In a recent interview with CNBC Television, Amazon founder and Prometheus AI co-CEO Jeff Bezos revealed that the startup's artificial intelligence model operates at a fundamentally different level of detail compared to traditional large language models (LLMs).
Bezos explained that Prometheus AI relies on training data that is "completely different" from what typical LLMs use. "We have to create our data sets and access data sets that are very hard to access," he said, emphasizing the proprietary nature of the data sources.
The comments shed light on Prometheus AI's strategy of differentiating itself in a crowded AI landscape by focusing on specialized, hard-to-obtain data rather than broad internet scrapes. Bezos did not provide specifics on the data sources but implied they offer a granular level of detail that could give Prometheus an edge in certain applications.
Prometheus AI is a relatively new venture launched by Bezos alongside other tech veterans, aiming to push beyond the capabilities of current LLMs by training on meticulously curated datasets. The approach reflects a growing trend among AI startups to target niche domains rather than compete directly with general-purpose models like GPT-4 or Gemini.
Bezos's remarks come amid intensifying competition in the AI sector, with companies racing to secure exclusive data and develop models that can outperform existing ones in specialized tasks.