How Artificial Intelligence is Used to Expedite Research
Our underpaid MBAs spend a lot of time doing research here at Nanalyze. We want to be as comprehensive as possible to ensure that we’re bringing our lovely readers the best stories about technology and retail investing. But that takes time. There are close to two billion websites out there on the interwebs, so tracking down the most relevant and reliable information is like trying to find an honest broker on Wall Street. And good data costs money. For example, data firm CB Insights is the gold standard when it comes to market research, partly driven by machine learning algorithms, but you have to be the New York Times to afford a membership. When it comes to more specialized research, Google often falls short. Now, a growing number of startups are using artificial intelligence (AI) to do the heavy lifting for online research. Let’s take a look at three examples.
Using Artificial Intelligence to Expedite Research
PatSnap – AI for Intellectual Property Research
Founded in 2007, PatSnap has headquarters in both Singapore and London, along with offices in the United States and China. The company closed a respectable $38 million Series D this month, led by venerable venture capital firms like Sequoia Capital and newcomers like Shunwei Capital. The latter is an investment firm from the co-founder of Xiaomi, a Beijing-based smartphone and electronics company on the cusp of a big-time IPO on the Hong Kong stock exchange. PatSnap has reportedly raised more than $100 million, but an exact figure is unclear because of an undisclosed Series C back in 2016, which Sequoia Capital also led.
PatSnap uses artificial intelligence to parse through big data for insights around intellectual property, patents, and R&D. The startup combines millions of data points from patents, licensing, litigation and other sources to provide intelligence that helps companies to streamline their R&D departments. The idea is that the PatSnap platform can provide near-instantaneous data about the patents and associated data surrounding any type of technology. In a sense, it allows companies to gather intelligence about their competitors using publicly available information. Clients could anticipate what sort of products a rival may be preparing to roll out based on patent activity, or discover a potential partner with a complementary technology.
Incidentally, this month marks the 10 millionth patent issued by the U.S. Patent and Trademark Office.
Example of PatSnap Platform in Action
Here’s a real-world example of how it works: A client called G2O Water Technologies has developed a graphene-based water treatment technology. G2O used the PatSnap platform to determine that one piece of intellectual property that the company wanted to patent had little competition. That meant it could tell its IP attorneys to draft a very broad patent, with a better chance of getting a large number of broad claims granted for its technology. As the CEO said:
Rather than trying to formulate a query to send to a patent lawyer’s firm, and wait to get some results back, we can just have a very quick look and see what’s going on. As we can do the searching ourselves, rather than having to get someone else, I would say PatSnap has probably saved us between £10,000- £20,000 a year.
That’s about $13,000 to $26,000 for you Yanks – significant savings for a small startup that relies mainly on government grants for funding.
PatSnap Platform Used by Chemists, Scientists and Attorneys
Last year, PatSnap rolled out a specialized IP search platform for companies that operate in chemical development called Chemical by PatSnap (apparently the startup’s IP platform doesn’t extend to registered trademarks for more interesting names). Chemical by PatSnap enables organizations to validate projects by providing AI-powered insights from its database, with links to more than 114 million chemical structures, clinical trial information, regulatory details, toxicity data, more than 120 million patents and other sources.
PatSnap also markets its IP platform to other industries and professionals, such as IP attorneys themselves or scientists conducting research. For example, PatSnap can identify technology sectors that are receiving the most investments. Researchers and scientists can then pinpoint their particular area of focus and see any funding associated with that research field, including grant information. That helps the cash-strapped scientist find a new deep-pocketed sugar daddy.
PatSnap claims that more than 8,000 companies, government agencies and academic institutions in 28 countries use its “invention” intelligence platform, with more than half of its clients in China. That list includes NASA, the city of London, and Quorn Foods, which produces meat substitutes in a marketplace facing increasingly sophisticated competition from companies fermenting fake meat or growing real meat in a lab.
Iris.ai – An AI Research Assistant
Speaking of laboratories, PatSnap isn’t the only startup offering research help to beleaguered scientists who need to spend more time answering the tough questions, like why you can’t rely on Facebook friends or how you can tell the difference between a liberal and conservative.
Founded in 2015, Iris.ai has created what it calls the world’s first AI-powered research assistant. The startup out of Oslo, Norway, raised $2.4 million in disclosed funding from angel investors and a pair of Seed rounds, the most recent coming last December. The process begins with a research question, paper URL, or TED Talk URL. Iris then “fingerprints the query using natural language understanding, and matches the fingerprint against more than 83 million open access papers.” You can try it for free here.
Iris.ai says it can shave 90 percent off the time it takes industrial researchers to build a reading list, which normally requires three weeks on average. That works out to a savings of about 432 hours per year, according to the startup. Its results are also more precise, with 85 percent of the materials relevant to the research against a self-reported 70 percent. The folks at Iris.ai hope that the AI platform will grow into a real research scientist someday, offering its own hypotheses and awkwardly flirting with graduate students. Easier said than done, of course, especially when you look at a mega-company like IBM that has yet to fully capitalize on perhaps the world’s most well-known AI platform, Watson, in the healthcare industry.
BenchSci – An AI-Powered Search Engine for Antibodies
Founded in 2015, Toronto-based BenchSci has raised $8 million in disclosed funding, with Google’s AI-focused venture fund co-leading the recent Series A. The company officially launched its machine learning platform last year for speeding up drug discovery by making it easier for researchers to find antibody usage data. The platform has since analyzed data on more than four million commercial antibodies. (You’ll recall from some biology class taken in the distant past that an antibody is a blood protein produced in response to a foreign invader to your body such as a bacteria or virus.). BenchSci claims researchers can find reliable antibodies 24 times faster and 75 percent cheaper than current methods.
BenchSci says its customer base has grown to include 14 pharmaceutical companies—including seven of the world’s top 10—and more than 900 academic research institutions, such as Harvard, UCLA, Stanford, and MD Anderson. Maybe now we can get that herpes vaccine we’ve all been waiting for.
Update 02/04/2020: BenchSci has raised $22 million in Series B funding to further develop its product suite and help expedite drug testing. This brings the company’s total funding to $44.6 million to date.
We’ve come across an oft-quoted statistic that says 83 percent of jobs making $20 an hour will be automated someday. The average salary of an entry-level research assistant is about $33,000. You can probably do the math: AI is coming for yet another white-collar job, whether we’re talking about occupations in the legal field or the much-maligned (mostly by us) human resources department. Companies are finding ways to use artificial intelligence to restructure all of that big data out there into more precise, easily digestible forms where finding pearls of wisdom on the interwebs will no longer be like trying to find the proverbial needle in the haystack. Those that do it successfully will be spinning that hay into gold someday.
Here at Nanalyze, we invest the majority of our savings in 30 dividend-paying stocks that increase our income every year and outperform during both bull and bear markets. Find out which ones in the Quantigence report freely available to Nanalyze subscribers.