What you see virtually could be real. This is a myth, says Igor Susmelj of ETH spin-off mirage Technologies. He points this at the increasingly high volume of manipulated content that claim to showing genuine information.
This very idea of filtering myths and reality on social media is the basis for his start-up. It was the recognition at the HackZurich competition in 2018 that paved the way for three engineering students. They designed a web application that can detect fake videos. After extensively traveling across the U.S. meeting experts and potential customers, they realized that this business model would not make money. Hence, they tweaked it a little.
Pre-programmed AI modules
Susmelj along with his colleague Heiki Riesenkampf together founded Mirage Technologies in September 2018. The idea is similar to the machine learning concept they presented in Deepbusters. Further, it claims to provide new deep learning methods that are user friendly. As a result, the company provides developers with trained and pre-programmed ‘rockets’. These rockets have specific functions and are segregated into different families.
Mirage trains its rockets using existing data. On the other hand, if there is no data, then Mirage uses transfer learning to input rockets with additional data. Further, Susmelj says that they do not have to rely much on data as they do not have to train their rockets from scratch. All that users need to do is use a couple of coding lines in their choice of programming language to set it off.
Why major companies are ahead of the curve
Tech giants like Google and Microsoft are ahead in the league. They have been in the business for long and have access to enormous volumes of data. This helps them understand the needs and train their models accordingly.