Years ago, Landing.ai founder and former Google Brain researcher Andrew Ng famously declared that artificial intelligence is the “new electricity.” In short, AI will revolutionize the way all businesses will work in the future. But as more companies race to integrate AI into their operations, many are finding that it’s not as easy as they thought it would be.
At VentureBeat’s Transform 2019 conference in San Francisco, Landing.ai VP of transformation Dongyan Wang explained why companies seem to fail so often and the steps they need to take to make meaningful progress. He reiterated Ng’s electricity analogy, saying that when electricity was discovered more than 100 years ago, companies scrambled to figure out what it would mean for the survivability of their business.
Wang said there’s a similar sentiment now about AI in industries all around the world. An explosion of data and accessible computing power have made even non-internet companies interested in AI. The number of AI-related jobs has increased significantly, as has the number of research papers being written on the subject.
“This is the third time AI has really come around. And we believe that this time, this is the real deal,” Wang added. “We’re going to see AI being adopted in the real world and provide business value. We’re going to see that impact for the next 30, 50, or maybe even 100 years.”
Landing.ai has seen this firsthand with its customers. It works with companies who want to see how AI can help improve not just their bottom line, but also their processes. Wang said it takes about 18 to 24 months to understand his clients’ needs and help them develop an internal AI team and strategy.
He brought up an agricultural company in China that wanted to make its harvester machines collect crops autonomously. Landing.ai figured out that while AI could make these machines drive in a straight line or do simple turns on the field, it would take too much time and resources to design more complex behavior — like avoiding utility poles or even ancient tombs, which are common in rural fields in China.
“I’m not sure we want to build the largest data set of tombs and then do the best AI models to recognize these tombs so that the harvesters can go around them,” said Wang.
The idea also brushed up against one of Landing.ai’s basic tenets: If you’re just starting on AI, you should work on one or two smaller projects first to build confidence. So Wang’s team came up with an alternate solution — an AI assistant that would provide detailed information about the crops to the human drivers so that they can make better decisions.
Wang used the harvesting example to show that companies need to think carefully about what the right use cases might be for AI. Ideally, they should be small projects that you can execute within six to nine months. That’s the methodology Ng used back at Google Brain, where his team first worked on speech recognition and Google Maps before tackling the company’s core advertising business.
Once you have something in mind, the next step is to make sure that you’re using AI to automate tasks — like any sort of grueling, repetitive work — and not entire jobs. Wang said the goal isn’t to replace your workers; it’s to make them more efficient. His final piece of advice: Combine your subject matter expertise with that of AI experts so that you can figure out the right use cases for your business.
What you ultimately decide to work on in those crucial first months may make or break your company’s approach to AI.
“I want to emphasize that it’s very important to pick the right one or two projects and make sure you’re successful. Why is that? Because for a successful company — there are a lot of doubts over AI adoption and any new technology,” said Wang. “If you really fumble on the first one or two projects, it may take you a couple of years or even longer to recover and start again. But then you’ve lost that very valuable survival time for the transformation.”