Pattern recognition using Neural Networks has come a long way – When I first got exposed to it almost 25 years ago, a 500 node Neural Network would take days to train and would perform pathetically compared to a standard Bayesian classifier. The explosion of computing power, especially the parallel processing power in the GPU’s, has made training of large, deep neural networks possible. Given the computing power and the advent of various flavors of Neural Networks, many of the pattern recognition problems are well solved today. Image Recognition and Speech recognition have reached near 100% accuracy because of these deep models.
As a fund, we were looking at pattern recognition problems using AI, especially as applied to industry verticals. Our thesis was that to create significant value, technology needs to go beyond images and understand higher level context – actions, sequence of actions … If a system can recognize a complex context, spanning arbitrary time frame, that would be powerful and can be applied across variety of settings and industries. However, as you go towards understanding of complex contextual situations, the problem gets exponentially harder and requires lot more than just computing power. We were looking for companies that would tackle this problem. We didn’t find very many companies.
About a year ago, Nagraj Kashyap and Samir Kumar from M12.vc introduced us to Roland, Moritz and the TwentyBN team. Right off the bat, in the very first meeting, we were impressed with their technology and vision. The team came with a great pedigree as Roland himself was one of Geoff Hinton’s students. TwentyBN had spent a lot of time at the problem of understanding actions and context and built a solution from the ground up. If recognizing context and actions itself is a hard-enough problem, they were taking it one step further and solving it in real time and at the Edge. To say that we were hooked, would be an understatement.
As we started working with TwentyBN, we got to know the team better. Every company is a journey. I have rarely seen a startup go in a straight line in terms of direction, focus, milestones or scale. There is constant evolution and course-corrections and setbacks and adjustments – learning from the market, customers, investors and team members. It is a fine balance though – management team should be willing to take feedback and course-correct, but not so fickle that they are flip flopping on focus every day. They need to have strong conviction on the value they bring to the market but not too stubborn to listen to what the market is saying. In my experience, the companies that reach greatness are the ones that manage this fine balance. Working with the TwentyBN team through the last 12+ months, we were impressed with the team’s willingness to take critical feedback while being steadfast on their conviction and execution. We watched the team launch the Digital Avatar product, Millie, based on their underlying technology with a focus on Retail. The company in turn, has been recognized by CB Insights as a Top 100 AI company and Top 5 Retail AI company.
One aspect we look carefully when investing is the composition of the board – it is often underestimated. Having investors around the table that help with the company’s journey and can make decisions quickly and decisively is essential. Between Peter, Samir, Christian, Alex and Lior, we are very excited to have a great set of investors and people on the board.
It is fair to say we are very bullish on TwentyBN and the value that they bring to different industry segments. We believe that they are creating a new category with their technology. I am sure there will be lots of excitement, learning, setbacks and wins along the way and we are excited to be part of the journey ahead with TwentyBN!