automotive manual assembly line

I live in the Bay Area. And in our bubble, the future is here. I can honestly say that I see self-driving cars pass by my house almost every day, often captured on my Internet-connected doorbell. I’ve driven past robotic carts delivering library books. My next door neighbor designed hardware for high altitude Internet-delivering kites. When I take my daughter to the park, two-thirds of the parents are wearing Google t-shirts. 

In other words, the technology my community members work on dominates the headlines. But when I meet them, here’s what I tell them: If you want to solve really interesting problems, go check out a factory.

That statement always elicits a pause. Manufacturing? They ask me. Wasn’t that industry solved 50 years ago? Isn’t it fully automated, like in the old Intel commercials? What innovation potential could there possibly be?

I have three answers for them.

First, because manufacturing is BIG, and transforming fast

  • It’s 15% of global GDP, driving $12 trillion in yearly revenues. 
  • Manufacturing is so big, in fact, that it’s strategic at a national level. India wants to add 100M manufacturing jobs by 2025. China wants to increase domestic content of core materials to 70% by 2025. The U.S. has a national plan to secure its manufacturing future. And Germany literally invented the phrase “Industry 4.0.”
  • Manufacturing is changing fast. Industry 4.0 is applying leading-edge technology from across the spectrum, including robots, IIoT, sensing, 3D printing, cloud computing and much more.

Second, because manufacturing has loads of problems to solve – and most of them revolve around the human factor:

  • In 2012, Foxconn announces plans to deploy 1M robots. By 2016, they’d only deployed around 40,000 robots – 4% of their target.
  • There are ~1.5M industrial robots in the world, with 250K added annually. Economists from the National Bureau of Economic Research estimate that one robot displaces six jobs. But with 345 million people working globally in manufacturing, new robots are displacing less than 1% of the workforce a year.
  • Drishti’s research with A.T. Kearney shows that humans still perform 72% of the tasks inside of a factory, while creating 68% of the defects. 
  • Manufacturing is shockingly complex. The average car has 30,000 parts – all of which have their own supply chain and assembly process. Consumers are demanding even more complexity, which is why there are at least 4.1 million possible build permutations for a Ford F-150.

Most importantly, all of that complexity has to be simplified in production to the point where just about anybody can pick up the work instructions and put the unit together. Which is why manufacturing is so well-suited to AI: Incredible complexity has to be simplified and made repetitive. 

Four reasons AI + manufacturing = perfect match

  1. The factory environment lets you solve harder AI problems in a production environment: In a factory, the dataset is repeatable. A small set of interesting things happens over and over. The background is not extremely varied, the environment is reasonably uniform and the environment is controllable. Contrast that with self-driving cars, and you see that many of the complicated factors are absent – which lets you apply more advanced AI techniques that are generally still in the laboratory. Action recognition, for example.
  2. The market demands you push the technology farther: Working in manufacturing, AI faces the inherent challenges of having to deal with human variability and idiosyncrasies. If you’re working in computer vision, you face interesting challenges of geometry and perspective. And cost is a factor: You have to be extremely efficient for your solution to be viable in the market.
  3. There are fascinating non-AI challenges for AI in manufacturing: If you’re going to make AI viable, you’re also going to have to build out an entire stack around it. Drishti, for example, has to solve heavy problems of pipeline (transferring data to the cloud, within the cloud, and back to the endpoint), scale (we create YouTube-like volumes of data), accuracy (customers have an extremely low tolerance for errors) and usability (the presentation layer for our data has to be utterly intuitive to a non-technical person).
  4. The market is ready: Is there any industry more primed for an AI solution? Factories live and breathe data. They are laser-focused on finding outliers, which is exactly what AI is perfect for doing. They are battling disruption (think of the electric car’s impact on Tier 1 suppliers of internal combustion engine parts) while embracing disruption (it seems like every major manufacturer has opened an office in Silicon Valley). In this industry, there are huge rewards for a competitive edge – and if you’re in the software industry, that means there’s a market there for the taking.

It’s easy to be drawn into industries that are building futuristic, sci fi-level tech. And there’s certainly a need for people in those companies. But there’s a massive opportunity in the manufacturing space that is often overlooked, because people don’t realize just how intriguing factories can be. 

Drishti was recently named a Technology Pioneer by the World Economic Forum. Find out why.