Carl Rost is the mind behind the AI-powered patent search tools at Patsnap.
Patsnap stands on the forefront of innovation intelligence, harnessing the ability of AI and machine learning to sift through billions of datasets, enabling innovators to make crucial connections. Their cutting-edge LLM technology, tailored for R&D and IP professionals, effortlessly navigates through billions of pages of patents every day. Patsnap’s AI assistant engages in conversational responses to novelty questions and may pinpoint specific answers inside extensive texts. As an illustration, it will possibly accurately determine whether a specific widget type is already patented.
Are you able to provide an summary of how Patsnap’s AI assistant works and its primary functions?
Sure! It’s an AI assistant called Hiro that lets you ask questions on a selected patent or perhaps a result set or our entire database! It’s been trained to grasp innovation and patent related questions and respond in a way that satisfies technical subject material experts and IP professionals. A recent advancement is that Hiro may even allow you to solve technical problems and propose novel directions for brand new inventions by applying inventive principles to technical solutions and problems which were present in our patent and literature database. Hiro works a bit in a different way depending in case you use it in our products which can be for R&D or for IP professionals.
I feel what makes Hiro unique is that it’s powered by Patsnap’s proprietary LLM, answers also link references and sources from Patsnap’s library of 200 million patents, 190 million pieces of literature, 254 million chemical structures, 879 million biological sequences, and a couple of billion news articles.
What problems is that this application solving for enterprises?
Great innovators should spend their time innovating, not determining novelty of products or doing preliminary research of the market. Patent data is one in every of our richest sources of technical information, rivaling journal data, especially in certain technology fields. For R&D, the time it takes to seek out and interrogate one of these data has been an enormous blocker to leverage this, but tools like Hiro can truly democratize one of these information for the primary time.
For legal professionals, it is common to spend hours, days, weeks, running prior art and freedom to operate searches. With AI tools this might be done more quickly, and with more accuracy, freeing up bandwidth for more strategic work.
Existing AI tools are one in every of two things: overly generalized and due to this fact not appropriate for the mental property space, or they’re black boxes, with no transparency as to resources, reducing confidence and obstructing decision-making. With Hiro, we link back to sources and ensure full visibility in any respect stages of the event process.
What were the primary challenges your team faced while developing the AI features for Patsnap, and the way did you overcome them?
We all know that individuals constructing latest inventions wish to keep them protected, so security was top of mind when constructing Hiro. Because the model powering Hiro is local and built into our app, no data leaves the environment to 3rd parties which can be hard to trust. Our competitors didn’t do the groundwork and bolted on third party models that don’t rise up to scrutiny. Once we say that we aren’t training models on customer data, we all know that to be true and may show our customers that and what we do as an alternative. In contrast, our competitors’ solutions expose you to risk through third parties who’ve a lower than stellar popularity, by way of transparency and handling of information.
Could you elaborate on how Hiro answers specific novelty questions and the impact this has on R&D and IP workflows?
With Hiro, users can ask questions like “What facets of this invention make it novel?” or “How might this patent delay in several legal systems?” and even “how you can construct a wearable jetpack” and get answers that talk to every step of the invention process. In comparison with generalist models, Hiro really gets what makes a patent special. Users don’t have to be patent experts to unravel what’s or isn’t novel inside their invention, and may understand in seconds which a part of their product or tool must be protected.
How does Hiro handle the vast amount of information from patents and non-patent literature to supply precise and relevant answers?
We did extensive training on that dataset, and rated the responses with experts. We then trained AI on the expert responses, had the AI rate output, and had experts review that. All in all, we’ve rated thousands and thousands of information points this solution to make sure the responses are meaningful for tech experts and patent pros.
How does Hiro utilize large language models (LLMs) to reinforce the efficiency of patent searches and IP evaluation? What kinds of data were used to coach Patsnap’s proprietary LLM, and the way do you ensure its accuracy and reliability?
Patsnap built an industry-specific LLM to power Hiro. The LLM has been trained on patent records, academic papers, and other innovation data, which helps it understand and retell info in a way that’s more helpful to professionals than generalist models. To make sure accuracy and reliability, we employed rigorous data preprocessing methods, including filtering out lowquality data, deduplication, and rewriting. We also synthesized latest data by combining different sources to reinforce the model’s understanding of IP-specific nuances. We supervised finetuning and reinforcement learning from human feedback to repeatedly improve its performance.
PatsnapGPT has been tested extensively and has outperformed GPT-4 in IP-specific tasks, demonstrating superior capabilities in drafting, classifying, summarizing, and reasoning inside the patent domain.
The proprietary LLM is transparent, linking sources and references, and it’s not trained on customer data. It’s the one industry player using an in-house tuned LLM, in an industry that is particularly reliant on data privacy and confidentiality.
How does Patsnap’s proprietary LLM compare to other general-purpose LLMs like GPT-4 by way of performance and accuracy for IP-related tasks?
Patsnap’s proprietary LLM outperforms GPT-4 in relation to mental property queries. Using the USPTO Patent Bar Exam, PatsnapGPT-1.0’s performed at the extent of an IP expert, while general LLMs didn’t reach the cutoff for patent lawyers taking the exam.
PatsnapGPT really stands out while you have a look at the way it performs in IP-specific benchmarks. Hiro consistently scores higher than general models like GPT-4 on the USPTO Patent Bar Exam. General LLMs fail to pass the 70-point cutoff on the exam, while PatsnapGPT 1.0 scored at the extent of an IP expert. This shows it has a greater grasp of IP fundamentals. Moreover, within the PatentBench, which is a comprehensive benchmark for IP tasks, PatsnapGPT excelled in several areas. It produced more accurate and relevant texts for patent writing, scored higher in classifying patents in line with the International Patent Classification system, and its summaries of technical effects, problems, methods, and abstracts were consistently rated higher by evaluators. It also shows faster speeds and lower memory usage in comparison with GPT-4 for long patent documents.
How do you envision the role of AI evolving in the sphere of mental property and research and development over the following decade?
I see AI playing an increasingly central role in mental property and research and development over the following decade. For one, AI will greatly enhance the efficiency and accuracy of patent searches and evaluation. Advanced AI models like PatsnapGPT will turn into even higher at understanding and categorizing complex technical documents, drafting top quality patent specifications, and identifying potential infringements or overlaps in existing patents. This can save an amazing period of time and reduce the margin for human error.
Furthermore, AI will revolutionize how we handle and interpret vast amounts of IP data. With the power to process and analyze large datasets quickly, AI can uncover trends and insights that may otherwise go unnoticed. This could inform higher decision-making and strategy in IP management and R&D, reminiscent of identifying emerging technologies, potential areas for innovation, and strategic partnerships.
In R&D, AI will drive innovation by aiding in the invention process. Machine learning algorithms can analyze previous research, predict outcomes, and even suggest latest lines of inquiry, accelerating the pace of discovery and development. AI also can simulate experiments and model complex systems, reducing the necessity for costly and time-consuming physical trials.
As AI technology continues to evolve, its integration into IP and R&D will enhance creativity, efficiency, and strategic planning.
