Patent applications for AI technology skyrocket
Large tech companies and other inventors have tried to sweep up a huge number of artificial intelligence-related patents in recent years. The result is that some smaller and less patent-focused firms are worried they could be shut out of the emerging technology.
A recent National Bureau of Economic Research study found that the number of patent applications mentioning AI and machine learning has skyrocketed since the early 2000s. That's also the case with patent applications involving neural networks, those computing systems modeled after the human brain’s learning process.
In 2004, just 58 U.S. patent applications referenced machine learning, with that number jumping to 197 in 2011. Then in 2016, 594 patent applications mentioned machine learning, and that number is likely to rise because the U.S. Patent and Trademark Office makes filings public 18 months after they have been registered.
The numbers aren’t quite as dramatic for neural networks, but those still rose significantly in recent years: From 1990 to 2008, the number of patent applications mentioning neural networks held fairly steady at 100 to 200 per year, while the number jumped to 346 in 2015 and 485 in 2016.
Big vendors, deep pockets
Meanwhile, many of the patents granted are going to large tech vendors. Tech vendors in the top 10 included the following: Microsoft and Microsoft Technology Licensing together received 206 machine learning patents between 2000 and 2015, according to the report. IBM was granted 117, Google 66, and Amazon Tech 37.
For neural networks, IBM was granted 112 patents in that 15-year period, Google obtained 78 patents, and Microsoft and Microsoft Technology Licensing were granted 71.
Some tech companies have used patent lawsuits as a tool against competitors in recent years, and that concerns AI users.
“As a start-up founder, I live in constant fear of litigation from enterprise companies with deep pockets filing patents and protecting their status quo by burying me in legal expense,” says Karrie Sullivan, principal at Culminate Strategy Group, a digital transformation consulting firm. “They can bury me in patent litigation, regulation, lobbying, etc.”
Sullivan’s second company is working on an AI and blockchain technology for the employer healthcare market. “I don’t have the cash yet to file 50 patents to protect my [intellectual property], so I have to prioritize,” she says.
Still, Sullivan is optimistic about the future of AI and the opportunities for innovation. There is “still a ton of white space” where AI can be combined with other technologies, she says.
Plenty of untapped opportunity in AI space
“The real thinking and application of AI is still untapped,” Sullivan adds. “As long as humans are shackled to low-value tasks, we won’t really understand the boundaries of human potential or the new economic models that will drive patents in the future.”
Other patent and AI experts downplayed the potential harm of the AI patent rush. The America Invents Act, passed by Congress in 2011, makes it easier for companies to challenge competitors’ patents, and a 2014 Supreme Court case has made it more difficult for companies to get software patents.
In addition, an AI patent in itself doesn’t necessarily translate into a useable AI technology, says Praful Krishna, CEO of enterprise AI vendor Coseer. A good AI system includes not only a basic algorithm, but also a team of data scientists and a storehouse of relevant data, he adds.
“While it is impossible to predict what will happen in a patent litigation, I believe that a large number of AI patents being filed now will turn out to be useless,” Krishna says. “The real value of AI solutions does not come from the technological innovation for the most part.”
Most companies using AI already own the most important piece of the puzzle: the data, Krishna says. “Each enterprise needs to craft their own AI strategy, work with smart vendors, and chart out their own solutions without worrying about getting locked out.”
Smaller businesses need a patent portfolio, too
Second-tier enterprises and even small businesses shouldn’t cede AI innovation to the big guys, other experts say.
Joe Lafata, an intellectual property attorney at law firm Harness Dickey, says AI users shouldn’t count on challenging patents and getting them overturned. Instead, smaller companies should focus on creating their own AI technologies, he recommends.
A smaller company’s own AI inventions may not only give it a niche market opportunity, but patented AI inventions can be used as bargaining chips in licensing deals or legal negotiations with other AI vendors, Lafata says. It’s the classic “We’ll let you use our technology if we can use yours.”
“It’s always better to have a patent portfolio than to not have a patent portfolio,” Lafata says. “It gives you possibly some leverage against bigger players that might try to assert their patents against you. You never want to go to a gunfight without bullets.”
Think of patents as trading cards, adds Alan Majer, founder of tech startup advisor Good Robot. “You need to have a strong hand to ensure that you don't get shut out yourself,” he says. “Many of those filing AI patents are probably already doing so as a defensive measure.”
Small companies shouldn’t overlook their own work on AI systems, Majer adds. If a small company has worked for two years on an AI technology, there’s likely an invention or two in that effort. “If the company has the wherewithal to apply artificial intelligence, I think they have the wherewithal to have innovation within the artificial intelligence space,” he says.
While large companies may focus on quantity of patents, smaller companies should instead focus on the quality of their inventions. “If they have a limited budget, then they can’t necessarily file 10 or 15 patent applications on a particular device or a particular solution to a problem,” Lafata says. “But they can file two or three.”
IP attorney Lauren Hockett at law firm Knobbe Martens isn’t concerned that AI patents will be concentrated in the hands of a few large companies. “Just based on my own practice, I see people across the board investigating new ways to use machine learning,” she says. “It’s not just a handful of companies; it’s large clients and small clients.”
But with the costs of filing for a patent—often running into the tens of thousands of dollars—some AI users may want to avoid the process altogether, Hockett adds. Companies should conduct a market analysis before deciding whether to file their own patents.
Another downside of a patent application is that the inventors have to make their invention public, Hockett notes. In some cases, AI inventors may instead want to rely on trade secrets to protect their technologies.
Majer recommends that AI users create their own innovations but also collaborate with other companies. Several existing projects provide open access to AI tools, he notes. “Supporting and using those projects is a great way to maintain your freedom to innovate,” he says. “Working with like-minded organizations to use—or even build and contribute to—open source tools or shared platforms is a great way to ensure that no single organization cuts off the possibilities for innovation.”
Note: The study referenced used IFI Claims patents dataset as its source of information on patents.This dataset contains the full text of all published U..S patent documents through February 2018 obtained from USPTO bulk files. Download the study for further info.
AI patents: Lessons for leaders
- The number of AI-related patent applications in the United States has skyrocketed in recent years.
- Smaller companies can develop their own AI technologies to avoid getting locked out of the field by other companies’ patents.
- Applying for patents may give your company leverage against other patent holders.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.