This was the provocative question posed at the 2017 Product Realization Group symposium which probes and engages Silicon Valley’s hardware ecosystem. Artificial Intelligence, or “AI” to insiders, was once relegated to sci-fi films. And although it’s not mainstream yet, there is no denying its potential impact on the hardware industry.

Keynote speaker Nahid Sidhki from SRI’s Robotics Center set the framework by describing mind-boggling applications. My favorite – “micro robots” – inspired by ant farms that will take 3-D printers to the next level for on-demand parts generation.

However, the question on everyone’s minds right now is how AI is shifting hardware products. Companies are trying to figure out their AI strategy as well as understand the broader implications of its use.

To sort through these questions, an expert panel provided their insights, including Greg Reichow, former VP of manufacturing at Tesla and now with the VC, Eclipse Ventures. Reichow described the two sides of the Tesla factory – half of the facility is highly automated and the other half is reliant on “very manual labor” for details like trim work.

Mark Jacobstein, the Chief User Engagement Officer at Guardant Health, expressed some of the limitations of AI. In “data starved” organizations, machine learning can be very challenging.

“Progress is possible when data sets start working together in a more structured fashion,” Jacobstein said.

Currently, he described the condition as “primitive.” Other reasons cited for slow adoption of AI include:

  • Investor Reticence
  • Regulatory Risk
  • Weak Infrastructure
  • Siloed Experts

Rhadika Dirks, Partner at, noted that products manufactured today are built with technology from five years ago, not for five years from now. Reichow talked about how the “speed of development is hindered by the speed of feedback cycles.”

One central question was whether or not hardware is still catching up to software. Some like Dirks say that the robots have caught up, and it’s the apps that haven’t kept pace. “Hardware can be better optimized,” she said. However, Dirks also acknowledged that AI for the future is waiting on the hardware. Jacobstein said that for healthcare, hardware is the laggard, but it’s not far away. Reichow offered that the real challenge is in developing algorithms, or quantum computing.

And since that put the discussion well out of reach for most businesses, the discussion turned to panelist predictions on moonshots. Here’s an unattributed list:

  • Brain-to-Brain communication and Brain-to-Machine interface
  • Machine learning on top of real world data (applied to clinical trials)
  • Genome mapping that’s forward looking (i.e., cognitive learning assists based on personal background)
  • AI-directed evolution (i.e., cure birth defects by “knocking out” particular genes/enzymes)

Finally, for companies that are thinking about their “AI Strategy,” panelists had the following advice. Dirks recommended that if AI isn’t your core business, use a vendor instead of rebuilding your team internally. Jacobstein echoed this statement, saying, “You’ll recruit the wrong people, or the problem will change after you hire.” And Reichow recommended that businesses forget about an AI strategy And should instead explore how to solve their customer’s problem. From there, businesses can figure out the right tool.