Generative AI is also a revolution for computer interfaces
Every revolution in tech starts with a better way for normal humans to access computing power — from punchcards and command lines to windows and a mouse to smartphones and touchscreens. AI is poised to be the next leap.
Why it matters: One big reason the tech world is so giddy over generative AI is that industry veterans see ChatGPT's conversational fluency as a key that can unlock the next level in this game of endowing everyday users with digital powers.
Think of the hours saved — and, in theory, productivity gained — if you can simply tell your chatbot to clean up your inbox, change your system settings or connect to a printer. The upside is, you won't have to know how to do those things yourself. That's a downside, too: It means individual users may end up with less skill and less direct control.
Driving the news: A number of big players are already using the power of large language models to make chat-style interaction a key way to interact with their software.
Microsoft this week unveiled a Windows Copilot which adds a side pane where users can both summon the AI-powered Bing search engine and change a whole host of settings that, until now, required knowing how to dive into an arcane system of control panels. "I think over the coming years, this will become an expectation for how all software works," CTO Kevin Scott said in a statement. Microsoft is also adopting the same standard for plug-ins that OpenAI is using for ChatGPT, meaning that the chatbot isn't just a user interface, it's also a new platform for developers to create apps — and a new front end for controlling other software. Adobe announced a new generative fill tool for Photoshop that lets people describe what they want to take place in an area, say put a mountain here or remove this object. The company sees big potential in using such descriptive commands to take all manner of actions.
What they're saying: Windows head Panos Panay told Axios that AI is bringing about a "generational shift" in computer interfaces.
"It will feel magical," Panay said in an interview.
Yes, but: It's going to take time for the full power and nuance of AI-as-interface to emerge, Panay said, pointing out that it took some years after the mouse before we got to a scroll wheel.
"It's a migration; this isn’t an overnight change," he said. "You learned to point and then you learned to scroll," he said. Panay described a future in which queries made in the chat-based Copilot can stay there or, as needed, move into apps in the main Windows interface. "We will put it where you want it," he said. Some of the computing work making all this AI-based chat happen will continue to happen in the cloud, but for privacy and other reasons, some work will shift back to the device, he said — noting that there are 200 million PCs out there with enough computing power to run some AI models locally.
Be smart: Using a chatbot to interact with software doesn't always mean you are seeing generative AI in action. Sometimes it's just about using natural language to say, change the settings or invoke a series of commands with a single instruction.
The other side: Magical as it often feels, the chat interface comes with its own drawbacks.
The empty chat window is something of a cipher. It doesn't provide users with much indication of what it can do or how to use it; there's no easy way to poke around and get a sense of its capabilities.
Flashback: Tech companies have talked for years about natural language interfaces. Back in the late 1990s, Clippy used to pop up in Word when it looked like you were writing a letter or resume and might need help.
But Clippy's abilities were constrained by the limits of computers to understand all the different ways that humans describe the actions they want to take. It's only with the arrival of large language models trained on all these ways that computers have reached a point they seem poised to speak our language, rather than humans having to learn to "speak computer." Perhaps as a sign of things to come, Microsoft brought Clippy back as an emoji in Windows 11, following years of exile.
What's next: Expect to see chatbot-like interfaces show up all over the place. Once the tech industry gloms on to a trend, it tends to happen everywhere, for better and worse.
For example, once smartphones ushered in an app-centered world, every company and organization felt obligated for a while to create their own app — even if it was just for a conference you were attending once.