By AI Trends Staff
The Ford Motor Co. has made a substantial investment in AI, from investing $1 billion in Argo AI in 2017 to advance its self-driving car efforts, to developing centers of excellence to focus on machine learning and AI, where engineers determine the AI tools and methods that can be dispersed throughout the company.
The use of AI for predictive maintenance, anticipating when a part may fail before it does, is proving productive for manufacturing at Ford, according to Jeanne Magoulick, Advanced Manufacturing Manager, Ford Motor Co.. She spoke as a member of a panel on Shaping Technology Innovation at MIT’s recent AI and the Work of the Future Congress 2020 held virtually.
“We are excited about predictive maintenance,” Magoulick said. “It will make us more efficient. We can identify when a machine is trending out of control and may need maintenance, so we can schedule at the next available window. It’s the next level of predictive maintenance from what we do today.”
It also helps in the ordering of needed replacement parts. “If we know the part is going bad, rather than holding the cash in our inventory, we can order it on demand,” she said.
AI is also being applied to vision systems, making for more powerful abilities to conduct inspections during manufacturing. ”We can find defects anywhere, including seeing paint scratches,” Magoulick said.
In addition, AI is being applied to further automate the auto manufacturing process, with research into where to apply the innovation ongoing. “We are using machine learning to try to reduce our cycle times,” she said. “We recently reviewed a use case for transmission assembly, which reduced the cycle time slightly the first time through.”
In addition, Ford is experimenting with the use of natural language for voice commands to communicate with machines on the shop floor. “It’s Siri for manufacturing,” she said.
Additional areas of research include studying audio to detect quality defects, “using AI to assess what is a good and what is a bad digital audio signature,” she said. Also, Ford is experimenting with collaborative robots on the shop floor, she said.
Domain Expertise Comes from Those Doing the Work Today
Asked by session moderator David Mindell, Co-chair, MIT Task Force on the Work of the Future, and MIT Professor of Aeronautics, where the domain expertise comes from, Julie Shah, MIT Associate Professor, Department of Aeronautics and Astronautics, said it is primarily from the people doing the work today. “The domain expertise is with people on the shop floor doing the job today, learned through years of apprenticeship in some cases,” said Shah. “It might look easy in some cases on first look but it can be challenging to program.”
She added, “Being able to learn from observation and demonstration is best done directly from someone doing the task on the shop floor, to see the key factors in doing the job successfully.”
Panelist Daron Acemoglu, MIT Professor of Economics, in response to a question from Dr. Mindell on whether AI will make better engineers, stressed the need for AI engineers to have a “concrete understanding” of the social implications of decisions they will make. He also stressed the importance of government policy.
“Government priorities are signals,” he said. “If the government gives up on the agenda of creating better technology, it’s natural for researchers to do that too.”
He is concerned that AI researchers maintain autonomy from the corporate world, and that big tech companies fund much of AI research in their own AI labs. “They have their own agendas,” he said. “If those companies set the tone for leading AI labs, how can we expect the AI research to do anything but parrot the priorities of those companies. It’s a difficult lesson. We are not really establishing our autonomy. We are saying good research means we are more integrated with Google, Amazon and IBM. Autonomy is critical in this area.”
Rus Sees “Problem-Driven” Research With Industry as Productive
He was challenged on this point a bit by panelist Daniela Rus, Director MIT CSAIL, and MIT Professor of Electrical Engineering and Computer Science, who said she has had some good experiences collaborating with researchers in private industry.
“I think there is a fair bit of autonomy and a number of programs that support problem-driven research,” she said. “Maybe there is not enough funding, and in some sense, where the government is lacking, the companies are stepping in.”
She added, “I find working with companies can be enriching and empowering,” mentioning a collaboration with Toyota Research Institute about five years ago to advance the science and AI and robotics research. Outlining her thoughts, she said, “When I think about how the university and industry research labs connect together, I have a mental model where the industry development lab works on products for today, the industry research lab works on problems of tomorrow, and the university research role is to think about the day after tomorrow, connecting to how those advances matter. The applications allow us to root our ideas into things the world cares about.”
Mindel asked if we should worry about AI taking over too many functions of humans. Prof. Acemoglu said, “There is a choice. There is no iron-clad rule on what humans can do and what technology can do. They are both fluid. It depends on what we value.”
Prof. Shah agreed with the sentiment. “The machines are still performing very narrowly-defined tasks,” she said. “Deep learning is a functional approximator, like algebra and calculus. It’s how we take those tools and use them for a purpose, and how we define success for those systems that matters. We might be trying to replace some aspect of what a human is doing today, but none of these systems operate truly independently. So asking what is the way we can have these technologies achieve our larger goals is the critical question.”
Rus ended the session on an optimistic note. “In the scientific community, we advance the science and engineering or intelligence and in doing so, we accomplish many things. We get a better handle on life, and we develop a broader range of machine capabilities. I am excited about using the latest advances in AI, machine learning and robotics to make certain jobs easier, to make life easier,” she said.
Technology has allowed people to come closer together during the pandemic, she said, “Despite the fact that the world is in the middle of a pandemic. And technology has allowed us to develop a vaccine more quickly, and that is helping us address the disease.”
Read the 2020 report from the MIT Task Force on the Future of Work
