Photo credit: venturebeat.com
2024 has emerged as a pivotal year for Perplexity. This innovative AI search startup, established by Aravind Srinivas, formerly of DeepMind and OpenAI, has secured substantial funding, with its latest investment round valuing the company at approximately $9 billion. Additionally, it has rolled out several key features such as Pages, Spaces, and enhanced shopping functionalities.
These advancements have strengthened Perplexity’s identity as an “AI-first” knowledge discovery platform, distinguishing it from traditional search engines like Google and Bing, which are integrating AI into their already established frameworks.
Despite these achievements, Perplexity faces a challenging landscape ahead.
To stay competitive, the company is expanding its offerings by incorporating Carbon into its ecosystem. Recently, Perplexity acquired this startup for an undisclosed amount, aiming to bridge the “data gap” experienced by enterprises utilizing AI search technology and streamline their knowledge discovery processes.
Carbon has engineered an advanced retrieval framework that facilitates the connection of external data sources to large language models (LLMs). Users can utilize the Carbon universal API or SDKs to integrate their data sources effortlessly, thereby accessing information stored in various formats, including text, audio, and video files. The framework is compatible with over 20 data connectors.
The Expanding Scope of AI Search
In today’s digital landscape, AI-driven search tools are utilized by a broad spectrum of users, spanning individuals to corporate teams. The core advantage of this technology is its ability to deliver immediate answers to queries, eliminating the need for users to wade through numerous links and content.
Perplexity has thrived by leveraging diverse large language models to source information from the web, simplifying the data retrieval process for users. The platform also enables teams to extract insights from various personal and organizational documents, such as PDFs and Word files.
Nonetheless, users often face challenges when trying to upload internal documents—such as reports, conversations, and images—particularly when handling vast amounts of proprietary data. This bottleneck can lead to a lack of context in the answers, which might compromise their relevance to specific organizational needs.
Addressing this “data gap,” Sanjeev Mohan, a former VP at Gartner Research, indicated that a significant trend in AI for 2025 will revolve around the extraction and transformation of unstructured data. This trend aims to empower teams to harness information from scattered internal sources efficiently, which in turn will enhance the capabilities of their LLMs to generate precise and actionable responses.
The integration of Carbon’s retrieval framework is expected to be instrumental for Perplexity. The company aims to incorporate Carbon’s engine and connectors within its platform, enabling users to easily connect various data sources, such as Google Docs, Notion, Hubspot, and Slack.
This development is projected to dramatically broaden the knowledge base that fuels the AI search engine, ensuring that responses are not only more relevant but also tailored to individual user needs.
What Users Can Anticipate from Carbon-Powered Perplexity
As Perplexity welcomes Carbon into its fold, the full potential of this integration has yet to be realized. However, it’s clear that the additional data connectors stand to greatly enhance the operational efficiency of teams utilizing the AI search engine.
For example, if a team needs to adjust a product launch date, Perplexity would have the capability to sift through all relevant documents in Google Docs, Notion, and Slack to ascertain the latest deadlines and guidelines that have been established.
This functionality will alleviate the need for users to manually gather context from a variety of platforms and documents, providing coherent answers autonomously.
“The significant advantage here is that our technology can identify the necessary information without requiring users to specify exactly where that data resides,” noted Sara Platnick, who oversees communications at Perplexity.
Another practical application would involve summarizing insights from customer meetings. Perplexity can quickly extract pertinent information from connected customer relationship management (CRM) systems.
By employing Carbon’s retrieval-augmented generation (RAG) workflows, Perplexity aims to simplify enterprise search, allowing businesses to forego the complexities of constructing their RAG infrastructures independently.
“By efficiently locating and interpreting proprietary data through the integration of Perplexity and Carbon, companies can tackle a multitude of intricate generative AI initiatives,” commented Kevin Petrie, VP of research at BARC US. “Leading organizations are particularly interested in applications related to customer service, document processing, image processing, and recommendation systems.”
Execution Will Be Key
The acquisition of Carbon marks just the initial step for Perplexity. The critical factor moving forward will be the effective execution of the integration, especially given the sensitive nature of the proprietary data that enterprises manage.
“Organizations are understandably cautious about sharing their intellectual property with external platforms. Consequently, both Perplexity and Carbon need to implement governance measures that ensure the protection of data within the organization’s secure environment. Companies are not eager to expose their private information or to inadvertently train public models on their proprietary knowledge,” Petrie added.
In response to these concerns, Platnick emphasized that all information sourced from internal and private channels will be encrypted throughout, as well as any data handled via Carbon’s connections. The company also has additional safeguards in place to ensure that confidential documents remain inaccessible to unauthorized users.
Currently, there is no confirmed timeline for the complete integration of Carbon with Perplexity. However, the managed API operated by Carbon will be discontinued on March 31, 2025, with existing clients duly notified and supported through their transition.
Source
venturebeat.com