Imagine that you own a space exploration company, and you need to design a suit designed to go to Mars. That is a tall order, and it will require a lot research—may begin with a few key questions: What materials should I use? How do I make the suit temperature control? What about blocking UV light and keeping dust out? Has anyone else tried to do this before?
Now, there’s an AI bot trained to get those answers for you.
Eureka is a new tool launching today at Patsnap, and it’s designed to simplify the early stages of product development. Patsnap (as in, “patents in a snap”) is an AI-powered patent intelligence company founded in 2007 with the goal of helping companies research intellectual property (IP) and scientific discoveries. If, for example, a company was working on developing a new jetpack, Patsnap could retrieve existing jetpack patents, summarize them, and track them in case new ones appear. These tools help companies ensure that they are not infringing on the IP of others while discovering the latest innovations in their field.
In 2021, Patsnap reached unicorn status by surpassing $1 billion in value, and, in 2023, it reached $100 million in recurring annual revenue. The company has a large list of clients that include companies such as Disney, PayPal, Tesla, Bose, and Huawei. Eureka will be Patsnap’s first AI-powered tool dedicated to research and development (R&D), and it could have a major impact on how businesses test new inventions—and how quickly those products can be brought to market.
An LLM specifically designed for product development
The basis of Patsnap’s various AI-powered tools is its unique language learning model (LLM), called PatsnapGPT. Specialized in product development and IP research, it draws on more than two billion news articles, 200 million patents, 254 million chemical structures, and other domain-specific information to help inform the innovation process. In May, Patsnap introduced its AI assistant, Hiro, which can answer questions and hold a conversation similar to ChatGPT, but answers specific questions about IP and R&D.
According to Frazer Kearl, senior product manager of Patsnap, there are two main reasons for Patsnap to invest in its LLM and chatbot rather than partnering with a large AI company, such as Claude or OpenAI: data security and accuracy of responses. First, Patsnap complies with data privacy laws including the EU’s General Data Protection Regulation (GDPR) and the AI Act, as well as the California Consumer Privacy Act (CCPA), and customer user data is not used for model pre-training or re-training. This assures clients that their sensitive data will not be used against their consent in any way.
“Then, in terms of performance, we are able to generate better responses with our AI technology, because it is specifically trained for the conditions in our domain,” said Kearl. “Reinforcement learning is done by experts in the field that are specific to our clients’ industries, and are very focused on those jobs.”
With Eureka, Patsnap actually dials that clarity up a notch or two. In a special demo with Fast companyKearl showed how this feature can help a product developer bring a new product idea to life, from a macro-level question to a high-resolution idea.
AI product design assistant
Eureka has several different features to help a product developer make that leap. To start, there is a simple Q&A chat section with Hiro, which is almost like a ChatGPT or Claude box. Here, the user may ask their most general question; For example, What is the latest space suit technology?
After that, Hiro will suggest follow-up questions to help narrow the scope of the search. If the user answers—say, “more about life support systems”—Hiro will report on the universe around the technology, breaking down the key players, what trends in the space look like, and how much is produced each year.
“All of these responses are extracted from our data sets,” Kearl said. “Since we were founded in 2007, we’ve been able to isolate these specific data sources—whether it’s patents, scientific literature, VC investment, market size, investment from a variety of government agencies—so it’s ready. [Hiro] to retrieve. That means the answers they provide are very useful because they’re not doing it ‘ad-hocly’ and trying to go online to find the latest news article.”
After a researcher digests Hiro’s wide-ranging report and decides on a narrow case study, Eureka’s next section, called Tech Mind, helps users formalize the problem they want to solve commercially. In this case, the user may enter information, What materials can be used to make space suits lighter?
Pulling from the PatsnapGPT database, Hiro will then respond with an overview of the many different options and their merits (in this case, he suggests airgel fabrics and aluminum-lithium alloys, first). The program will raise application cases and conditions, as well as other parameters to reduce the field of study even more. On the side, the “Expand Your Knowledge” bar allows users to dig deeper into the resources used for more information.
All of this background research is combined in Eureka’s final step, called Find Solutions. Here, the user takes everything he has found so far and explains to Hiro the exact problem he wants to solve. That would be something like this, I want to significantly reduce the weight of space suits by introducing airgel composites while maintaining the thermal performance of the suit.
The software then produces a number of practical solutions that can be tested in the lab, including specific programs that set the path for in-lab testing. To qualify for a Mars suit, Hiro suggests trying an aerogel-metal composite insulation (among several other options).
As an initiative of Tech Mind, all sources of AI and its methods of solving problems are laid out for users to check whether they agree with Hiro’s thinking or not.
“If for any reason, a user says, ‘Hey, that looks suspicious, or it’s like a magic bullet, let me go double-check what’s going on with it,’ they can go ahead and do that,” Kearl said. AI for the user, because ultimately, our goal is to make their R&D results.”
Future use
Essentially, Eureka serves as the ultimate partner for brainstorming and researching new ideas, allowing human researchers to avoid the time-consuming manual work that comes with collecting data across the board. Alternatively, Kearl says, “we go to a website, we do a search query, we get hundreds of results, we try to search those results manually, we’re exhausted, we probably ignore 80% of them—which happens more often than you’d think. — and then we move forward and hope for the best.”
Eureka helps product designers get started, but people are still the ones doing the real testing and testing. If their initial ideas don’t work, they can go back to the drawing board with specific constraints for Hiro based on their findings.
DOW Chemical Company, a Michigan chemical manufacturer, has been using Eureka in beta for the past few weeks. According to Scott Matteucci, capital manager for coatings, plastics, and additives, the tools help reduce the time spent searching for market trends and new materials from “weeks to seconds,” giving the R&D team “more time to brainstorm, experiment in the lab, and drive R&D.” ours.”
Given that Eureka is new to the market—and the fact that it takes at least a year to sell a new product in most industries—the tool hasn’t been used to actually produce new things yet. But Patsnap’s senior head of marketing Micky Teng says the tool, alongside Patsnap’s tech suite, can have a big impact in industries where turnaround times are critical.
“I think there is such a huge global benefit to what we do for our clients, because they are working to be successful,” said Teng. “With [COVID-19] vaccines, for example, BioNTech is one of our customers, and they work with the Moderna vaccine. [Our customers are also working on] electric cars, such as Rivian and Tesla, and low-emission aircraft technology. It is the customers who shape the future of the entire industry.”
Source link
