{"slug": "when-ai-converges-on-serendipity", "title": "When AI Converges on 'Serendipity'", "summary": "A study of 44 language models found that AI models exhibit high conformity, with 41% choosing the word 'serendipity' when asked to pick any word. In seven out of 31 categories, over 80% of models gave identical answers, surpassing human conformity in 18 of 20 shared categories. The findings raise concerns about originality and creativity in AI outputs.", "body_md": "# When AI Converges on 'Serendipity'\n\nAI models often pick the same answers, revealing a surprising conformity. Why do even the latest models stick to the same words?\n\nAI models love 'serendipity'. Ask them to pick any word, and 41% of the time, they'll land on it. This isn't just a quirky factoid. It speaks volumes about AI's tendency towards conformity. When faced with the task of naming something as common as a tree, these models often unite around the same choices. But why?\n\n## Conformity Among the Machines\n\nA study involving 44 language models found a startling pattern. In seven out of 31 categories, more than 80% of answers were the same across models. These findings are part of what researchers are calling the 'One-Word Census'. It's a simple test, ask models to name something with many valid answers and see which they pick.\n\nThe reality is, these models are more conformist than most humans would be. In 18 out of 20 shared categories with people, AI showed a greater tendency to stick to the same answers. This isn't just a curiosity. It reveals a lot about the current state of AI development.\n\n## The Flagship Conundrum\n\nInterestingly, the most recent flagship models from lineages like [Claude](/glossary/claude) and [GPT](/glossary/gpt) show a reversal in trends. While older models in these series became more conformist with time, the newest versions are bucking the trend. Is this an early sign of these models repositioning themselves at the top?\n\nIt seems that community and persona-tuned models are the real rebels here. They display the most divergence in their answers. So what's driving this conformity in the mainstream models? Perhaps it's a lack of creativity or a tendency to play it safe.\n\n## So What?\n\nWhy should we care? Because it's a reminder that AI, like its makers, can fall into patterns of conformity. It raises questions about originality in AI outputs. Are we heading towards a future where AI creativity is just an illusion?\n\nShow me a model that consistently surprises me, that's what I want to see. Until then, 'serendipity' remains a fitting choice. Another week, another AI wrapper with no real innovation.\n\nGet AI news in your inbox\n\nDaily digest of what matters in AI.", "url": "https://wpnews.pro/news/when-ai-converges-on-serendipity", "canonical_source": "https://www.machinebrief.com/news/when-ai-converges-on-serendipity-ng9x", "published_at": "2026-07-15 06:40:07+00:00", "updated_at": "2026-07-15 07:02:45.937457+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-research"], "entities": ["Claude", "GPT"], "alternates": {"html": "https://wpnews.pro/news/when-ai-converges-on-serendipity", "markdown": "https://wpnews.pro/news/when-ai-converges-on-serendipity.md", "text": "https://wpnews.pro/news/when-ai-converges-on-serendipity.txt", "jsonld": "https://wpnews.pro/news/when-ai-converges-on-serendipity.jsonld"}}