Gemini 3 Approaches The Uber-Software Point: AI As The New UI
Google's Gemini 3 provides generative software interfaces.gettyAI innovators have been telling us that AI is the new UI, or user interface, for over a year. That just became even more obvious with Google’s release of its new Gemini 3, which the company says is its most intelligent model ever. Gemini 3 incorporates a new generative UI mode that will automatically provide unique user interfaces for understanding complex and multimodal answers to questions: a significant leap forward in generative user interfaces.“Gemini 3’s unparalleled multimodal understanding and powerful agentic coding capabilities are also unlocking more bespoke generative user interfaces,” says Elizabeth Hamon Reid, Google’s VP of Engineering for Search. “Now, Gemini 3 in AI Mode can dynamically create the ideal visual layout for responses on the fly — featuring interactive tools and simulations — tailored to your query.”Gemini 3 offers stronger reasoning capabilities and deeper understanding of human intent, Google says. It’s better at finding relevant answers in Google’s massive corpus of search results, and it’s less likely to tell you what you want to hear versus telling you what is most likely to be true.The big innovation, however, is generative user interfaces: custom-built interactive software interfaces to answer our questions. One example Google offers is explaining the three-body problem. To answer, Gemini 3 generates an instant physics demonstration of how three orbiting objects in space generally trend to instability. Another example Google provides is an answer to mortgage payback options, for which Gemini 3 instantly generates an interactive mortgage calculator that allows users to input different timelines, interest rates, down payments, and home prices.Google's Gemini 3 provides a generated software interface to deeply explore the 3-body problemJohn KoetsierAt some level, this is AI approaching the uber-software point: the stage at which a high-end AI model can dynamically emulate or create an optimal user interface for any given set of requirements. All the big LLMs such as ChatGPT and Claude tend to depreciate custom-built user interfaces because we tell them what results we want, rather than click around in an interface to achieve a desired output. But Google’s Gemini 3 is likely now the most capable at inventing useful new UIs on the fly to show answers, boost deeper understanding of processes, and enable rapid adjustments to answers based on changing inputs.This is significant, according to computer-interface expert Jakob Nielsen, who says that AI is only the third user-interface paradigm in computing history.A dynamically-generated software interface for mortgage calculations generated by Gemini 3John Koetsier“With this new UI paradigm, represented by current generative AI, the user tells the computer the desired result but does not specify how this outcome should be accomplished,” Nielsen said in 2023. “Compared to traditional command-based interaction, this paradigm completely reverses the locus of control.”This is reminiscent of how Steve Jobs sold the all-screen change-on-the-fly software-defined user interface of the first iPhone, comparing it to the set-in-stone user interface of prior smartphones that dedicated half their space to a physical keypad and buttons.But now, instead of user interfaces being defined by software engineers once and users simply using what they provided for many months or years, AI engines are instantiating new interfaces instantly, and just as quickly deleting them when they are no longer needed or useful.It’s essentially a revolution in computing: as Nielsen said, only the third major shift in all of computing history. (The first was batch processing; the second command-based interaction.)“Perhaps you’re researching mortgage loans: Gemini 3 in AI Mode can make you a custom-built interactive loan calculator directly in the response so you can compare two different options and see which offers the most long-term savings,” says Google’s Reid.While new computing models don’t necessarily obliterate old models – we still use keyboards even though computers have been able to take dictation for decades – we’ll likely see this kind of generative interface become a significant component of all software in the future. And, theoretically, a sufficiently advanced AI model could become whatever software interface we might need for any given task, whenever needed.That’s likely a long way off, and it may never be the most efficient or effective way to accomplish certain tasks. But Google is continuing to research and refine generative interfaces over time, Reid says.
