Tech

Artificial Intelligence technology

Currently the cutting edge in AI research and in industrial applications builds on a method called deep learning, with impressive results. There are however still bottlenecks before truly intelligent, autonomous systems (for instance self-driving cars) are possible. One of the main problems currently is the need to teach the AI system, e.g. by labeling all data.

Unsupervised machine learning

Curious AI applies unsupervised learning methods to solve the bottlenecks in deep learning. Our work takes a lot of inspiration from the human brain, but adapted to silicon (think of the difference between flapping wings and the solution the Wright brothers came up with).

It is generally understood in the field of machine learning that unsupervised learning is key in the future. Curious AI merges unsupervised and supervised deep learning (technically, this is known as semi-supervised learning) for superior performance.

old-cogsUnsupervised learning really makes a difference in demanding applications since nobody is able to label, frame-by-frame, 10,000 hours of video recorded from a car, for example.

Our long-term plan is to build artificial general intelligence. Based on our already published research and results, we believe we can make a breakthrough.