AiThority Interview with Chris Nicholson, Co-Founder and CEO at Skymind

Know My Company

How did you start your journey into Artificial Intelligence? What made you co-found Skymind?

It was an accident. I moved into a hacker house in San Francisco in 2013 (just like in the HBO series), and met my co-founder, who was already focused on Machine Learning. I had covered AI as a reporter, and I was working at a Sequoia-backed startup at the time, learning all the hard lessons of hypergrowth. So we teamed up. He would build the product and I would build the company. We were both fascinated by it, and we knew AI would be huge.

Read More: AiThority Interview With Rick Halton, VP of Product and Marketing, Lumeon

What is the role of the AI products at Skymind?

Skymind builds AI software that solves infrastructure problems for data scientists and Machine Learning engineers. We built the most popular AI tool for Java, Deep Learning, and we help build Keras. A lot of people generate ML models in Python and need to deploy them on the JVM. They need to do distributed training on Spark. They need to pull data out of Hadoop, or connect to Kafka, or store their results in Cassandra. Those are all JVM big data tools, and we make ML “just work” with that stack so it can save teams a lot of time with boring integrations. We can promise companies that they will get to insight faster, and deploy their AI models to production faster. We have a special focus on simulation as well as natural language processing.

How is Skymind different from other service providers in the same landscape?

Skymind is the only startup in the world to have built a major AI framework. We’re in the ring with giants. That means we control our bits all the way down to the silicon, which gives us tremendous freedom to innovate and respond to our customers and users. We’re not waiting for Google to give us permission, to fix a bug in TensorFlow or to add a new op. Most AI startups are just bundling someone else’s open-source work and slapping together a marketing web site. A recent study in Europe showed that only 40% of “AI startups” have anything to do with AI at all. A few are truly original in their approach, totally authentic in their commitment to hard problems, and deeply talented. We are in the latter group.

Read More: AiThority Interview With Tony Pepper, CEO and Co-Founder, Egress

How do you see the raging trend of including ‘AI in everything’ impacting businesses?

Achieving real success in AI takes commitment, patience, and a talented team. Near term: a lot of companies haven’t spent the time to formulate a strong use case with a clear ROI and to gather the right data for the predictions they want to make. They need to do both of those first. And they need to be thinking about AI from the very beginning when they start gathering the data. AI starts with the data. It shouldn’t be an afterthought. Long term: people will get more familiar with what AI is, and what they need to make sure it succeeds. Electricity took many decades from the moment of its discovery before its use became widespread. We haven’t seen half the things AI can do yet.

What are the biggest challenges and opportunities for Artificial Intelligence companies in dealing with rising technology prices?

The biggest challenges for AI companies are obtaining the right data, hiring the right talent to analyze it, and then breaking through the hype to prove that their technology stands apart. It’s easy for people to get confused about tech. There are so much jargon and so many new companies all the time.

Read More: AiThority Interview With Cyril Perducat, EVP IoT and Digital Offers, Schneider Electric

What makes deploying Machine Learning/Deep Learning capabilities so hard?

ML/DL is new, and it’s a new kind of workload. IT doesn’t speak that language and doesn’t know what to do with the profusion of new tools. What’s a Machine Learning model? How do you monitor its performance? Data scientists know that, but the people in charge of production stacks are still learning to speak that language and master those tools. Data science and other engineering teams have to really try to understand each other’s concerns, and to take each other’s problems seriously.

Where do you see AI/Machine learning and other smart technologies heading beyond 2025?

Nobody knows, not even Geoff Hinton. That’s just too far out. I can barely tell where we’ll be next year, the field is moving so fast. The best minds in the world are working on this, and enormous amounts of capital are flowing in to back them. People outside of AI are shocked by what AI can do now, but they’re usually shocked by yesterday’s news. If they knew how fast it’s moving, they would be speechless.

Read More: AiThority Interview With Don Zereski, Vice President of Engineering, Indico

What is your opinion on “Weaponization of AI”?

AI has already been weaponized and humans will continue to find new ways to weaponize it. In that sense, it’s like any other technology. The internal combustion engine is used in minivans as well as tanks. Nuclear fission can help cities keep the lights on, or raze them to the ground. Humans weaponize all technology, which inevitably makes each new invention a double-edged sword, so to speak. The same governments that we are asking to regulate AI are the most advanced in its weaponization. There’s no irony in this: It’s just the state’s long-running legal monopoly on violence, applied to AI.

What start-ups and labs are you keenly following?

I follow the well-known labs that everyone does: DeepMind, OpenAI, BAIR, FAIR… There are too many startups to count.

Read More: AiThority Interview With Jim Scott, Director, Enterprise Architecture, MapR

What technologies within AI and computing are you interested in?

I’m interested in Meta-Learning, the research into algorithms learning how to learn. That’s one step back from solving a specific problem and is pointing the way toward solving lots of problems. I’m interested in computational creativity, all the work being done by people like Robbie Barratt and Mario Klingemann to create novel images. I’m interested in self-supervised learning, or teaching algorithms about the world even when you don’t have “answers.” I’m interested in how deep neural networks can be combined with symbolic reasoning to extract knowledge from data in flexible and expressive ways. I’m interested in natural language understanding and having an algorithm that I can finally have a decent conversation with. It hasn’t happened yet and it seems like it’s still a long way off.

Could you share a piece of professional advice that changed your life?

It’s better to sense and seize an opportunity than to follow your passion. Opportunity is rare, and passions are common. Artists and musicians starve because it’s really easy to enjoy making art and music. Making music is so fun that lots of people will do it for free. If you seize an opportunity, very often you can work your passion in later, or at least bankroll it.

Read More: AiThority Interview With Stefan Nandzik, VP, Corporate Communications, Signifyd

Tag the one person in the industry whose answers to these questions you would love to read.

Chelsea Finn at Berkeley/Google Brain.

Thank you, Chris! That was fun and hope to see you back on AiThority soon.