For our next specialist interview in our series speaking to technology and IT leaders around the world, we’ve welcomed Jeff Stephens, co-founder and CTO of EyeQ Imaging Inc.
Jeff has been in product development and technology since graduating from the University of Texas at Austin with a Mechanical Engineering degree and started out in hardware development for Dell, Inc in 2006, but always was interested in software and web development as well.
Tell us about the business you represent, what is their vision & goals?
EyeQ’s aim is to improve the quality of digital imagery for businesses and photographers. We’ve been at this since the early 2000s and are leading the industry experts in automatic image correction and optimization.
What was your experience like moving away from mechanical engineering to software engineering and AI?
To a certain extent, all engineering is about the same thing: solving practical problems with a dearth of information. But - I have noticed one interesting similarity between Mechanical and AI product development that “normal” software engineering sometimes lacks:
When making physical things, “builds” are big, expensive events and they are not undertaken lightly. Prototype products can cost 100x the final product cost - so every prototype you make needs very careful consideration, careful study of what you are building, and what you’ll be able to learn from it. In the “standard” software world, builds are basically free. Nightly builds run without any thought, and the artefacts just pile up, ready to be tested. This allows much more rapid development, but also makes each “build” less important - given less consideration.
In the AI world, “training” is analogous to the physical build: they can take days to complete and are far more costly than the essentially free software build. This makes each training far more important and requires more forethought and test planning.
This might seem like a trivial or semantic difference, but I think it’s important. Standard software engineering’s “nightly build” approach can be very useful, but it can also introduce a more lax change review process. This can permeate the whole development process - not just “cheap builds any time” - but less emphasis is put on specs, goals, testing, discrete development phases, etc.
Can you tell us a bit more about your experience with Bibble Labs?
I left Dell in 2006 to join a 5-person company that was building the fastest digital RAW file workflow application. Huge corporate behemoth to tiny startup. This was a culture shock - and one I badly needed! Bibble Labs was an amazing experience - a small, fiercely dedicated team building the tool we most wanted in our own photo workflows. I learned so much from this time: technical, team management, M&A, running a business, etc.
What do your day-to-day responsibilities look like at your organization?
As CTO - I don’t build the products; I build the team that builds the products. So my main job is tending the team.
Each person needs to be focused on the right work, and the right time. We’ve been fully remote for over a decade and geographically dispersed as well. So a big part of my job is coordinating the efforts of the team when multiple people are working on the same project. This bleeds over into the HR world as well - is everyone happy in their jobs, do we have the right people for the workload we need to deliver, and the right tools and processes.
Then, there’s the product management side: are the specs detailed enough to give to the developers? Will the end result actually meet the market needs? How best to plan and deliver the larger development tasks - year-long projects, etc. Is the general direction of our progress towards something exciting and valuable?
Then, there’s customer support: are all of our customers thrilled by our products? Is the experience with them as good as it could be? Are the docs and on-boarding of our SDK’s clear and concise? Do we resolve issues quickly?
Can you share some of the proudest achievements you've experienced in your career?
The “achievement” that means the most to me is my current engineering team and process. Building a great team is hard; a group of great people does not simply make a great team. EyeQ’s overall development and innovation capacity is a result of hard-fought wins on many fronts.
In which industries and processes do you see the greatest opportunities for the application of artificial intelligence?
AI is a fascinating space; the results can be truly amazing, but it never seems to keep up with the hype. Where are the self-driving cars, after all!
AI is focusing now on the problems that a 4-year-old can trivially solve: is that a dog, or a cat? Is that a stop sign close to me, or a billboard a long way away? Is that person happy or sad? Is that signal or noise? So, a huge amount of effort is focused on these pattern-recognition chores.
The exciting thing for me is to find complex, real-world problems that can be solved, wholly or partially, by these “trivial” pattern recognizers. So, so many of these problems start with digital imaging, visible light, lidar, sonar, radar, X-rays, seismic, etc. These are all just “images” - and discerning information from them - at the right time - can lead to so many exciting possibilities.
What are the most significant changes you expect to see in business as a result of AI?
I don’t expect AI to fundamentally change how businesses run or compete with one another. AI is the steam engine: it’s a new, powerful tool - but it does fundamentally the same thing as the water-powered paddle wheel. Yes, the steam engine changed the course of humanity - and AI has that potential as well. I expect we’re decades away from ‘Artificial General Intelligence’ - so for the time being, it’s evolutionary, not revolutionary.
So, the impact on businesses will be determined by how each business adapts. Which ones jump too soon or too far? Which ones are too timid and slow to change? Which ones best add the new capabilities that AI affords to their existing capabilities? In other words: corporate change management is like any other change.
One specific thing that I find exciting in the AI / ML field is the level of “openness” of the field. AI has been deeply connected with open source. Open-source models, open-source data, open-source processes. Academia and business have never worked as closely and cooperatively, and to me, that’s what has enabled the speed of adoption of AI by non-research organizations. Businesses that adapt to this best will benefit the most.
What is your experience with using AI-backed data analysis or log management tools? What do you think is the benefit of using a log management tool that has machine learning capabilities for an organization?
Modern life generates a huge amount of data. Every GPS tracking everyone walks with their phones, every call or key-press, every router connecting us. Every weather sensor. All the dozens of telemetry sensors on billions of cars and IoT devices. That data is useless until it can be collected, cleaned, reduced, and finally understood.
Once it’s understood, then it's up to people to make use of it - make decisions to affect that data. This is a mining operation - seeking the valuable nugget from the mountain of rubble. The better and faster this can happen, the quicker we can find those patterns, and the faster we can react.
What is EyeQ doing differently than anyone else when it comes to AI?
EyeQ blends AI-based imaging with classical imaging in a way no other organization does - that I’m aware of. AI is an excellent hammer, but not everything is a nail. What EyeQ has been able to do is to cleverly use all the tools in the toolbox - not only the shiny, new ones.
Many of our customers are excited by the possibilities that AI affords, but likewise, they are nervous about the change. The ‘just right’ solution was the right blend of known / classical imaging paired with the right amount of new / AI imaging.