Permutable’s co-founders Wilson Chan and Alexandr Medvedev gathered together with other thought leaders as part of Open Ocean‘s Data Series programme. The exclusive round table also featured other key thought leaders in the field, addressing the exciting advancements, challenges presented and innovative new approaches in the small data revolution, as well as predictions about what we can expect further down the line.
Over the past few years exciting progress has been made in the industry with regards to the use of small data sets – as opposed to the “big data” sets that machine learning has become synonymous with.
Across the industry, there has been a clear advancement in the design of systems able to use data sets more effectively by using real-time human feedback. Helping to capture this cognitive data which is often locked up inside our heads is fundamental to addressing big data complacency within machine learning which has become so prevalent.
Co-founder Alexandr Medvedev commented, “Due to the specifics of our clients in FinTech, our products allow us to create pattern recognition engines based on five to 50 examples. We originally started to investigate this problem because a trader’s time is quite expensive. By using human feedback loops we are able to substitute large amounts of required training data with expertise domain user knowledge.”
He added, “We are all very excited about whether it’s possible to gain a human level speed of learning, especially given the energy and resources being funneled into research at the moment, particularly in the field of Natural Language Processing.”
Founder Wilson Chan observed, “One of the things that we found is when we went to meetings with AI companies is that most companies focus on models which is the back end while they assume they have all the model data required. But what they don’t assume is how do you transition the knowledge from the domain user, and get the cognitive data back into the model in the most efficient way possible via a real-time feedback loop.”
At Permutable, we focus as much on the front end as we do on the machine learning models. By using a continuous human feedback loop, we achieve real-time data labeling. We also use an algorithm like zero-shot learning, which can learn from as little as one example. This presents an exciting opportunity for smaller companies to take advantage of the benefits of artificial intelligence without having to go through a large data harvesting exercise.
For more insights into how our AI technology can transform your business why not arrange a call to talk with Permutable today.