How AI Will Change Frontend and APIs
Lyman dismissed the sci-fi notions that large language models and generative AI are in any way intelligent, calling them “pattern printers” that are just “our own babbling mirrored back to us.” Instead, we should look at their potential to be a new type of user interface, he contended.
“As a replacement for human beings, they fall short,” Lyman wrote. “But as a replacement for, say, a command-line interface? They’re a massive improvement.”
It’s an intriguing idea that led us to wonder: How will generative AI change the frontend and the user interface? Will chatbots take over every website?
AI on the Frontend
AI won’t necessarily lead everything to become a chatbot, but AI is already changing some frontend functions, said Lee Robinson, head of developer relations at frontend platform Vercel. First, Robinson is seeing more organizations implement natural language processing capabilities into site search.
“On one side, we’re seeing more interfaces that have natural language as an entry,” Robinson told The New Stack. “Instead of going in manually clicking a bunch of filters and doing a bunch of manual searching, [users are] able to talk to the app, to talk to the software, in natural language — ‘find me shoes that are green that are size 11 that have a Nike logo.’ I can just type that in and it works. We’re definitely seeing more of that.”
That’s not the only change Vercel is seeing due to AI. The second is to integrate AI into the product in ways that aren’t necessarily apparent to the end user.
“We’re also seeing a rise of companies who use large language models. They use AI under the hood to craft a better product experience, but you might not realize it at first,” Robinson said. “It’s behind the scenes. It’s part of creating a better product experience. And I think this is it’s a rapidly growing part of companies infusing AI into their existing product offering.”
AI and the API
Lyman argued the sweet spot for AI isn’t the output box, but the API. And it’s best use isn’t for intelligence, but as a user interface between humans and computer.
“AI may not truly understand us, but it can deliver our intentions to an API with reasonable accuracy and describe the results in a way we understand,” he stated.
“One area where I see a lot of potential is the simplification of complicated graphical user interfaces,” Asthana wrote. “For complex tasks, graphical user interfaces often become hard to use, and actions hide behind rows of buttons, menus, shortcuts, and procedures. They require years of training for people to become proficient, and even then, most people struggle with them. Generative AI trained on domain understanding has the potential to simplify those experiences.”
These AI bots won’t be limited to chat, but will be deeply embedded in the existing workflows through which humans interact with computers, he added.
“For instance, bots will begin helping with intensive UI and data tasks, interacting through voice, and of course, interacting through a chat mechanism,” he wrote. In this scenario, APIs become the “hands and legs that power the ‘thinking’ that the AI is doing, according to Asthana. However, that may require some changes to how APIs are deployed, he cautioned.
“Until now, we have primarily been designing APIs for applications that are used by humans, but designing APIs for machines will become an increasingly important area,” Asthana wrote. “If you are the leader of an organization, what does this mean for you? Well, if your organization doesn’t have APIs or has poorly designed APIs, you are invisible to these bots.”
It’s a shift Postman is already seeing, Ankit Sobti, co-founder and CTO of the API platform, told The New Stack.
“We are increasingly seeing AI-based consumers,” Sobti said. “We’re using that to either discover information or finding different connection points between the systems; and our opinion is that [with] APIs, you have to be conscious as a designer of details, to be conscious of both human consumers and AI consumers, as you build them out.”
One tangible way APIs must change to support AI is by supporting streaming responses, Robinson said.
“These models can take some time to think up a response and generate the response,” he said. “A lot of APIs need to support being able to stream responses, because it could take 10 seconds, 30 seconds, a minute to actually generate that image or to write the blog post. So the API has to be capable and enabled to do that.”
That will also require pairing with modern platforms that can support streaming for the best user experience, Robinson added.
However, other ways in which APIs need to adapt are still being worked out, Sobti said.
“I see definitely a point where we are in the awareness spectrum of the fact that there are a newer set of consumers that are consuming APIs now — how do the design models change? How do the development practices change? And how do the deployment practices change around it, is something that we believe we have to see and discover as an ecosystem,” he said.