Patsnap, located in Singapore, launched a global patent search database in 2007 to enable IP and R&D professionals to work faster. The debut of its AI assistant, CoPilot, expands its AI product line.
CoPilot speeds up patent and non-patent literature searches, which Patsnap claims will speed up IP and R&D workflows throughout its product portfolio. Its proprietary LLM provides references, journal links, and patent links.
The business claims to have spent millions and has over 50 AI engineers. Patsnap CEO and co-founder Jeffrey Tiong tells Eltrys that Patsnap removes friction in the innovation process for its clients, both inside and across IP and R&D teams. IP analysts and attorneys must run prior art and freedom to operate searches to decide where to invest time and money, and Patsnap’s products make the process easier.
Over 180 million patents and 130 million articles from 170 nations are in Patsnap Analytics. Tiong said Patsnap’s AI solutions can help IP teams study markets and defend ideas at scale.
CoPilot enhances Patsnap’s AI solutions. It helps IP and R&D teams locate information faster in patents, non-patent literature, and technical news. Tiong suggests acquiring automatic patent claim summaries, learning about a technology, and linking to a company’s patents. It solves certain issues too. CoPilot can translate patents and discover relevant material for R&D professionals interested in battery energy density.
CoPilot also helps R&D and IP teams keep up with rapidly changing sectors, provide content analysis to guide strategic patents and research, extract key details from patents and literature, and secure AI without customer data. Tiong claims that CoPilot’s LLM prevents client data from leaving Patsnap’s firewall and entering external networks.
Tiong says Patsnap’s LLM was educated on patents, academic publications, technical reports, and corporate news, including mergers and acquisitions. It also includes IP expert and Patsnap product annotations.
Pre-train, post-train, and self-training fine-tuning teach its LLM general, specialized, and aligned data, giving it patent and non-patent data accuracy. Because of retrieval-augmented generation, Tiong says CoPilot’s model is better at in-depth analysis and less likely to see things that aren’t there than GPT3.5.
SoftBank and Tencent invested $350 million in Patsnap. More than 1,200 personnel serve 12,000 consumers. It serves life sciences, automotive, consumer products, technology, manufacturing, engineering, and legal customers.
