Why KnuEdge took the long route in artificial intelligence instead of the startup fail-fast approach

  • author image

Editor’s note: KnuEdge, a San Diego-based startup that launched in June after 10 years in stealth, aims to enable the next generation of artificial intelligence computer processors based on neural networks that underpin machine intelligence. At the helm is founder and CEO, Daniel Goldin, whose previous exploits include leading NASA across three presidential administrations, where he initiated robotic exploration of Mars, oversaw the redesign and launch of the International Space Station and drove the repair of the Hubble Telescope. In this interview, he discusses the benefits and challenges of taking a longer-term approach in neurobiologically based algorithms and silicon R&D, rather than the short-term route many of today’s software startups employ.  

What led you to found KnuEdge? 

Dan Goldin: In mid-2000, once I became convinced that we had the space station on track for orbital success, I wanted to get back to a visionary human expedition to Mars. A significant problem with that mission was the communications delay between the spacecraft and mission control on Earth. At the speed of light, it ranges from about 8-48 minutes for a round trip. As a result, real-time Earth-based mission control would not be viable to handle real-time spacecraft issues. Platforms with onboard intelligence –  that could learn and adapt like biological systems – would be required for mission survival and effectiveness. At the time, no such technology was on the horizon. 

I explored the possibility of joint investments between NASA and the industry to undertake a long-term research effort to develop next-generation neurologically based machine intelligence capabilities, which could address the Mars operational issues while at the same time meet future commercial needs.  However, there were deep investments in legacy technologies, so a joint research program didn’t appear feasible. At that moment, I decided when my term as NASA administrator was up I would dedicate the next phase of my career to engaging in the development of next-generation machine intelligence systems that operated at the intersection of neurobiology, computer science and physics. 

Upon retiring from NASA, I joined with Nobel Prize winning biologist, Gerald Edelman, at the Neurosciences Institute in San Diego to obtain a basic understanding of neurobiology and participate in early research on brain-based devices. While at the Institute, we built a robot that could play soccer and subsequently won a DARPA challenge.  With this testimonial to the power of the neurobiological approach, I decided to found a company and raise the necessary capital. 

How did you get your initial investors for your startup venture? 

Dan Goldin: I advised my initial investors I had a specific large commercial application in mind that we would build from scratch and it would take about a decade. Along the way, we would benchmark and sell early technologies to assure we were on track. I knew we had to develop neural biologically based algorithms and we needed a new silicon-based computational platform – one that operates like the brain in heterogeneous sparse matrix calculations. 

This longer-term strategy you took is very different from the “short-term iterations and fail quickly” strategy of many Silicon Valley startups. Did you just realize it would take longer to develop your technology, or did you think the typical Valley model was not effective? 

Dan Goldin: People have been highly successful in bringing outstanding products to market using the traditional Silicon Valley approach. For goodness sakes, look at what entrepreneurs like Mark Zuckerberg have done with Facebook. The technology didn’t drive the application as much as he had a brilliant idea with social networking. I stand in awe of what he has accomplished. I don’t pretend that our approach is better; I’m just saying that for where I felt I wanted to take this company, the longer-term strategy was the right approach. I wanted to accomplish a specific technological transformation. I wanted to apply new biological principles to algorithms and silicon and then bring a new capability to the marketplace, and I knew we needed time to accomplish that. 

Besides being a risk-taker, what leadership skills or mindsets do you think are essential for taking this longer strategy? I assume you need to hire the right talent that is willing to go on that longer journey, but how do you manage a company from this approach that is different from how you would manage it as a CEO of a company in the typical Silicon Valley startup approach? 

Dan Goldin: One of my basic operating principles is to visualize the future and then work backwards to see what specific steps you have to take to get to your starting point. I did what the late Stephen Covey said: “Begin with the end in mind.”   

The second thing is you have to have well-defined intermediate milestones. We started out by taking the easier of the two steps to develop credibility – developing heterogeneous sparse matrix algorithms – before we addressed the more challenging and expensive silicon architecture. 

Is that what led to KnuVerse, the voice-recognition technology your company has developed? 

Dan Goldin: It became clear to me that state-of-the-art voice-recognition systems worked pretty well in good conditions, but they could not perform effectively in real-world noise. Our first objective was set: we wanted to be able to accurately identify a person speaking in the presence of high levels of background noise into most any platform type. We tested our KnuVerse authentication solution in battlefield conditions, and it has worked beautifully. 

