Perplexity Unveils Enhanced AI Search with GPT-3.5 Turbo Fine-Tuning and Introduces Code Llama Chat
Perplexity has revealed significant updates to its AI search feature, Copilot, incorporating GPT-3.5 Turbo fine-tuning from OpenAI and the introduction of Code Llama to its LLaMa Chat.
These advancements promise a more efficient and responsive AI-powered search experience while maintaining high-quality results.
For developers, this means gaining access to Meta’s latest open-source large language model (LLM) for coding within just 24 hours of its release.
Enhancing Speed and Accuracy with GPT-3.5 Turbo Fine-Tuning
Perplexity’s testing has demonstrated that the fine-tuned GPT-3.5 Turbo model is on par with the human ranking of the GPT-4-based model.
This enhancement offers Copilot users increased confidence in the accuracy of answers provided by Perplexity.
One notable achievement is the remarkable reduction in model latency by approximately 4-5 times, resulting in nearly an 80% decrease in the time taken to deliver search results.
This swifter response time could significantly elevate user experience, particularly for individuals seeking rapid solutions to critical inquiries.
Transitioning to the fine-tuned GPT-3.5 model is also anticipated to lower inference costs, enabling Perplexity to invest in further improvements to continuously deliver new features and enhanced performance.
Moreover, Perplexity users now have the capability to upload PDF files, fine-tune search outcomes to specific sources, and create an AI profile to personalize search results.
Similar in concept to the initial part of ChatGPT’s Custom Instructions, the AI profile is a tailored approach to refining the search experience.
Logged-in users can test Copilot for free up to five times per hour. Alternatively, Perplexity Pro subscribers, at a cost of $20 per month, can enjoy up to 300 Copilot queries daily and have the option to switch to GPT-4.
Introducing Code Llama Intstruct to Perplexity Labs LLaMa Chat
In conjunction with the enhancements to Perplexity’s AI-driven search, the company has introduced Code Llama within Perplexity Labs LLaMa Chat.
Code Llama is a coding-focused LLM developed by Meta AI and is seamlessly integrated into Perplexity’s LLaMa Chat to enhance responses to technical queries.
Users can tap into this open-source LLM and request assistance in writing functions or clarifying segments of code. Code Llama’s Instruct model has been meticulously tuned to comprehend natural language prompts.
The swift integration of Code Llama into Perplexity Labs LLaMa Chat ensures that developers can promptly assess its utility following Meta’s announcement of its availability.
A Glimpse into the Future of AI Collaboration
The latest strides in Perplexity’s AI-powered search through the fine-tuned GPT-3.5 Turbo model, coupled with the introduction of Code Llama chat, underline the remarkable progress achieved when major tech companies engage in collaborative efforts.