Singapore-based PatSnap, founded in 2007, has launched CoPilot, an AI assistant to streamline intellectual property (IP) and research and development (R&D) workflows.
PatSnap initially started with a global patent search database and has expanded its suite of AI products to include CoPilot, which enables faster searches within patents, non-patent literature, and technical news.
CoPilot offers automatic summaries of patent claims, helps understand technologies, provides links to company patents, and addresses specific R&D queries. The AI assistant is designed to enhance efficiency for IP analysts and attorneys.
CoPilot’s Role in PatSnap’s AI Product Suite:
CoPilot is integrated into PatSnap’s suite of AI products, supporting IP and R&D teams in various ways.
It assists in staying updated with rapidly changing sectors, offers content analysis for strategic patents and research guidance, extracts key details from specific patents and literature, and ensures AI security by keeping customer data within PatSnap’s firewall.
The proprietary language model (LLM) behind CoPilot has been trained on a diverse dataset, including patents, academic papers, technical reports, and recent company news.
Key Features and Applications:
- Automatic Summaries: CoPilot provides automatic summaries of patent claims.
- Technology Learning: Users can learn about a technology and access links to a company’s patents.
- Problem-Solving: The AI assistant addresses specific R&D problems, such as finding information on improving battery energy density.
- Sector Updates: CoPilot helps teams stay informed about rapidly evolving sectors.
- Content Analysis: It offers content analysis for guiding strategic patents and research decisions.
- Key Detail Extraction: The AI assistant extracts details from specific patents and literature.
AI Security and Model Capabilities:
CoPilot’s proprietary LLM, trained on pre-training, post-training, and self-training fine-tuning stages, ensures specialized accuracy in patent and non-patent data.
The model, developed in-house, eliminates the need for customer data to leave PatSnap’s firewall, enhancing AI security.
CoPilot’s model is better suited for in-depth analysis than GPT-3.5, with a reduced likelihood of hallucinating answers.
Company Background and Funding:
PatSnap, founded in 2007, initially focused on a global patent search database. The company has raised $350 million from investors like SoftBank and Tencent.
With over 1,200 employees and over 12,000 customers across various industries, including life sciences, automotive, consumer goods, technology, manufacturing, engineering, and legal, PatSnap aims to remove friction in the innovation process for its customers.