“We know that the devil is in the details. So we spent months and months choosing the right industry and sub-industry where we can solve a big problem with reinforcement learning”.
To predict the future with AI and deep learning, Roy Cohen and Uri Yerushalmi helped create Fetcherr, an AI-powered price intelligence engine that can predict a price on anything. We had the opportunity to pick their minds and take away some great insights.
Promising Potential In The Airline Industry
Yuri: “The airline industry is quite interesting in the sense that it has competition. It’s very competitive… Your competitor sometimes matches you. Sometimes there’s a time lag. These things are very interesting to be taken into account in reinforcement learning systems, because you don’t look only at the next step, you look several steps away. It’s like a chess game”.
Yuri: “the current traditional systems, because of the complexity of, of the, the market, they took it to a totally different direction… The current situation in the market was very far from what the technology currently allows”.
Roy: “It’s very traditional. I’m talking about the infrastructure side, like banking and Swift. Nobody’s brave enough to change Swift. It is still completely manual. So we also investigated and we interviewed almost 60 airlines to give you a lens of how much we spent.
“We interviewed major hotel chains. We interviewed insurance companies. We interviewed high officials in banking. We interviewed car accidents to predict the possibility of a car accident… So a goal-based system approach, or a glass-box AI approach for full transparency within the organisation destroys all these siloed departments… And creates for the entire stakeholder real-time transparency. So the CEO can see what each flight his organisation can do and not wait for the end of quarter report”.
Removing The Trivialities Out of AI
Yuri: “From our perspective, the airline does not need to understand anything in neural networks or AI or the data… From the customer’s perspective, they don’t care whether we are using AI or not. They have a goal. It’s a goal-based system. And their goal is a lift-up of revenue or gaining some market share”.
“Piggybacking on the current data structures, as is. No need to change anything in their IT systems and showing them what is going to be the effect of every change. We show the customer this specific area should be raised by, for example, 2%. And we also explain the reason. ‘That’s because the COVID situation has changed, or your competitor did x and y and z. And because of that, we think that you should lower your price.’ From the customer’s perspective, it’s to understand why that is happening”.
More on Deep Learning & Price Predictions
This episode is not to be missed!
Give it a listen. Share your thoughts with us. And enjoy!