The Indian founder relied on artificial intelligence in patent analysis and use even before the ChatGPT hype.
(Photo: Private)
Tokyo How can artificial intelligence (AI) bring real benefits to companies? Indian series founder Komal Talwar has been working on the question for a long time. She and her team have been researching generative AI for years and launched it in their service XLScout before the hype surrounding ChatGPT.
Their goal is to refine and combine the latest AI technologies such as large language models in such a way that the results are really usable for companies. So far, the problem has often been that the results of these programs contain too many errors, explains an engineer responsible for AI analysis at a Japanese chemical company. “80 percent accuracy when recognizing chemical formulas, for example, doesn’t help us.” It should ideally be 100 percent.
But in order to tickle these last percentages out of the systems, the cooperation of specialists and AI experts is necessary. As a result, many companies are springing up in this area that focus on specific industrial applications.
The most important thing is that companies and providers define exactly which problems and tasks they want to solve, the entrepreneur tells Handelsblatt. At least one topic was obvious to them: patents. “We want to democratize their analysis and use,” explains Talwar. At the AI summit of the business newspaper “Nikkei” in Tokyo, she recently presented in a keynote how the technology can already be used in companies today.
She is active in this field with her company TT Consultants, which claims to employ more than 250 people with expertise in various industries. been on the road for many customers worldwide for 17 years. The 44-year-old has now started at points that were previously very complex and expensive. One task is to automate the validation of patents in order to reduce the high rejection rate of almost 50 percent.
The service should also help to generate new ideas from one’s own patents. Another point is the monetization of idle patents, for example by detecting patent infringements or products for which one’s own patent could be of interest. The latest idea is to let AI write patents.
Canadian research and Indian expertise
To implement it, Talwar left India, founded XLScout in the USA and settled the company’s development in one of the world’s leading centers of technology: Canada. “The country is the rising star in AI,” says Talwar. “We work closely with universities and professors there.”
The cooperation gave her access to staff and the latest AI technologies, which she was then able to optimize with the expertise of Indian consultants for various sectors such as the automotive, steel and chemical industries. She was also able to tap into state subsidies, says the Indian woman. “We became part of the ecosystem, which then helped us develop our products.”
The experts do not rely on one technique, but on different methods depending on the problem. The team uses open source language models, which are then fed with industry-specific databases. The system has access to over 150 million patents and more than 200 million specialist articles.
More about artificial intelligence
“In some places they already work very well, in others they still need to be fine-tuned,” says Talwar. In other cases, a task is approached with supervised or unsupervised machine learning or natural language processing.
At first the offer was not a sure-fire success. “It was a difficult task for us to even explain to people what AI can do in the different use cases,” says Talwar. “But over the past year, ChatGPT has suddenly made our job so much easier because now everyone wants to use generative AI.”
This also includes German customers. She now wants to open her own office in Munich for them. Talwar has had an agency in Japan for some time.
More: Official replacement or bureaucracy turbo? How AI should save the state.