Humanloop raises funding to teach AI to learn from humans faster
Humanloop, a game changing UCL spinout, has attracted investment from UCL’s Technology Fund (UCLTF). Based on the world-leading research of two notable academics at UCL’s AI Centre Humanloop is teaching AI to improve performance with 10 times less human training than existing systems.
The UCLTF has invested in Humanloop to help the team, which includes Professors David Barber and Emine Yilmaz, to build software that can train AI models to reach high levels of accuracy while using smaller datasets through better interaction with their human users.
The potential of AI is still far from being realised. Current AI performance is severely limited by the volume of annotated training data required to teach systems. Whereas humans can learn new concepts from just one or two examples, most AI systems require thousands of examples to reach adequate levels of performance. Solving this problem is a growing priority for the AI community as data labelling is time-consuming, expensive and a bottleneck in the adoption of AI. Humanloop’s technology enables AI systems to query human experts on only the most valuable data and learn much faster.
Humanloop’s team has a unique mix of leading AI academics and experienced start-up founders. David Barber, Professor of Machine Learning and Director of UCL AI Centre and Professor Emine Yilmaz, an expert in intelligent sampling who has worked for Amazon Alexa and Microsoft Cortana, are joined by three exceptional co-founders. Raza Habib and Peter Hayes are UCL PhD students with previous start-up experience. They are joined by Jordan Burgess, a machine learning engineer who was previously at Amazon and Bloomsbury AI, a UCLB spinout also supported by UCLTF investment which agreed to join Facebook’s London research team in July 2018.
Professor David Barber, Co-Founder of Humanloop and Director of the UCL AI Centre, said: “What makes Humanloop’s systems different is that we use research developed at the UCL AI Centre to make the system learn a lot faster. Our models also have well-calibrated uncertainty so when they are not confident they defer to a human, working alongside people and augmenting them, rather than trying to replace them. UCL Technology Fund’s vital investment means we can put our research into practice and take this exciting vision forward.”
David Grimm, Investment Director, UCL Technology Fund, said: “This initial investment allows us to support a highly talented team to create what could be a game changing technology for improving performance in AI. We’re excited to be joining them right at the beginning of their journey and look forward to supporting them as they grow. Our fund exists to invest in the best ideas emerging from UCL’s world-leading research community so they can go on to have a societal and market impact, and AI is an area where UCL is truly amongst the very best.”