I’m an inventor on several of Zebra’s patents, but I’m not an inventor by trade. I actually spend my days managing business operations for Zebra's Chief Technology Officer, from tracking sales and revenue progress on our new technology solutions to drafting new intellectual property (IP) submissions, strategizing for our vertical customer advisory boards, and fielding requests to evangelize our new technology to partners and customers. It keeps me busy, to be sure.
But I have been around career innovators like Charles Swope and Dr. Biswa Sengupta long enough to know the value of both sharing and protecting brilliant ideas. That’s why I thought we could spend World Intellectual Property (IP) Day talking about the lessons they’ve learned* as inventors about IP rights. I want to ensure the next generation of dreamers and doers know the value of their creativity and understand the actions they should take to maximize the impact of their original ideas.
Matt: You’re both inventors, but your roles at Zebra are different. Tell me about your current focus areas and responsibilities.
Biswa: Along with being a senior technical leader at Zebra, I am part of the Chief Technology Office's Artificial Intelligence (AI) Leadership team, advising the company on topics ranging from machine learning (ML) expertise to computer vision, natural language processing (NLP) and reinforcement learning to research and development (R&D) collaborations, patents, and mergers and acquisitions (M&A). My work is as much about being an evangelist for strategies and technology as it is about the technology itself. At the coalface, I lead special projects (retail, warehouse, and healthcare verticals) that sit at the intersection of sequential decision making, computer vision and NLP.
Charles: My current focus areas are all about disruptor technologies for current verticals and possibly new business units. This includes being part of the Zebra Fellows Community Steering Committee, supporting current business unit needs, and creating the “art of the possible” technologies in proof of concept (POC) or another demo format. My expertise is turning both of these into reality and is another way of saying I am a "research" person. I have been in R&D for my entire career, and currently I am in Zebra’s Chief Technology Office as a Fellow of Engineering.
Matt: When did you secure your first patent? (And what was it for?)
Charles: My first patent was an electrically activated RF switch used by a two-way public safety radio. It was used for several years in one form or another to switch between a radio or public safety microphone used by police and firefighters as an accessory option. But my first real innovative breakthrough was a company award on an apparatus I invented that could be used to determine how robust a Voltage Controlled Oscillator mechanical design was during production and eliminate bad or faulty products.
Biswa: My origins are deeply rooted in academia; thereby, a research paper – instead of a patent – is the currency we typically trade our ideas with. My first paper was titled “Optimisation of timing properties in a platform-independent manner” during the third year of my undergraduate degree, which looked into hardware-software codesign of real-time systems such as fighter planes. It won 2nd prize by IEEE and has been in IEEE’s Hall of Fame.
Matt: Do you always apply for patent protection for your ideas?
Charles: Not always. Depending on how narrow or broad the invention is and its potential business impact on Zebra, I may or may not submit a disclosure. I often seek counsel when I am on the borderline with submitting. Another measure I use to make my decision is to assess if my innovation is enabled to a level that merits submission. Specifically, I ask: “Is there enough detail in the disclosure to completely define my invention such that the novelty is clear to the reviewer?”
Biswa: From my perspective, the field of machine learning is progressing at lightning speed, and so are the corresponding innovations that can be patented. For the most part, my group focuses on patenting features or aspects of a product that are enabled or enhanced by algorithmic improvements. For example, we would like to pursue patents on the way we put together a time-series model for a retailer with specific inputs and outputs. So, many of the ideas my group submits for patent consideration end up being a clockwork handover of data between a few different bespoke algorithms.
Matt: Have all your inventions resulted in or been incorporated into a tangible product?
Charles: Nope. I estimate that 10% may have been incorporated. Many of the inventions I have been involved in are part of existing or future development, so I expect those will eventually be incorporated.
