There are many companies working on artificial intelligence. The vast majority of these are looking at how artificial intelligence systems can recognise things and tell them apart – differentiate between an apple and a mango for example. But what do these artificial intelligence systems do with that information once they’ve perfected a method of gathering and defining it?
This is where decision trees come in. Decision trees are used to map various consequences and possible actions an artificial intelligence system can take. For example in an autonomous driving system, if sensors were to recognise an object in the road ahead, a decision tree could include options such as slow down, emergency stop or swerve. One of these would then be chosen based on the information gathered and the perceived outcomes.
And while many companies are looking at the best ways to gather that kind of information, no-one else – as far as can be determined – is currently looking at ways to instil machine learning and automation into the decision making process through the introduction of machine learning algorithms.
Prowler.io, a new UK startup launched earlier this year, believes it is the first to do so. The company recently closed a £1.5 million seed funding round supported by noted tech venture capitalists such as Amadeus Capital.
Prowler.io will use that money to develop a platform that can improve artificial intelligence decision-making systems. If successful, it could develop a solution worth billions to the tech industry.
Currently companies in all areas linked to decision marking spend £10 -£15 billion creating decisions trees, says Vishal Chatrath, co-founder and chief executive officer (CEO) of Prowler.io. “Simplifying the process even by 50% could save billions,” he adds.
The startup is initially planning to work in the area of game development but Chatrath sees the base platform as something that could easily be used by a wide variety of sectors. Prowler.io is already working with three partners on an early stage proof of concept. He expects it to be ready by March 2017.
“A decision tree in any vertical can be replaced by our technology,” he says. “Anything that comes down to a fundamental decision making problem.”