A new technology is there to allow machines to gather knowledge from less dataset and recreate more knowledge from the new data fed to them. As we know, today AI systems use deep-learning algorithms to acquire knowledge from a data set and then they practically apply this knowledge themselves.
This new technology is aimed towards reducing this tedious method. Through Gamalon, a Boston company, this technology has been applied on two of their products. Gamalon states that the technique is based on Bayesian program synthesis, a mathematical framework named after Thomas Bayes who was a mathematician in the 18th Century. They use probabilistic programming where a code uses probability instead of the specification. This is based on Bayesian probability since predictions about the surrounding are made from experience.
On an interview with MIT Technology review, Gamalon CEO and co-founder showed a demo which was a drawing app that uses this new technique. This app relies on probabilistic programming to identify whatever a user is trying to draw unlike a similar app developed by Google that uses previous or known sketches to predict objects being drawn. In this way, the Gamalon app makes correct predictions even if the drawn patterns are unfamiliar to these.
Gamalon also released two apps, Gamalon Structure and Gamalon Match, to prove that their technology can be applied commercially. Gamalon Structure uses this technique to recognize various approaches from just raw text. For example, by just inputting the manufacturer’s description of a product to it, it is able to determine its brand, product name, and other features.
Gamalon CEO also stated that there are many other areas to apply this technique. For example, when a smartphone or a laptop is setup with this technique, they wouldn’t share confidential user information with companies in order to determine user preferences. But this can be done through probabilistic calculations within these devices. This technique could also help the automatic or self-driven cars adapt to their environment.