Particle uses Large Language Models (LLMs) in several ways, including the following, to:
Summarize content into bullet point summaries, as well as other summary styles
Identify the topics an article is about
Identify which articles are about the same story and should be clustered together
Answer questions from users
Identify interesting and relevant quotes
Power the audio experience in Listen to the News
Particle uses LLMs from OpenAI, Cohere, and Anthropic, with whom we have agreements such that they do not train on any data we send them. Particle does not train any foundational AI models.
Particle also uses traditional AI models from Google Cloud.
For our Listen to the News feature, Particle uses Elevenlabs for audio.