Everybody’s talking about machine learning and artificial intelligence right now, but the real conversation needs to be about how to move business processes based on these techniques through proofs of concept and into production. To do that requires two things: a proper understanding of the right techniques and tools and an ability to build an operational solution that works within the organisation’s existing IT environment.
Until this month, Brooklyn-based research and advisory company Fast Forward Labs sat firmly in the first of these boxes. “We aim to be publishing research six months to two years ahead of when specialist technology will be useful in a deployable solution,” explained founder and data scientist Hilary Mason in a phone call with DataIQ from Helsinki, where she was due to speak to a telco conference about how networks can drive more value from their high-volume data assets.
What changed in September was the acquisition of her business by machine learning, analytics and data management vendor Cloudera. “It is a natural fit because of their ability to build and deploy solutions for the data of today and the data of the future,” explained Mason. It also means that her consultancy now has the platform onto which clients can migrate machine learning models once they have demonstrated their potential.
While Mason says that, “we are going to continue the slow, linear growth we have been doing since 2014,” there is no doubting that the ML and AI space has hotted up this year. Like big data back in 2012, it is fast becoming a must-have among companies wanting to deploy some disruptive technology and a must-do for data scientists who see it as a chance to prove their mettle.
Mason is more cautious about the potential. “The media coverage gets quite frightening and rarely focuses on the reality. It sees AI as a magic box that absolves companies of their responsibilities,” she said. Initiatives like the UK’s Data Ethics Committee, due to launch this Autumn, are part of efforts to set boundaries around what machines are set up to do and who is liable for what they end up doing.
“This is the beginning of a movement and a change in the way people think about data-driven product development.”
Understanding how these new techniques and tools will impact on business - and their employees, as well as society more broadly - is not easy because of the scientific nature of the practice. “It is not like software engineering where you know the thing you are setting out to build and that it is possible before you start, even if you don’t know how it will turn out. With machine learning, you don’t know if it is possible to build a product. It is much more of a risk than software engineering and that is what people are getting nervous about,” explained Mason.
By integrating her research and advisory business with a proven technology provider, the issue of how to put ML models into production at least becomes more straightforward, since the engineers can be involved with the scientists from the outset. It may also help companies to identify where the true value is to be derived without the risk of giving away their IP, one of the key issues which has emerged around the Google Deep Mind and Royal Free data sharing.
“Data is the core of a business. I would never advise a company to give it away. But we are seeing cases where a vendor says it will take the data, build something from it and then want to keep the data,” she warned.
For now, the new Cloudera Fast Forward Labs finds itself in a hot spot from where it is well placed to advise clients on creating value without losing an asset, while also improving their chances of successful ML and AI delivery. As Mason said: “We are in an exciting place with machine learning - people are recognising its value now, which was not the case five years ago. This is the beginning of a movement and a change in the way people think about data-driven product development.”