it. They don’t have to go through an onboarding screen either, unlike when they install an app with its own unique sets of design quirks.
More importantly, this latest breed of natural language chatbots is capable of performing a variety of jobs, and a dialogue interface affords users free rein on how to use it. One user may want to code their personal website inside ChatGPT, while another may be looking to build text-based adventure games.
That’s because generative text tools work by making predictions based on similar examples in their source pool. How do sentences similar to the one you entered end? What does the code for websites similar to the one you requested look like? ChatGPT answers your input by making a prediction based on similar content patterns elsewhere on the web. But your input could be anything.
Meanwhile, graphical user interfaces (GUIs) are deliberate and specific—tap on an icon, scroll on the web, click a button—and their rules vary from app to app. No one GUI can accommodate limitless possibilities like a chatbot.
These seemingly primitive chat windows can be seen as a versatile universal interface for a wide range of tasks at tech companies—inching them closer to manifesting the “everything app,” says Dr. Hamed Zamani, PhD, a computer science professor at the University of Massachusetts Amherst, which tech leaders like Elon Musk have been lusting over for some time now.
“You don't need to have one app for translation, one app for writing assistant, one app for learning a new language,” adds Zamani. “It's all in one interface.”
Silicon Valley’s quest to build text-based conversational interfaces that feel as natural as speaking to a human has been a long time in the making, from one of the earliest chatbots, Eliza, in the 1960s to the virtual assistants on our phones today.
Most recently, in 2016, Meta (then Facebook) famously bet on chatbots to replace apps and envisioned people would soon do everything from a chat interface, like making a restaurant reservation and sending someone flowers. Later that same year, Google released a dedicated messaging app that allowed people to include an AI assistant in their conversations, and ask it questions to, say, plan a trip in a group chat.
“Chatbots will have as profound an impact as previous shifts [GUIs, touchscreens] we’ve had,” said Microsoft CEO Satya Nadella in March 2016.
The reason the conversational interface keeps cropping up every few years is that it’s only the form of computing interaction that speaks the same language as we do, and that immediacy feels “magical” and “humane,” says Matt Webb, a longtime designer, and co-founder of the now-shut pioneering studio, BERG.
For centuries, language has been a tool of communication, and the chat interface, with its powerful simplicity, is an extension of that, transcending platforms and generations, Christiane Paul, curator of digital art at the Whitney Museum of American Art told Shaping Design.
We’re still seeing a lot of sameness, as tech companies look to rip off ChatGPT and acquire an AI “innovation badge,” believes Dan Grover, a product manager who has worked on WeChat and Facebook Messenger chatbot teams.
Over time, though, designers will develop a stronger point of view, he says. As designers acquire more literacy in this developing domain, they’ll move on from simple textboxes and iterate on the actual performance, mechanics, and usability of AI-based features.
For designers today, conversation design and designing the bot’s characteristics—like its voice, tone, and flow of the interaction—will prove vital to building a successful chatbot and ensuring users have productive conversations, says Mike Myer, CEO and founder of the AI platform, Quiq.
“Conversation design is to conversations [and chatbots] what website design is to websites,” Myer added. One simply can’t exist without the other.
