TL;DR:
- Writer introduces Knowledge Graph, a tool to address the issue of hallucinated content in LLMs.
- Knowledge Graph acts as a content validator by leveraging an organization’s existing source content.
- It automatically identifies content requiring fact-checking and suggests replacement statements based on relevant information.
- Writer provides transparency by displaying the source of the replacement content for human evaluation.
- Knowledge Graph is a structured representation of information stored in a graph format, allowing fast retrieval of specific details.
- It can connect to an organization’s content sources, or content can be manually transferred into the graph.
- Companies have the option to self-host the Knowledge Graph or let Writer manage it.
- Knowledge Graph and self-hosting capabilities are now available in open beta.
- Writer aims to reduce the impact of hallucinated content and provide businesses with greater control and accuracy in AI-generated content.
Main AI News:
As businesses delve deeper into the realm of generative AI, one of the most perplexing challenges they face is the issue of hallucinated content. When confronted with queries for which it lacks the answer, the AI model often fabricates information, regardless of its validity. Finding a solution to address this problem is crucial for businesses to operate effectively.
Today, Writer, the renowned generative AI writing platform, has announced the public beta release of Knowledge Graph, a powerful tool that empowers companies to aggregate their existing source content into a comprehensive graph database. According to May Habib, the CEO of Writer, Knowledge Graph serves as a content validator by leveraging the customer’s own content to scrutinize the generated output.
“The beauty of an LLM is that when it possesses factual information relevant to the task at hand, it incorporates that knowledge,” explains Habib. “Knowledge Graph enables easy accessibility to the information that matters—databases, knowledge bases, shared drives, chats—embedding them for the LLM’s usage in meeting our customers’ specific use cases.“
Knowledge Graph automatically identifies content requiring fact-checking and proposes alternative statements based on the most pertinent information contained within the graph. Writer diligently presents the source of the replacement material, allowing human employees to evaluate and accept or reject the suggestion.
The company defines Knowledge Graph as “a structured representation of information stored in graph format, facilitating rapid retrieval of specific details.” It can directly connect to an organization’s content sources, or users can manually transfer content by dragging and dropping it into the intuitive Knowledge Graph interface.
Importantly, Knowledge Graph operates independently from the content systems that contribute to the graph, ensuring a clear demarcation. Furthermore, companies have the flexibility to choose whether they want to host the Knowledge Graph themselves or rely on Writer for management.
While not foolproof, Knowledge Graph represents a significant step towards assisting companies in mitigating the impact of this well-known problem. Customers have the option to determine the frequency of Knowledge Graph updates, with weekly updates being the typical choice. However, configurations can be adjusted to align with specific requirements, keeping in mind that longer intervals between updates may result in a larger amount of content being excluded from the accuracy check.
In addition to Knowledge Graph, Writer has also announced the availability of self-hosting capabilities for the Writer LLM. This empowers large organizations seeking greater control over their internal content to host the LLM themselves. The Knowledge Graph seamlessly integrates with self-hosted LLM instances.
Both the self-hosting feature and Knowledge Graph are now available in open beta, following a closed beta phase. Writer, established in 2020, has raised $26 million in funding to date, as reported by Crunchbase.
Conlcusion:
The introduction of Writer’s Knowledge Graph to mitigate hallucinated content in LLMs carries significant implications for the market. By addressing the challenge of inaccurate generated content, businesses can enhance the reliability and credibility of their AI-driven outputs. The availability of a content validator and the option for self-hosting empower companies with greater control and accuracy over the information generated by LLMs.
This development reflects a growing demand for AI tools that ensure the integrity of generated content, positioning Writer as a key player in meeting market needs. As organizations strive for more trustworthy and accurate AI-generated content, the adoption of solutions like Knowledge Graph will likely become essential, driving the market towards higher standards of reliability and enabling businesses to leverage the potential of AI with confidence.
