The easiest way to train your algorithm. Machine learning made by humans
One of the most revolutionary technologies currently being developed is artificial intelligence. It is a collective noun for algorithms that can perform human tasks, which can change entire businesses. In order to train these algorithms, you need data that is well-labeled. Labeling is not a simple task and requires human involvement. Amazon developed Mechanical Turk, a ‘marketplace for work that requires human intelligence’, and new players like Dbrain are searching for solutions too.
The Ghent-based RoboVision - which uses machine learning in agriculture, manufacturing and media - developed Alltag. This platform enables employees to help ‘training’ the AI models by labelling data. Until now, this was done via a browser on the computer, which is a very labor-intensive process. RoboVision organizes sessions, in which people are guided in the labeling of the data.
To simplify this in the future, In The Pocket and RoboVision developed a unique mobile platform. People no longer have to work with the computer mouse, but use touch to label data. User testing has taught us that this is a much more natural way of working.
"Machine learning and AI will have an enormous impact on our lives and on how we work. The data economy can bring with it enormous benefits but only if the human factor is taken into due consideration in the development of machine learning platforms. We feel that a partnership made a lot of sense; reinforcing our AI knowhow with In The Pocket's digital product and platform expertise"
Jonathan Berte, RoboVision
The application is currently being used to label plants and computer chips. With enough data correctly labelled, the algorithm can be trained accurately. Alltag still has a lot of untapped potential. For example, it might be interesting to let users know if they’re working accurately or not, or even pay them directly when they’re using Alltag to train the algorithm.
Alltag’s product vision is unique in the sense that we are building a decentralised platform to guarantee that the people working on it are treated fairly and receive fair pay. With that in mind we approached Vluchtelingenwerk Vlaanderen. Refugees are ideal beta testers for this product because the market is very international. And the actual work of structuring data is easy-to-learn, language-independent labour that can be done remotely. With the support of Vluchtelingenwerk Vlaanderen we recruited refugees to help with the data labelling.
The collaboration between In The Pocket and RoboVision resulted in the Roadmap to AI, a 4-step program that aims to uncover potential use cases with artificial intelligence, build a first pilot and deliver a clear roadmap as a final result.