tech
LLMs are stuck in a groupthink groove. This startup is trying to get them out.
Chatbots are far more predictable in their responses than you might expect. That's fine for research or coding, but it's a problem if you're looking for something new.
TL;DR
- Mainstream LLMs like ChatGPT and Claude often give similar, predictable answers to the same prompts.
- A research paper titled 'Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)' highlighted this issue, finding significant repetition across different LLMs.
- Researchers speculate that similar training data and methods lead to this homogeneity.
- Springboards has created an LLM called Flint, built on Alibaba's Qwen 3, designed to generate more diverse and unexpected responses.
- Flint is trained to identify points in its output where more variety is possible and insert random elements, rather than simply increasing overall 'temperature' settings which can lead to incoherence.
- While Flint is still a prototype and not always perfect, it is seen as a valuable tool for creative professionals in fields like advertising and marketing to brainstorm ideas.
- The company aims to provide users with more choice and encourage broader thinking, pushing back against a potentially 'gray, boring world' dominated by repetitive AI output.