Perplexity Unveils LLMs With Real-Time Data

Perplexity, a San Francisco-based startup, has revolutionized the field of language models with the introduction of their online large language models (LLMs). These groundbreaking models, known as pplx-7b-online and pplx-70b-online, are designed to overcome the limitations of traditional LLMs. By tapping into real-time information from the internet, Perplexity’s online LLMs provide users with the most up-to-date and accurate responses to their queries. Let’s delve deeper into the features and benefits of these cutting-edge language models.

What Are Online LLMs?

Perplexity’s online LLMs are a breakthrough in the world of natural language processing. Unlike their offline counterparts, such as GPT-3.5, which rely solely on their training data, online LLMs have the ability to access the latest information available on the internet. This real-time grounding enables the models to generate responses that depend on recent events or data. Whether you need the latest sports scores, stock prices, or Google news developments, Perplexity’s online LLMs have got you covered.

To ensure the utmost accuracy and reliability, Perplexity employs various techniques to minimize the generation of false information. The pplx-7b-online and pplx-70b-online models are built on top of the open-sourced mistral-7B and llama2-70B models. Perplexity has fine-tuned these models using diverse, high-quality datasets, optimizing them for helpfulness and factuality.

Leveraging In-House Search Technology

One of the key features that sets Perplexity’s online LLMs apart is their in-house search technology. The company has developed a robust search, indexing, and crawling infrastructure that allows the models to augment their responses with the most relevant and valuable information available on the web. With a large and regularly updated search index, Perplexity’s online LLMs prioritize high-quality, non-SEOed sites using sophisticated ranking algorithms. This ensures that users receive responses grounded in accurate and trustworthy information.

Advantages Over GPT 3.5

Early testing conducted by Perplexity indicates that their online LLMs match or exceed the capabilities of leading proprietary models like GPT-3.5. When benchmarked on measures of robustness, helpfulness, and knowledge across academic subjects, Perplexity’s online LLMs consistently perform at a high level. The ability to tap into the latest online information gives these models a competitive edge, allowing them to provide timely and accurate facts and data in response to user queries.

Accessing Perplexity’s Online LLMs

Perplexity’s online LLMs are publicly accessible through their API and Labs web interface. This means that developers can integrate this cutting-edge technology into their own applications and websites. By making these highly capable models affordable and easily accessible, Perplexity is driving the democratization of AI. They aim to level the playing field between large tech firms and smaller organizations, enabling everyone to benefit from the power of AI.

With ongoing improvements and advancements in their performance, Perplexity envisions a future where search and information discovery are centered around conversational interfaces. The online LLMs developed by Perplexity provide a glimpse into this future, where users can query an AI assistant much like consulting a human expert. The responses provided by these models will be timely, factual, and nuanced, revolutionizing the way we interact with AI.

See first source: Search Engine Journal

FAQ

1. What are online Large Language Models (LLMs), and how do they differ from traditional LLMs like GPT-3.5?

Online LLMs, such as Perplexity’s pplx-7b-online and pplx-70b-online, are a new generation of language models that have the capability to access real-time information from the internet. Unlike traditional offline LLMs, which rely solely on their training data, online LLMs can provide responses based on the latest information available on the web. This real-time grounding allows them to offer more up-to-date and accurate responses.

2. How does Perplexity ensure the accuracy and reliability of their online LLMs when accessing the internet for information?

Perplexity employs various techniques to minimize the generation of false information. The pplx-7b-online and pplx-70b-online models are built on top of open-sourced models (mistral-7B and llama2-70B) and have been fine-tuned using diverse, high-quality datasets. This fine-tuning process optimizes the models for helpfulness and factuality, ensuring that the responses they generate are accurate and reliable.

3. What is the significance of Perplexity’s in-house search technology in their online LLMs?

Perplexity’s online LLMs leverage their in-house search technology, including search, indexing, and crawling infrastructure. This technology allows the models to enhance their responses with the most relevant and valuable information available on the web. With a large and regularly updated search index, these models prioritize high-quality, non-SEOed sites using advanced ranking algorithms. This ensures that users receive responses grounded in accurate and trustworthy information.

4. How do Perplexity’s online LLMs compare to other leading language models like GPT 3.5 in terms of performance and capabilities?

Early testing conducted by Perplexity suggests that their online LLMs match or surpass the capabilities of leading proprietary models like GPT-3.5. These models consistently perform well in measures of robustness, helpfulness, and knowledge across academic subjects. The ability to access the latest online information gives them a competitive edge, enabling them to provide timely and accurate facts and data in response to user queries.

5. How can developers access Perplexity’s online LLMs, and what is their vision for the future of AI-powered search and information discovery?

Developers can access Perplexity’s online LLMs through their API and Labs web interface, making it possible to integrate this advanced technology into their own applications and websites. Perplexity’s vision for the future of AI-powered search and information discovery centers around conversational interfaces. They foresee a future where users can query AI assistants in a manner similar to consulting a human expert, receiving timely, factual, and nuanced responses. This vision aims to revolutionize the way we interact with AI and democratize its power, leveling the playing field for organizations of all sizes.

Featured Image Credit: Photo by Bernd Dittrich; Unsplash – Thank you!