Biswa: A lot of the work done by the CTO group members amounts to what Zebra calls Horizon 3 (H3) work – a set of bleeding-edge algorithmic work which diffuses into product over time. For example, we have published a lot of work on multi-agent reinforcement learning that is being used in development in an Orchestration AI area. Orchestration AI is like a mastermind/oracle for any task requiring planning and scenario-based assessment. Similarly, many of our bespoke computer vision algorithms – localization, segmentation and rendering algorithms – are readily added to our current hardware devices on the other end of the spectrum. So, the short answer is it varies, depending on whether the invention is Horizon 3 type or ready utility as in Horizon 1 type inventions.
Matt: Which one of your ideas or inventions are you most proud?
Biswa: A theme closest to my heart is how we can take inspiration from nature and create energy-efficient neural networks. Remember that a human brain takes roughly 10 watts of power while a supercomputer can often demand megawatts of power – this is what is so unique about the deep neural networks that the evolutionary engineer has crafted. One of my inventions investigated the principles of energy-efficient neural network design. Such inventions reiterate that the semiconductor industry has a long way to go beyond graphics processing units (GPU), tensor processing units (TPU) and other architectures that have a whole deal to learn from primate brains.
Charles: Patent #US 6,633,095, which describes a gearless motor using NiTi material as the mechanical motion technology. I figured this would change the world. (It hasn’t yet, but it will!)
Matt: What does it take to secure a patent? Is there something the patent review committee looks for when reviewing applications?
Biswa: Having a clear-cut protocol that takes you from input to output is the first step as you are thinking about patenting an idea. As a next step, a 15–20 minute discussion with a patent counsel clarifies whether a given concept is a good candidate for a patent application. Remember that you cannot patent differential calculus, but you can patent a warehousing localization and mapping product with the ubiquitous embedding of calculus!
Charles: As Chairperson for one of the patent review committees at Zebra, I can tell you the best way to secure a patent is to provide a thorough writeup of your innovation that highlights your novelty against the current art. This can be achieved by providing a better way of doing it or showing how it is lower cost, lower power, faster, etc. Without that information, patent review committees have a difficult time approving submissions for the pursuit of patent protection.
Matt: Biswa, I know you and your team do a lot of open-source research publishing – and that you also leverage insights shared in the public domain in your research and design (R&D) efforts. I appreciate the collective mind is critical to innovation. Yet, it’s also important to safeguard intangible creations of the mind from theft. Proper credit must be given as due. So, I’m curious how you decide what becomes open source versus closed source?
Biswa: We prefer to seek patents on specific applications of an idea, including AI applied to a particular technical application. When it comes to fundamental AI/ML algorithms or mathematical methods, we are more likely to publish in, for example, academic journals and conferences. We can then use bespoke algorithms as pieces to the "product puzzle," where more often than not, we may end up combining multiple algorithms. Interestingly, when you get the timing right, we can transfer ideas via open-source publishing and simultaneously protect underlying IP via the patent procedure. They are not mutually exclusive.
Matt: This year’s theme for World IP Day is “IP and Youth: Innovating for a Better Future.” So, I’m curious what you were both taught as a youth about IP rights. Did you learn about IP in school? Or is that something that you were more cognizant of once you entered the corporate world?
Charles: I was exposed to patents and innovations as a youth from my father who was a Nuclear Physicist and Electrical Engineer in GE’s Research Division. He always advised me on how to document my work and what value a patent has in industry.
Biswa: On that note, I feel the rubric of IP has evolved quite a lot in the last 50 years. IP as a construct became more evident when I entered the corporate world. As academics, we share everything we invent via papers, talks, and mentorship. Once an idea is published as a peer-reviewed academic paper, one can quickly establish the line of thought via citations. In AI, the speed of invention is mind-boggling – what is a state-of-the-art algorithm today may become obsolete next year. So, the legal framework should evolve to assimilate the innovations quickly or else there is an impedance mismatch between the velocity of inventions being produced and the velocity of consumption by the legal infrastructure. Another recurring theme that organizations often forget is the data they have collected over decades is their IP – this information is as important as the model. Lots to think about here.
