Patenting AI: Let’s start with a history lesson
When I start talking about artificial intelligence (AI) I go into full geek mode. This stems from a fascination with technology and all of the ethics and human factors involved. As someone who started thinking about the trajectory of artificial intelligence and the impact it could have on social collaboration tools back in 2014, I developed a personal point of view that has stood the test of time: unlike typical software systems where you’re clicking buttons to dependably achieve your goals, with AI you’re working with systems that cannot offer 100% reliability — therefore trust becomes the new currency to measure success. Which begged the question: What does it mean to design for trust? Lots of research and design investigations ensued to explore that question.
It was while I was forecasting the future of social collaboration when I stumbled across patenting at IBM. I reached out to an IBM Watson developer, Corville Allen, to vet the feasibility of my ideas and he just so happened to be one of the most active inventors in the Research Triangle Park area. What luck! He quickly became my mentor and opened me up to a whole new world of design; playing with the art of the possible, unbounded by deadlines, resources, or budgets and powered by smart systems that could meet my end-users on their terms, not the other way around. My kind of playground!
I currently have 13 AI-related patents under my belt and want to start transitioning into a mode where I can give back to the IBM patent community and bridge the gap between design and AI patenting. Naturally, designers tend to not be technical experts and as a result AI can feel daunting and out of reach. Let’s cut the crap and feel empowered to help drive this movement! Shall we?
I cannot emphasize enough how fascinating it is to look at the history of AI patenting activity. It tells a story of a transition from research to application, a global arms race with investments in different sectors and the, literal, number one role IBM has played.
Fortunately, getting this information came gift wrapped with a bow in the form of the World Intellectual Property Organization’s (WIPO) 2019 Technology Trends report on Artificial Intelligence—looking at trends in AI innovation since the field first developed in the 1950s .
Let’s kick-off with the most interesting finding: 50 percent of all AI patents have been published in the last 5 years. That’s based on nearly 340,000 AI-related inventions and over 1.6 million scientific publications to date. I mean…wow! Just think about it. The first AI patent gets issued in Japan in the early 1980s, yet we’re just now seeing technology realizing its potential. This is a really exciting time to be an inventor, so get out your chisels we’re part of a renaissance!
Right now, companies represent 26 of the top 30 AI patent applicants worldwide. IBM is in its 26th consecutive year in overall U.S. patent leadership and it maintains that lead even when we isolate its AI patent portfolio—with 8,290 inventions, followed by Microsoft with 5,930.
When we break this down into the 25 categories and sub-categories for AI techniques you can see that IBM leads in 10 of them.
Up until now, I had always heard AI broken down into techniques such as regression, classification, unsupervised, and reinforcement learning models. The WIPO trends report opened my eyes to how patent law categorizes AI efforts with lots of sub-categories I had never heard of. (Bio-inspired? — sounds interesting.) As an IBMer, this exposes me to a more nuanced understanding of AI in relation to my company’s investments.
“Technology trends can be discerned through patent analytics.” — WIPO
Now play along.
I want you to think about the world around you. Pay attention to some of our everyday modern conveniences. I voice a command to turn on the lamp from my bed and moments later the room brightens. A Tesla preemptively accelerates, anticipating a car accident that hasn’t happened yet. My phone gets a traffic alert based on a typical upcoming destination. I start to respond to an email and the rest of my thought spills onto the page in the form of predictive text. These are just a few embodiments of machine learning. As you can see from the figure above, it’s the AI technique that makes up more than one-third of all AI patent disclosures (134,777). This dominance in the patenting world has a clear connection to where narrow AI is currently excelling, where the world has been investing its efforts, and what was inevitably going to become market ready first.
Everything we experience that uses machine learning today is directly linked to the innovations first being pushed in the scientific research and patent community. It’s a beautiful symbiotic relationship that everybody benefits from. We need inventors to push the technology, they need consumers to secure investments.
Currently, there are 4 fields and industries that are referenced most in AI patents and it realllllly shouldn’t surprise you. Would you be shocked to find out that telecommunications (think internet of things and methods for exchanging information) is highly referenced? Would your jaw drop to hear that transportation (think planes, trains and automobiles) is being heavily explored? Would you slap your grandma silly if you heard that the field of life & medical science was being innovated? Would you… Well, you get the point. Of course not. The impact from these industries is practically ubiquitous at this point and we could have seen it coming just by looking at historical trends in scientific publications and patent disclosures.
It’s also fun being armed with these patent trends and then looking, generally, at a timeline of Machine Learning…
You see how the evolution from scientific research to company-driven innovation is reflected. You can decipher the different AI techniques that map to each category and sub-category driving each milestone. By looking at who is working the hardest at what in patenting, you can start to understand the world around you. For example, according to trends in AI functional applications, computer vision is the most popular which has manifested in hundreds of ways—it’s actually pretty hard to surprise us with computer vision anymore. Deep learning has the biggest recent growth in the field, at 175 percent average annual growth between 2013 and 2016 along with other machine learning techniques such as multi-task learning (49 percent) and neural networks (46 percent), so we can be confident they will drive ubiquitous manifestations in our near-future. I seriously better be around to see this.
“We are now in the age of AI implementation.”— Kai-Fu Lee
As designers, it is our goal to leave the world better than we found it. Part of our process in Design Thinking is to explicitly focus on our end-users and not the technology that drives our solutions. This was an easy feat with traditional computing where everything was just a simple algorithm or graphical interface adjustment. But now we’re in the big leagues. We would be remiss not to focus on technology. The history of patenting AI not only helps us understand the opportunities that artificial intelligence is unlocking but keeps us one step ahead by seeing what’s coming next. Expanding our toolkits with the art of the possible keeps solutions from falling off the radar for seeming too far fetched—instead we can be drivers of innovation manifesting high-value business solutions. Our grandkids will thank us. 🔮
I want to leave you with a nice little cheat sheet, of sorts, where WIPO outlines not only what each AI technique means, but a summary of all functional application and application fields.
The entire report can be found here: https://www.wipo.int/tech_trends/en/artificial_intelligence/
