At ABR, we know that our team members are the foundation of our success. That's why we're excited to introduce you to Dr. Trevor Bekolay, our Co-Founder and Senior Research Scientist II. With a passion for exploring the intricacies of learning and memory, Trevor brings a wealth of software engineering knowledge and research expertise to our team. In this blog series highlighting our talented colleagues, we will be showcasing our team members' diverse skills and experiences, and provide an inside look at the innovative work being done at ABR. Join us as we dive deeper into the fascinating world of brain-tech and get to know the brilliant minds behind it all.
My graduate school experience in theoretical neuroscience paved the way for my work at ABR. I was always interested in how the brain works and had a natural inclination toward the world of technology and programming. The field of theoretical neuroscience provided me with an ideal platform to apply my skills and contribute to advancing our understanding of the brain.
My fulfilling experience collaborating with Chris Eliasmith and other colleagues at The Centre for Theoretical Neuroscience, University of Waterloo, set my career path in motion.
“I aspired to see our work evolve into a commercially viable product that could enhance the daily lives of individuals by being implemented in their homes. The prospect of establishing ABR as a company was an exciting opportunity for me."
Previously, I held the position of Lead Developer for Nengo, our software development environment for neuromorphic systems. I’m currently one of the main developers working on the development of NengoEdge.
“NengoEdge is a tool that helps developers create state-of-the-art neural network applications that can run in real-time on low-power devices.”
In the AI industry, many companies are users of AI technology, whereas, at ABR, we're working to push the boundaries of what's possible with AI through our own cutting-edge research and development. Other companies take algorithms, networks, and models and apply them in novel situations. For example, they might use existing models like transformers or LSTMs and ask themselves, "Can we apply this to parsing legal documents to give value to lawyers?"
ABR stands out from other companies in the AI space because we not only develop new and innovative algorithms but also focus on developing models that are tailored to specific use cases and applications. When working with clients, we not only train existing models using their data, but we also evaluate other models and assess the feasibility of applying our LMU model to the specific problem at hand. If those approaches are unsuccessful, we have various other tools in our toolkit, such as SSPs and neuromorphics, which we have experience with and can use to address these problems.
“That’s one of the things I find enjoyable about working at ABR – the depth of experience and knowledge that everyone in the company possesses, particularly when it comes to different AI techniques.”
In the field of AI, there is often a great deal of complex mathematics and statistics involved, which can make it challenging for developers to prioritize creating models that are easy for others to understand. At ABR, we not only have the expertise, but we also have the ability to explain it to each other in a way that allows us to create user-friendly interfaces like NengoEdge.
When I completed my Ph.D., I had a lot of technical expertise in the use of scientific Python. It has been interesting to learn about all the things there are to be concerned about when scaling up software to production, particularly in terms of building and administering containers. Being more familiar with that in my degree would have helped a lot with reproducibility!
Since I came out of my Ph.D., mostly having academic experience, it's also been valuable learning a lot about the business side, especially as a Co-Founder. I've gained little bits of experience with all aspects of running the business. In particular, I've gained extensive experience in managing software licensing and intellectual property (IP) matters. On the marketing side, I’ve honed my skills in website analytics and improved my ability to communicate effectively.
The current project I'm working on right now is adding the ability to talk to our TSP chips in NengoEdge. I'm excited about this project because it spans all levels of the software stack. You can observe the entire software stack in action when working on this project, from the front-end code interacting with the browser to the super low-level C code handling the TSP chip. It's quite interesting seeing all the levels of that and being able to talk to the hardware team to get their insight into why they have the chip working this way.
“I believe that this project represents a culmination of the diverse range of expertise that we have here at ABR. The versatility of the TSP chip we are working on makes me think of many potential applications, such as integrating it into my home automation system, being able to control my garage door through voice commands, and lots of other uses.”
I'm also excited to receive user feedback after people start to use NengoEdge, which will help guide future development.
It has been fascinating to witness the emergence of ChatGPT and GPT-4 and how rapidly they have transformed how people perceive which tasks are uniquely human. Obviously, this has happened in many different fields over time; various machines now do things that used to be done by humans. There are some things, like writing emails, that I'm now happy to have semi-automated. It's interesting that that's one of the tasks now being offloaded to computers. As we experiment more with AI, we realize that for fields requiring substantial expertise, automation becomes limited. I'm excited to see how labour changes over time as we learn to use AI for what it's suitable for but maintain what's uniquely human.
I find ABR's role in this area particularly interesting because, although current ChatGPT text-based and image-based interfaces are popular, they don't currently leverage the potential of voice-based interaction, which is an area where ABR could make a significant impact.
“There's an emotional component to using your voice that changes how you interact. ABR's ability to provide speech-enabled tools will change how we engage with AI in fundamental and hopefully positive ways.”