Matt: What do you want young inventors and entrepreneurs to know about IP? Maybe it’s something that you wish someone had told you sooner? Or something they should keep in mind as they share their ideas?
Charles: I get this question a lot. First, I want the young inventors to know that "there is no such thing as a dumb idea.” Don’t let others say that your idea, concept, or thought process is bad. Then, I want all inventors to know – not just the young inventors, but all inventors – that "if you don’t try, you don’t get.” So, do not let fear of rejection keep you from submitting your innovations. It may get rejected, but you can always get feedback from me or the reviewer to improve on your submission.
Matt: That’s something I have learned all too well during my time at Zebra. In fact, I got my first patent by being persistent – taking others’ feedback and trying again. Biswa, what about you? What advice do you have for those with big ideas?
Biswa: My advice to young scientists and engineers is to first concentrate on their inventions and then examine them from the legal lens of IP protection, especially if you are working in the corporate world. IP protection and open-source publishing must go hand-in-glove for any company to take from the open-source pool of knowledge and contribute back to that knowledge pool. It must not be a one-way street. In my opinion, patenting products with all the plumbing of algorithms and data is the way to go forward rather than individual algorithms. Data and algorithms must be seen holistically.
Matt: What personal value have you derived from IP? Were you able to secure jobs because of your demonstrated abilities to invent products or solutions that led to IP protection for companies or organizations?
Charles: Well, at first, I just wanted to get a patent, one patent, and I would have been good with that. But then, I received some money for it as part of a company award program. So, I’m thinking, this is a great way to get some extra money and help my company out at the same time. Then as I became more familiar with the process, I was able to participate in the review committees and eventually chair a couple. Getting IP provides exposure to leadership in a critical area that defines a company’s success. Getting IP attaches you to that success and opens the doors for opportunities that others without it do not have.
Biswa: Along those same lines, in AI, your job prospect is directly proportional to what you have contributed as a scientist, the products you have delivered, and the types of teams you have managed. Employers have headhunted me based on the value they think I can bring based on my published work. So, I advise all my students and colleagues: do not just use open-source work and alter it to the product you are delivering. Please give back to the knowledge base – be it your bleeding-edge algorithm or simply a blog post talking about what has not worked. Celebrate your failures and discuss them so that others learn from you. The joy of seeing your invention materialise as a product that many people and companies use gives you a sense of immeasurable satisfaction.
Matt: Did you ever have any qualms about sharing credit with other individuals or perhaps your employers?
Charles No. We often work as a team and it’s common to co-innovate. The key word is “co-innovate” and is what I understand what you mean by asking about sharing credit. If a person is not a co-inventor, then they should not be on an invention disclosure.
Biswa: Charles is right. Innovation these days is a group sport. Even if one person in the group is the leading innovator, often, you will need a handful of people to produce a working product. You will divide and conquer, and sharing credit becomes a no-brainer.
Matt: I know you both are trying to make positive changes in the world, and you spend your days coming up with new ways to make others’ jobs and lives better. You are true inventors and innovators. But that means you have a tremendous amount of IP and only so much time to focus on the patent process. So, in your opinion, what is the best way to protect IP while not slowing down?
Biswa: There is no doubt that human creativity and innovation must be protected, especially in AI. As I mentioned earlier, it is about matching the velocity of inventions in AI with the momentum of the patent process. That suggests a question: should the legal community think of an alternative for AI-infused products and solutions?
The UK's IPO is currently running a consultation on copyright and patents and how they should be protected, especially when the inventor is the AI algorithm itself. For example, AI (non-human) is inventing a new molecule for drug development. Should this be accredited to the algorithm developer or the algorithm (AI)?
Charles: Another vehicle we find useful to protect IP in a time-crunch situation is a provisional patent application, which gives Zebra (or your employer) time to evaluate our options while maintaining our rights to pursue a full non-provisional patent application. We can also put a Non-Disclosure Agreement (NDA) in place while working with other companies or individuals to protect our IP as confidential while we evaluate whether to file a patent application.