New AI Breaks Free from Chatbot Groupthink Problem

Ask ChatGPT to pick a random number between one and ten, and you'll almost always get seven. Ask Claude the same question, and guess what? Seven again.

Australian startup Springboards noticed this weird problem and decided to fix it. Their new AI model called Flint generates far more variety in its answers than mainstream chatbots, offering fresh ideas when others get stuck in repetitive loops.

The groupthink issue goes deeper than party tricks with numbers. When CEO Pip Bingemann asked major chatbots to name a type of car, they predictably said Toyota or Honda. Flint suggested a Ford F-150. When prompted for running shoe taglines, both ChatGPT and Claude responded with identical words: "Run your way." Flint came back with something completely different.

Recent research confirms this isn't just coincidence. A paper called "Artificial Hivemind" won best paper at NeurIPS, a major AI conference, after exposing shocking repetition across 25 different language models. When researchers asked these models to write metaphors about time, most defaulted to "Time is a river" or "Time is a weaver." Real humans gave six completely different answers when asked the same question.

The sameness shows up everywhere once you notice it. Ask chatbots to name your band, and you'll likely see suggestions involving "glass," "neon," "velvet," or "static." One test of ChatGPT's "original" band name suggestions found that Sofa Astronauts already existed as an actual band.

Scientists believe this happens because most AI models train on similar data using similar methods. The result? Everyone gets the same answers to open-ended questions, even though they feel like personal conversations.

The Bright Side

Springboards built Flint specifically for creative professionals who need actual variety when brainstorming. Zoe Scaman, a business strategist, tested it against major chatbots using a classic challenge: reinvent a finance company for young people. The mainstream models all suggested teaching financial literacy in fun ways. Flint took a completely different approach, questioning the entire concept of wealth accumulation.

The startup now offers a tool where marketers and advertisers can drag around text from different AI models, mixing and matching to spark genuine creativity. Flint serves as the wild card option when other models start sounding too similar.

This breakthrough matters beyond advertising campaigns. As AI becomes central to how we work and think, having models that can suggest truly different paths helps us avoid collective blind spots.