The New ChatGPT Is More Show Than Tell

OpenAI unveiled an "Intelligent UI" update for ChatGPT on Wednesday that generates visual, custom elements for answers, powered by its recent GPT-6 model, which rolls out to paid users today and free users tomorrow. OpenAI product manager Aarush Selvan said a design team spent significant time determining when adding a diagram, chart, or buttons adds value versus when it feels cluttered, and users can ask ChatGPT to generate fewer visuals. The update follows Google's earlier "generative UI" expansion for Search and signals a broader shift toward generative user interfaces for ChatGPT's more than a billion users.

When you ask a chatbot a question, you’ll often get a wall of text back, maybe with an image or two pulled from the web. That may no longer be the case soon as OpenAI https://www.wired.com/tag/openai/ overhauls how ChatGPT https://www.wired.com/tag/chatgpt/ answers prompts. On Wednesday, OpenAI unveiled an “Intelligent UI” update for ChatGPT that generates visual, custom elements—when helpful—for answering a user's questions. The company's new experience is powered by its recent GPT-6 model https://www.wired.com/story/openai-says-gpt-6-can-use-a-computer-better-than-a-human/ , which rolls out to paid users today and free users tomorrow. I got a quick test of the user interface before launch, and these visual elements felt like a welcome change. When I asked for a diagram of a slug, ChatGPT generated an anatomy-focused visual with buttons I could press for each different part, from the upper tentacles to the muscular foot. I also generated a San Francisco https://www.wired.com/tag/san-francisco/ apartment affordability calculator with movable sliders to adjust income and other expenses, showing what percentage of my take-home pay different units would cost. My favorite part of this quick hands-on with ChatGPT’s Intelligent UI was an airplane https://www.wired.com/tag/airplanes/ seat explorer. I’m not a fan of sitting in the exit rows, so I asked ChatGPT to generate a custom tool to compare the cabin layouts between four different Alaska and Delta planes. In its output, I could tap any seat on the map for more details, and ChatGPT highlighted the exit rows in green. While useful in some results, every ChatGPT answer won’t suddenly include a chart or graphic. “We have a design team that spent a lot of time figuring out when adding a diagram, chart, or buttons adds value versus when it's starting to feel cluttered,” says Aarush Selvan, a product manager at OpenAI. If a user isn’t interested in visuals, they can tell ChatGPT to generate fewer of those outputs. These visual results are similar to what Google https://www.wired.com/tag/google/ announced earlier this year for Search https://www.wired.com/tag/search/ when it expanded the bespoke results https://www.wired.com/story/google-search-goes-agentic-and-doesnt-need-you-anymore/ users may see. Google’s version is called “ generative UI https://research.google/blog/generative-ui-a-rich-custom-visual-interactive-user-experience-for-any-prompt/ .” One example: if you want to understand deep space better, Search can output an adjustable graphic that breaks down how black holes work using multiple sliders. Chatbot experiences have increasingly become multimodal https://www.wired.com/story/google-lens-multimodal-search/ , meaning that text, images, and voice blend into one user experience. OpenAI's launch of an updated ChatGPT that's easier on the eyes and more interactive for over a billion users signals that the multimodal era is here. However, a larger shift toward generative user interfaces https://www.wired.com/story/these-companies-have-a-plan-to-kill-apps/ may spark an identity crisis for designers, as AI dictates more of what software looks like for individual users. With AI tools now able to quickly generate custom software https://www.wired.com/tag/software/ , the future of chatbots https://www.wired.com/tag/chatbots/ is further personalized and attuned to what I want to see, and what I want to see is slug diagrams. Good riddance to walls of text.