Halfpoint - Fotolia

Google's Eric Schmidt on digital success -- be first, perfect comes later

Success for digital products doesn't follow the same trajectory as for physical goods. Google's Eric Schmidt laid out the case for a rush-to-market approach at an MIT event on AI.

In the fast-paced business environment of IT development, there is little time to add finishing touches, and getting a product to market quickly should be the priority, according to Google's Eric Schmidt.

"The world is so competitive now that if you think you have enough time to quote 'fully finish' your product, you're going to lose," the former executive chairman and CEO of Google said Thursday before an audience at the Massachusetts Institute of Technology.

"What I tell people when they start a company is," Schmidt added, 'You have to figure out who started the same company the same week and figure out how you're going to compete with them.'"

Few have achieved the level of success that Schmidt has in his field, and for about a half hour he imparted some of the wisdom picked up over his career to an audience of investors, entrepreneurs and educators at Thursday's "AI and the Future of Work" conference.

'Power-law distribution'

Gaining an advantage over competitors is of supreme importance in the digital marketplace, where the spoils for the market leaders tend to be many times greater than lesser competitors. Indeed, in the technology sector, being first matters hugely, according to Erik Brynjolfsson, director of the MIT Initiative on the Digital Economy and a MIT Sloan School of Management professor.

Unlike other sectors of the economy where companies' market shares are closer to one another, the most successful digital products tend to eclipse their lower-tier competitors.

MIT, Google, AI and the future of Work, machine learning, power of distribution
Erik Brynjolfsson and Google's Eric Schmidt discuss how the market approaches for digital products differ from those for physical goods at the 'AI and the Future of Work' event at MIT.

"Digital goods tend to follow the power-law distribution," Brynjolfsson said, using a statistical term for the way the top players in a digital marketplace can dominate. In contrast, "physical services or physical goods tend to follow more of a normal distribution, like a bell curve," he said.

Schmidt said he first learned about that principle from someone at Microsoft 30 years ago.

"We were talking about how the PC market share was shaping, and it was, in his view -- 90%, 9% and .9%," Schmidt said. "I said that's not true. Market shares are always 60-40. He was the first person to me that said, 'This thing has a different shape.'"

Continuous improvement

After a product is introduced, engineers should begin measuring its success and working on improvements, using real-world data gleaned from the product itself, Schmidt said.

The world is so competitive now that if you think you have enough time to quote 'fully finish' your product, you're going to lose.
Eric Schmidtformer executive chairman and CEO, Google

"If you're connected to everyone all the time, you should be able to get instantaneous feedback all the time," Schmidt said. "A correctly engineered product of any kind is giving telemetry and so forth about its use and then strong engineering teams are watching and saying, 'That was a bad idea. That was a good idea.'"

When Schmidt became Google's CEO at the turn of the century, he was surprised by the company's somewhat unstructured but urgent approach toward product development, which he said is now the norm in the industry.

"The thing that was most bizarre to me joining Google was that it didn't have a normal project schedule," Schmidt said. He said, "We had a rule that everything would come out in beta because we didn't want to defend some of the mistakes."

Marissa Mayer effect

Schmidt credited Marissa Mayer, who headed up Google's user interface program and then led Yahoo, with measuring how actual users responded to products.

"Her basic theory was we couldn't be sure about what the right UI was, but we could certainly test it," Schmidt said. He said, "This measurement concept is a key to product success in every industry."

To learn more about the rush-to-market approach, Schmidt recommended the audience read "Blitzscaling: The Lightning-Fast Path to Building Massively Valuable Companies" by Reid Hoffman and Chris Yeh.

But acting with urgency does not mean throwing caution to the wind, Schmidt said, and artificial intelligence and machine learning systems often need human direction or oversight.

Ryan Robidoux, who teaches computer science at Somerset Berkley Regional High School in southeastern Massachusetts, asked Schmidt about the limits of the strategy.

"Machine learning and AI is going to be in all these kinds of fields, especially ones that affect finances and the health and safety of people," Robidoux said. "Would you still recommend this idea of blitzscaling even though these systems may have a negative effect on the safety of people?"

"Blitzscaling is a comment about how you would run your company, not necessarily how you would build an AI-based system for a life-critical operation. So there are still significant errors as you know in machine learning systems that we tolerate, so they're advisory," Schmidt responded.

 "You wouldn't want to have a pure machine learning system in a life-critical situation -- flying the airplane or whatever -- you want it to be advisory," he said.

AI-first strategies

Google is owned by parent company Alphabet, where Schmidt remains a technical advisor. Technology companies have already switched to a mobile-first development strategy that prioritizes how products will work on smartphones, and he said companies are embarking on AI-first strategies to incorporate machine learning into all software.

Consumers have grown used to machine learning's integration into products like Netflix, where it recommends movies and TV shows to the individual viewer, Schmidt said.

"One of the things that subtly happened was we all started using these recommendation engines, which are machine learned, without really thinking about it," Schmidt said. "And why did we use it? Because it made our experiences better."

Schmidt said that machine learning can help map out more efficient processes in the enterprise that are invisible to human analysts. Google improved its energy efficiency by 15% by using machine learning to study patterns in the data centers, Schmidt reminded the audience.

"What we know from machine learning is there are subtle patterns that computers can detect in the training data that we just don't see," Schmidt said.

Dig Deeper on Digital transformation

Cloud Computing
Mobile Computing
Data Center
and ESG