How did you manage to find the right talent for your venture? 

Dan Goldin: Sometimes startups hire prematurely, seeking the smartest people in their planned field. I didn’t want to bring people onboard until we knew what problems they were going to address and established a clear corporate vision. The detailed planning process I previously discussed was essential in identifying critical skills and setting the corporate culture, so I could tell the people I was going to interview where the vision was going and what would be the stepping points. Having established these core values, we were then in a position to begin the hiring process in earnest. 

Given we wanted brilliant risk takers, we advertised on Monster.com, and the initial ad read something like: If you’re an incrementalist, interested in improving on the work of others, looking for 20-30 percent improvement, you need not apply. If you are not afraid of failing, you want to take risks and you’re shooting for orders of magnitude improvement, this is the place for you. This allowed us to engage some of our star employees including an internationally recognized neuroscientist. 

Another key issue in hiring is location. I chose San Diego because it had an environment conducive to long-term research and a very strong neurobiological establishment, rich in academic and business organizations. It had a culture where people understood our technological needs and had the patience for our longer-term journey. 

Did you run into any unanticipated challenges that arose from taking this longer route than you would have with a shorter route? 

Dan Goldin: Yes. It has been very stressful, but I’d do it all over again. It’s a little stressful when it takes this long, but I have never been happier. This is the first time in my life when I didn’t work for someone else. 

How did you deal with the stress? 

Dan Goldin: I exercise every morning, I try to sleep seven to eight hours a night, I eat really healthy, love the people I’m working with and have a wonderful family. I like to body surf, ski, and I try to have as full a life as possible; but I work 60-80 hours a week. 

Other than stress, were there other anticipated challenges because of the length of time it took to do this? 

Dan Goldin: Yes. In the early years, we almost ran out of money a few times. I felt personally responsible to the people who had invested in the company, and my wonderful co-workers who  dedicated their lives to our challenging objectives. This has been a heavy weight and inspired me to be more creative in moving the enterprise forward. 

At one point, I contributed a significant portion of my own limited personal finances to show our  investors I was serious about what we were doing. Another time, one of the investors and I lent money to the company, hoping we would be able to raise more – and we did. Those things happen when you’re in a 10-year quest. 

My father taught me to only do one thing and do it well; don’t do many things and have an excuse for failure. So I’m deeply committed and do what has to be done. 

With the faster pace of new technologies entering the market today, how can a CEO leading a long-term strategy deal with the risk that somebody will get to market faster? 

Dan Goldin: You go as fast and as smart as you can while maintaining as tough yet realistic goals as possible. Especially when you’re going up against brilliant competitors with access to significant resources. You have to think and work smarter than they do. 

The reason the pace is speeding up is the tools that are available today are significantly better than the tools of a decade ago. The fact that the tools are so good increases productivity and opens up opportunities for many.  

At the time I started, I knew it would take about a decade, and I hoped that we wouldn’t be overtaken. So far, we’ve been one step ahead. It’s scary. If you asked me my deepest worry, it’s getting overtaken by someone else who didn’t pay their dues but had better tools and more resources. But our people are exceptional, and I believe we will prevail. 

What’s next for your company over the next 12-18 months? 

Dan Goldin: We had our launch on June 6, and we’re working with many interested companies. We hope in another few months to announce exciting relationships with industry leaders. Banks, aerospace and insurance companies have been experimenting with our algorithms and silicon products and obtaining excellent results. I’m not yet ready to talk about what we’re going to do with these technologies, but there is a bigger picture to what we’re doing. 

Read Dan’s commentary, “Former NASA chief explains why businesses can’t afford to play it safe. 

Dan Goldin is the founder and CEO of KnuEdge. Prior to founding KnuEdge, Dan was the longest-tenured NASA chief, serving across three presidential administrations. At NASA he guided the redesign and launched the International Space Station, initiated bold robotic exploration of Mars, formed the Astrobiology Institute to better understand the origin, evolution and destiny of life in the universe, and without incident put a record number of people into space.

Post Your Comment

Leave another comment.

In order to post a comment, you must complete the fields indicated above.

Post Your Comment Close

Thank you for your comment.

Thank you for submitting your comment, your opinion is an important part of SandHill.com

Your comment has been submitted for review and will be posted to this article as soon as it is approved.

Back to Article

Topics Related to this Article