We will always need interfaces
ChatGPT will eat up a lot of the software we use every day. I’m not particularly interested in using Uber Eats if I can just ask a bot to order lunch. Not having to wade through dozens of options on Google Flights or Expedia will be a welcome relief. I’d even give the bot access to my email if it can use that to automatically schedule meetings for me. The future is here!
But while a lot of tasks are well-suited for the chat bot interface, many others are not. I know because I’ve tried. Back in 2016, we built a company called Yala, which helped businesses share posts to social media through a Slack interface. In fact, we grew to serve 3,000+ businesses, including Walmart, IKEA and Stack Overflow.
They did not like our bot. It was clunky. They wanted a calendar view, where they could see the content going out each day. They wanted to be able to see mockups of their posts. They wanted to use Yala in the browser.
Many application simply do not lend themselves to the chat interface.
Video (TikTok, Youtube)
Social networking (Twitter, WhatsApp)
Productivity (Monday, Airtable)
Games
These are immediate examples that come to mind, but there are many others. These apps can’t live inside a chat interface - they’d be crippled, no matter how powerful the underlying AI is.
A large number of applications that cannot be chat-ified will still be GPT-powered, either on the frontend as part of the GUI (for example, the new Github Copilot X) or on the backend. In fact, it stands to reason that many backends will be entirely replaced with GPT logic (see this library that scrapes the web using GPT as an example).
But we’re all still going to need software, if only for the GUI.