More than an OpenAI Wrapper: Perplexity Pivots to Open Source

Imagine a bright and hopeful illustration, adopting a charming, light-hearted animation style. At the center, portray a whimsical tiny robot assistant representing the innovativeness of AI search engines. The little robot is looking at a vibrant computer screen that displays some code, referencing the open source software that companies like Perplexity are using. Above the screen, a symbol represents Perplexity. The backdrop should be a night setting with a soft ethereal glow coming through a window, symbolizing the 'golden hour' of AI developments. Furthermore, show abstract neural network connections swooping across the scene. This is an illustration for an article about the progressive AI search engine, Perplexity and its use of open-source software, its own search index, and unique search rank mechanism.

Perplexity, an AI search engine, is gaining attention as an alternative to ChatGPT. Unlike ChatGPT, Perplexity provides citations by default, addressing the hallucination issue in generative AI. The company has become a strong player in a market dominated by OpenAI, Microsoft, Google, and Meta. Perplexity has evolved rapidly, developing its own search index and building its own LLMs based on open source models. They have also introduced Copilot search, which delivers pinpoint results by considering user preferences. Perplexity’s default model is based on a fine-tuned version of GPT-3.5 and LLaMA-2. They aim to move away from relying on OpenAI and are considering using LLaMA-2 or Mistral as their base model. The search index size is not as important as the quality of the data, and Perplexity’s search rank mechanism relies on citations from a large language model. Despite competition from Google and Microsoft, Perplexity’s pivot to open source LLM models positions them well in the market.

Full article

Leave a Reply