Diffbot, a Silicon Valley company known for its extensive web knowledge index, has recently unveiled a groundbreaking AI model focused on factual accuracy. The new model, based on Meta’s LLama 3.3 and dubbed GraphRAG, marks a significant advancement in the field of artificial intelligence.
Unlike traditional AI models that rely solely on preloaded training data, Diffbot’s LLM leverages real-time information from its Knowledge Graph, a dynamic database containing over a trillion interconnected facts. This approach aims to enhance accuracy and transparency by retrieving information from the web as needed, rather than relying on static training data.
The Knowledge Graph, created in 2016, continuously crawls the web to categorize entities such as people, companies, and products, extracting structured information using computer vision and natural language processing. With regular updates every few days, the Knowledge Graph remains current and reliable for the model to query in real time.
In benchmark tests, Diffbot’s model has shown impressive results, outperforming other popular AI models in accuracy scores. Additionally, the company has made its model fully open-source, allowing organizations to customize and run it on their own hardware, addressing concerns about data privacy and vendor lock-in.
This release comes at a crucial time in AI development, as there is increasing scrutiny on large language models generating false information. Diffbot’s approach, grounded in verifiable facts from its Knowledge Graph, offers a promising alternative to the trend of scaling up model sizes.
Industry experts see the potential value of Diffbot’s Knowledge Graph-based approach for enterprise applications where accuracy and auditability are critical. The model is already being utilized by major companies like Cisco, DuckDuckGo, and Snapchat.
The open-source availability of the model on GitHub and a public demo at diffy.chat make it accessible for testing and deployment. With versions optimized for different hardware configurations, organizations can easily integrate the model into their AI workflows.
Looking ahead, Diffbot envisions a future where AI evolves beyond just larger models to better organization and access to human knowledge. By emphasizing real-time information retrieval and data provenance, the company is paving the way for a more accurate and transparent AI landscape.
As the AI industry grapples with challenges around factual accuracy and transparency, Diffbot’s innovative approach offers a compelling alternative. While the impact of this release on the field remains to be seen, it signals a shift towards prioritizing accuracy and verifiability in AI systems.
