Check out the on-demand sessions from the Low-Code/No-Code Summit to learn how to successfully innovate and achieve efficiency by upskilling and scaling citizen developers. Watch now.

Artificial intelligence software does some amazing things. It enables computers to identify brands in images, translates conversations between speakers of different languages in real time, and some hope it can predict movements in the stock market.

But whether you’re talking about especially powerful technologies like deep learning algorithms, or more general AI software, traditional approaches trump cutting edge solutions under certain conditions.

At a Transform 2018 breakout session in Mill Valley today, executives explained that when considering AI solutions, companies should question the cost-benefit ratio.

“Some things might be slightly less accurate but have better cost-benefit,” said Rahul Todkar, vice president of enterprise data science, research and marketing analytics at Charles Schwab.


Intelligent Security Summit

Learn the critical role of AI & ML in cybersecurity and industry specific case studies on December 8. Register for your free pass today.

Register Now

An artificial intelligence algorithm might be more accurate than less advanced software, but less advanced software might still be a better value because it is accurate enough — and at a much lower cost.

Keyword search is an area where a simpler approach can work better. That’s because, when people search a keyword, they expect to find something that matches the exact word, said Tom Pinckney, a vice president at eBay overseeing applied research on buyer experience. Deep learning would, in a sense, overthink it.

“Our search ranking is still basic decision trees and stuff like that because that just works better for search,” he said.

Sometimes the reason not to use an advanced technology has to do with regulations, not the technology’s fitness for a certain application. Todkar said the “black box” nature of deep learning, where software engineers can’t explain how or why an algorithm works the way it does, could create problems for Charles Schwab if the company used it in the wrong ways.

“That’s a big consideration for us, especially in the financial world with a lot of regulations, compliance,” he said.

Regulators aren’t the only people who might take issue with a result that a company can’t easily explain. Pinckney said eBay has to consider sellers who ask for explanations about why a competitor’s product is featured instead of theirs.

He also said you have to think about the impact of AI on employees in a situation where software and people are both tasked with the same job. eBay tries to avoid conflicts where “the humans and the machines are kind of arguing about who gets to make the decisions,” he said.

Impacts on people may be hard to measure, said Kaleido Insights analyst Jessica Groopman. Algorithms can influence human behavior in ways that don’t immediately or obviously show up in your bottom line.

“Measuring revenue impacts is not the end of the story,” said Groopman.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.