The company is also working on a mobile pick-and-place robot that can roam warehouses on its own.
These tasks won’t come easy for a robot, but Tutor’s chief executive Josh Gruenstein says that with enough training from remote human helpers, machines will eventually figure it out. And the same goes for just about any industrial job.
“Our ambition is any task that a human being is doing in the warehouse or factory . . . we expect to be able to scale into that work,” Gruenstein said.
If he’s right, those predictions that robots will eliminate millions of jobs might, at last, come true.
“[W]e’re building the solution to the most manually-intensive tasks on the line,” reads a statement on the company’s website. “We’re building an infinite labor supply, to deliver value directly into the organizations that move the world’s goods.”
Cassie is a stationary, one-armed machine with a large array of suction cups on the business end. The suction gripper is Tutor’s own design, but most everything else is off-the-shelf hardware, largely made in China. Tutor’s chief contributions are the software that drives the robots and the humans who train them.
Gruenstein came up with the idea for Tutor while working on his master’s in AI at the Massachusetts Institute of Technology, along with cofounder and fellow MIT grad student Alon Kosowsky. They’d worked on the idea of training robots through digital simulations. They also built a virtual robot made entirely of software, then operated it inside a virtual warehouse. The team hoped that this kind of simulation could replace real-world training.
But it couldn’t. “You realize that the real world looks different,” no matter how accurate the simulation, Gruenstein said.
So Tutor is putting its robots into the real world and training them on the fly. Founded four years ago in stealth mode, the company has placed prototype robots in a variety of industrial settings. The goal was to “create this virtuous cycle: you deploy more robots, you collect more experience, you become smarter, and that enables you to deploy more robots,” Gruenstein said.
Their first customer was a perfume factory in New Jersey. “It required the full-time assistance of human beings sitting in Boston to teach that robot all the things it needed to do that job,” said Gruenstein. “We’ve bootstrapped that learning and that experience into an operation where today we operate robots from coast to coast across the continental United States.”
The company doesn’t sell its robots. Instead, customers “hire” the machines for $14 to $18 per hour, less than they might have to pay a human worker.
Ideally, the Cassie machines just sit there stacking boxes, with no need for human involvement. But if there’s an issue, the robots can request assistance from “tutors” who sit before large, flat-panel screens. Tutor Intelligence operates two tutor teams — one in Watertown, the other in the Philippines.
During a recent visit, a trio of tutors sat under soft lighting, gazing at large computer monitors. Multiple windows presented a robot’s-eye view of jobs being performed at various locations. At that moment, the robots were doing fine on their own. But if one of them had encountered a problem — say, an odd-sized package it wasn’t sure how to lift — one of the tutors would take control and guide the robot’s movements.
It’s not the only company training this way. For instance, if one of Pickle Robot’s truck-unloading robots gets confused, a human can be called in to help.
“We take the data created by the pilot’s remote assist and use it to teach the robot to ‘unstick’ itself when similar situations arise in the future,” said Pete Blair, vice president of marketing at Charlestown-based Pickle.
Some roboticists say that remote training has its limits. “There’s no world in which an AI operator can train the system for every situation that may occur,” said Taskin Padir, a professor of electrical and computer engineering at Northeastern University.
That’s especially true for tasks that require a lot of manual dexterity, like installing an electrical panel full of wires and circuit breakers. Today’s robotic hands don’t have the agility and precision for such work, Padir said, even with help from humans a thousand miles away.
But Padir estimated that AI-powered robots like those being developed by Tutor Intelligence will someday be able to do 50 or 75 percent of the things people can do. Maybe some of the unemployed will find work training robots, but it won’t last. You only have to teach them once.
🤖 IRobot files for bankruptcy, will be sold to Chinese manufacturer. Read more from business reporter Jon Chesto.
📊 Electronics or elder care? See which Massachusetts job sectors are rising or falling. Read more from data reporter Scooty Nickerson.
👋 These Brandeis students created a friend-matching app meant to get Gen Z off their phones. Read more from Globe contributor Yogev Toby.
📦 DraftKings and Wayfair plan big moves for their headquarters in Boston. Read more from business reporter Jon Chesto.
📸 N.H. inventor Dean Kamen pictured in newly released Epstein image. Read more from New Hampshire reporters Amanda Gokee and Steven Porter.
🎂 MIT’s startup mentoring service was supposed to last a few years. Try 25. Read more from business reporter Jon Chesto.
🚌 Electric school bus operator Highland Electric Fleets in Beverly raised up to $150 million from Aiga Capital Partners.
🌊 Autonomous underwater drone maker Vatn Systems in Portsmouth, R.I., raised $60 million in a deal led by BVVC and including Propeller Ventures.
💸 Boston-based OnCorps AI, developing applications for investment fund operations, raised $55 million from Long Ridge Equity Partners.
⛏️ Mining tech startup Allonnia in Boston raised more than $20 million in a deal led by Viking Global Investors, Bison Ventures, General Atlantic, BHP Ventures, and Pivotal Capital Partners.
🩻 AI imaging startup a2z Radiology AI raised $4.5 million from investors Khosla Ventures and SeaX Ventures.
📹 Marketing video company Goldcast in Boston was acquired by Virginia-based events tech company Cvent. Terms of the deal were not disclosed.
🎰 DraftKings president and cofounder Matt Kalish is leaving his executive role at the company. Kalish, who is launching the creator media platform HardScope, will remain on the board of DraftKings.
🛍️ Ecommerce marketing company Klaviyo in Boston added Chano Fernández as co-chief executive officer, effective Jan. 1., alongside co-founder and co-CEO Andrew Bialecki. Fernandez previously was co-CEO at Workday and co-CEO at Eightfold.
✈️ Travel software firm Mobi.AI hired Stuart Barwood as chief revenue officer. Barwood previously was chief commercial officer at Tripstack.
📍 Automation and testing company Teradyne in North Reading is opening a new robotics manufacturing facility in Detroit next year.
Teens, Social Media and AI Chatbots 2025 (Pew Research)
The Publicity War Between AT&T and T-Mobile Is Getting Ugly (Wall Street Journal)
From puppies to superheroes, Chinese AI toys are bringing hugs — and hesitation (CNBC)
👋 Thanks for reading.
❓ Have a tip? Email Aaron at aaron.pressman@globe.com.
✍🏼 If someone sent you this newsletter, you can sign up for your own copy.
Hiawatha Bray can be reached at hiawatha.bray@globe.com. Follow him @GlobeTechLab.
