CO-FOUNDER & CEO, SKYMIND
Deep learning library.
Chris Nicholson is co-founder and CEO of Skymind, the company behind Deeplearning4j, which is the most popular open-source deep-learning framework for Java. Previously, Chris led communications and recruiting for FutureAdvisor, a Sequoia-backed Y Combinator startup that was acquired by BlackRock in 2015 for $200 million. He also spent a decade as a journalist, reporting on tech and finance for The New York Times, Bloomberg News and Businessweek, among others.
By accident. I moved into a hacker house in San Francisco in 2013, and the guy who became my co-founder was sleeping on a bunk in the next room. We were both in startups, he had been working in machine learning for years, and we decided to build a company together.
Artificial intelligence is basically a set of statistical models and logic gates that make good predictions about data. AI algorithms are becoming more sophisticated and accurate really quickly on an expanding range of tasks: including everything from winning at board games to steering autonomous vehicles. Those algorithms are also getting better at predicting human behavior, and they will affect how we live and behave. They can see us click through pages on the Internet and learn about our preferences. That’s going to make the data that companies gather more powerful and prescient.
When companies wanted to analyze data, they had a choice being accurate, expensive and rare human experts, or less accurate, less expensive and infinitely replicable algorithms. Now those infinitely replicable algorithms are more accurate than humans on a lot of problem sets. That chokepoint no longer exists, and companies face a new landscape of incentives. These algorithms will help them create new products and perform existing operations more efficiently.
We were inspired to found an open-source software company by Red Hat and the many excellent open-source software tools that developers use every day. (Open-sourcing code makes it human readable. That means your users can also become contributors to your code base, and help improve it. That’s one reason why an open-source operating system like Linux won over Windows.)
AI and robotics are going to combine to create more functional machines. That’s going to affect primary care, manufacturing and warfare. AI gives robots faculties like machine vision or the ability to respond to some human speech.
We’re going to keep on hiring really smart people wherever we can find them in the world. For us, marketing usually happens when we explain complicated new technology is a clear and simple way. That’s what drives people to our website, so we will continue to do that. Marketing and product development are intertwined, so the great people we hire will be focused on creating a product that makes AI easy to use and powerful.
Most moments in business are difficult, frankly, because there’s always a problem to solve, and most of them are hard problems: How do we build the right product? How do we convince customers that it’s the right product in the face of fierce competition? How do I raise venture capital? How do I hire great people? How do I fire someone who isn’t a good fit? Those are all pretty hard problems, and you have to solve them under pressure and do the right thing both for the company and all its stake holders. Learning how to sell something is hard, and one of the hard things about it is the amount of knowledge you need to be an authority on the subject, someone that customers can trust. Selling isn’t just convincing someone to spend money, it’s making sure that your product is the right fit for them, and steering them elsewhere if it’s not.
Our ideal client is a large corporation, and within that corporation a principle software architect or an executive in a line of business. Those people have problems to solve, and when those problems involve predictive analytics, deep learning can give them better results than almost any other algorithm. So the ideal experience is for the client to try our open-source software, tell us about the problem they want to solve, and then work with us to build a solution using their data. What we do then is we support that solution long term. Open-source software companies are like insurance companies: we sell warranties on our own code. So the ideal experience is for the software never to break, but if it does, to have it backed up and repaired quickly by its creators.
Clear communication. You need to be able to explain to people where the company is going and why in plain language without business-speak. You need to cut the crap, the doublespeak. You need to listen to them, trust them increasingly as they show what they can do and give them autonomy wherever possible. That’s very motivating. Startup people like the ones who join us are self-starters. They don’t need a lot of micro-management. Their job description and the skills they need keep evolving as the company grows. So we offer them the chance to work with a small, talented team at a growing company on a set of hard and important problems, and when the solutions for those problems are completed, the world will be changed. This is work with impact. It’s open-source AI.
Learning how to communicate about AI, how to explain the way it works and the limits to what it can do, is really important. There’s a lot of confusion and misunderstanding about AI, even more than other technologies, and the people who leave their listeners and readers with a feeling of clarity, epiphany, understanding, are rare. Way too rare. Anyone who can do that work of understanding the tech and translating it for intelligent outsiders who don’t know the jargon will go far.