Artificial intelligence is no longer just a growing fad in computer gaming and an experiment in fields including industrial manufacturing. Instead, it is also a still-developing technology that continues to see new uses in e-commerce and mobile apps creation as software developers find new ways to use the power of AI in innovative and useful ways.
One believer in the use of AI in a growing number of mobile apps in the future is Eyal Lanxner, the CTO and co-founder of Feedvisor, a Tel Aviv, Israel-based vendor which uses algorithms to constantly re-price merchandise for e-commerce marketplace sellers so their goods are properly priced against competitors.
By using machine learning, AI and data science techniques, Feedvisor helps e-commerce sellers to ensure that their product pricing is always competitive in real time around the globe, Lanxner told ITPro.com.
"We try to do it in a more algorithmic approach than competitors," he said. When buyers on sites such as Amazon.com search for a product to buy on the website, they get results and listings from a variety of sellers each offering the item at different prices. One such seller is also chosen to be a selected vendor to show off the item on the site. That's where Feedvisor uses AI to help level the market for sellers.
"You must make sure that you are getting exposure in that [selected vendor search page position] while also making sure you are not cutting your prices too low," said Lanxner.
Feedvisor's use of AI in its algorithmic pricing services is an important example of the idea that AI is more than just computer vision, which is probably the most common form of AI, he said.
For app developers, the possibilities of AI are still wide open, he said, but the promise starts with knowing what they will want to do with it and how they can solve specific app challenges for businesses with the technology.
"Developers will be able to solve more business problems with AI, but they have to have an idea of what they want it to do with it" before they even get started, he said. That can mean adding specific capabilities to apps such as being able to use AI to measure a person's clothing sizes by analyzing a 2D photograph of them or creating other new outside-the-box thinking about how to solve other difficult problems.
"If you want to solve things correctly with AI, you need to understand AI and its possibilities," he explained. "Computer vision is one thing to understand, but also segmentation, identifying anomalies, clustering groups of objects" and more. "AI or machine learning is a toolbox, but you need to first understand the problem" you are trying to solve.
At the same time, apps developers also have to think about what can go wrong when they are using AI, he said. In a previous job, he was working for a data security company that had a banking client which wanted a fraud detection application to help protect it from attackers.
The bank gave the security company a long list of transactions, which needed to be sorted into legitimate transactions and fraudulent transactions by the application. The first version of the application, however, had a false positive rate of 1.5 percent, which was way too high for the bank to accept. The client instead demanded a false positive rate of no more than 0.1 percent.
But instead of using real data, the security company used simulated data to run its application and the algorithm they used produced data that was way off the mark, said Lanxner.
So what was the lesson learned from that early AI experiment?
"You can't simulate data," he said. "You have to have real-life data. Until we had that, we really couldn't provide the value that we wanted to provide."
Lanxner said he expects to see more developers in the future using AI to build better e-commerce recommendation systems within applications, especially if improvements can be made to keep refining Amazon's algorithms.
One major driver for the coming AI innovations in app development is the availability and continuing growth of computer power to run and develop the complex algorithms that will make the ongoing advances possible, he said.
"We have much more computer power and now much more data" around the world, said Lanxner. "Those changes made AI much more possible."