CMU starts virtual panel series on the future of AI

Artificial intelligence (AI) is starting to permeate every sector of the U.S. economy. According to the National Security Commission on Artificial Intelligence (NSCAI) in their March 2021 “The Final Report,” it will "transform virtually every aspect of our existence." Carnegie Mellon has partnered with the NSCAI in a virtual panel series starting this year discussing artificial intelligence's impact on "economy, national security, and welfare."

The panel series invited a variety of speakers from Carnegie Mellon, Pittsburgh, and Pennsylvania robotics and AI institutions. The discussions so far have revolved around robotics, semiconductor production, Pittsburgh industries, and other ways the U.S. can use artificial intelligence to improve domestic competition.

"We've had an unfortunate narrative in the public where people have akin artificial intelligence to Terminator and Skynet, and that's not the case," NSCAI Commissioner Katharina McFarland said during the first panel discussion. "This technology is here to stay, and you'll see in everything that you read, the ubiquitous nature of this can be a challenge for us, or it can be an opportunity."

The first panel discussion on Aug. 18 was "AI and the Future of Manufacturing: A Synergetic Relationship," which discussed how artificial intelligence has begun integrating into supply chains. Gary Fedder from Carnegie Mellon's Manufacturing Futures Institute guided the conversation, which addressed many possibilities for AI to be involved in the supply chain, including in semiconductors and in robots that fuel many manufacturing processes. In the opening remarks, McFarland framed the discussion by mentioning insights from the NSCAI report, such as how national security is tied to economic strength and that AI could improve U.S. economic strength by improving manufacturing capabilities.

Robotics has been widely integrated into the manufacturing process, and the panel discussed different ways that AI could improve these robotics. Ira Moskowitz from the Advanced Robotics for Manufacturing Institute mentioned the various ways that AI can be applied to robotics, such as predicting human intent or developing simulations for robots in situations where there is limited real-life data. Elizabeth Ann Holm from the Air Force Center of Excellence on Data-Driven Discovery of Optimized Multifunctional Material Systems mentioned how AI could support the move to agile manufacturing, prioritizing customizability over scale, as providing customizability requires finding patterns from accumulation of data which AI excels at. Carnegie Mellon alumnus Jared Glover from CapSen Robotics also mentioned that right now, robots often have to be programmed for very specific tasks such as moving each individual finger, so robots cannot easily be repurposed for different tasks.

"When we talk about 'artificial intelligence,' intelligence is sort of with a lowercase 'i' right now," Glover said. "It's definitely not at human-level intelligence, no matter how you measure it."

Semiconductors are also a key part of the manufacturing process, powering robots and other necessary machinery. Shailesh Patkar from semiconductor company II-VI Incorporated mentioned how his company is using AI for inspecting chips and improving product efficiency and performance. Moskowitz added that since the semiconductor industry already runs so well, the only aspects left to improve are more complex, and AI excels at more complex operations. Holm also mentioned how the industry is shifting away from large, monolithic manufacturers to smaller manufacturers. Glover, however, pointed out that the adoption of these technologies have been slower in smaller manufacturers because of high costs, unfamiliarity with the technology, and difficulty repurposing robots.

The discussion then shifted to how the U.S. could invest in educating both workers and students in related industries. Holm also mentioned how, in order to support these supply chains, there must also be enough students in data science and AI to meet these demands. McFarland, Moskowitz, and Fedder added that the unique nature of the COVID-19 situation allows us to think about more hybrid, non-traditional classroom approaches to training technologists.