Welcome to the Age of Garage Robotics

A couple of weeks ago, Chris Dixon of Andreessen Horowitz published a blog post about his most recent investment in Comma.ai, a new AI/autonomous mobility company. Comma.ai is the genius of hacker extraordinaire George Hotz, who built his own self-driving car in his garage that was featured on Bloomberg News below (Chris actually rode shotgun!).

Chris remarked that George’s self-driving car is similar to what happened with social media leapfrogging telecom, as stated, “WhatsApp was able to build a global messaging system that served 900M users with just 50 engineers, compared to the thousands of engineers that were needed for prior generations of messaging systems. This ‘WhatsApp effect‘ is now happening in AI. Software tools like Theano and TensorFlow, combined with cloud data centers for training, and inexpensive GPUs for deployment, allow small teams of engineers to build state-of-the-art AI systems…I came away convinced that George’s system is a textbook example of the ‘WhatsApp effect’ happening to AI.” 

However, Chris was only partially right, the WhatsApp effect is happening across the robotics spectrum as well. As Rodney Brooks predicted years ago, we are entering the age of personalized robotics, with tinkerers and hackers becoming founders of the new autonomy economy. Earlier this week the Wall Street Journal ran a front page article about high-tech farmers upgrading their tractors with autonomous features.

Matt Reimer, in the spirt of Hotz, modified his eight-year-old John Deere tractor with drone parts, open-source software and a Microsoft Corp. tablet to create his own, home-built, autonomous combine. While John Deere has been selling autonomous equipment for tens of thousands of dollars for over a decade, Reimer’s investment was a fraction of the cost, roughly around $8,000. He said that’s about how much he saved on wages for drivers helping with last year’s harvest.

Mr. Reimer’s alterations, which he hopes to replicate for other farmers this year, are part of a technology revolution sweeping North America’s breadbasket. Farmers, many of them self-taught, are building their own robotic equipment, satellite-navigation networks and mobile applications, moving their tinkering projects out of machine sheds and behind a computer screen.

This homespun hacking—which sometimes leapfrogs innovations by big equipment companies like Deere & Co. and navigation specialists like Trimble Navigation Ltd.—reflects dwindling farm incomes, the low price of electronic hardware and, sometimes, off-season boredom. Such projects could eventually compete for farmers’ dollars in the farm-technology market, which generates an estimated $2 billion in annual sales, according to data and research firm IBISWorld.

“Even if they release [autonomous tractors] next year, it’s probably going to be 15 years before that technology trickles down to every farm,” said Mr. Reimer, referring to the big farm-equipment companies. What’s more, his version would be “a lot cheaper than if somebody’s got five to 10 engineers working full-time on something like this,” he said. He added that his system doesn’t require altering Deere’s own software or coding.

The engineering-school dropout says he picked up programming from online forums and coursework archived on the Massachusetts Institute of Technology’s website, though he said he didn’t complete that program either.

Technology is already firmly rooted in modern farming, allowing a shrinking number of farmers to oversee more acres. Advances like auto-steering tractors have freed some farmers to trade futures contracts and gripe about the weather on their smartphones from inside a tractor cab, pausing only to turn and stop their machines. Defectors from Silicon Valley powerhouses like Google Inc. and Yahoo Inc. are building software to analyze soil and manage fertilizer use.

A three-year slump in major crop prices, however, threatens the big bet on farm technology by companies like Deere, Monsanto Co. and DuPont Co., which have sunk hundreds of millions of dollars into precision-guided planters and algorithm-powered advice on managing crops—plus cloud-based systems to manage the data. Sliding prices for everything from meat to melons has U.S. net farm income on pace to fall to $54.8 billion this year, down 56% from 2013 and the lowest level since 2002, according to government forecasters.

Kinze Manufacturing Inc., an Iowa-based farm-equipment maker, is working on driverless tractors, including a system that allows several to load grain on the same field. Kinze continues to invest in the research, but Phil Jennings, service manager at the company, said turning a profit from such cutting-edge machinery gets harder when the price of corn, the most widely grown U.S. crop, is at less than half its summer 2012 peak.

With less money to spend, some farmers say they can build their own tools, suited to their farms, at a lower cost.

“Poverty is the mother of invention,” said Jim Poyzer, 65, who returned to farming six years ago after a few decades in computer programming. During the winter months four years ago Mr. Poyzer began tinkering with a microprocessor, eventually developing a system to monitor and adjust how many seeds his planter places in his fields near Boone, Iowa. The system tailors the flow of seeds to the soil’s ability to produce healthy crops.

Matt Reimer, whose farm is south of Killarney, Manitoba, checks the control box for his autonomous tractor. He estimates the system’s cost at about $750, versus commercial versions that retail for around $5,000, and says it helped him save about $1,000 a year on seeds. “That’s not much, but farmers are trying to optimize everything,” he said.

Now, Mr. Poyzer is working on other projects, like a solar-powered sensor to monitor soil temperature that he says could help him get a jump on planting.

Some companies, including Deere, have taken steps to prevent anyone from modifying the software that runs their equipment, and also warn that altering a tractor’s systems could put farmers and workers at risk. Deere uses copyrights and other intellectual-property measures to protect its software.

“We always have producers wanting to build on top of those solutions. I think it’s a great thing,” said Cory Reed, head of Deere’s intelligent solutions group. But when it comes to Deere’s software, Mr. Reed said, “there has to be a limit, both for regulatory safety and for proprietary reasons.

Deere doesn’t object if farmers create their own systems to work with Deere equipment, without altering the company’s embedded software, so long as these meet industry standards, he said.

Some farmers-turned-techies aim to reap profits on their innovations. Dirt Tech, a startup run by two farmers and two software engineers, is developing a range of mobile applications that help map soil fertility across farmers’ fields, or mark rocks to avoid damage to machinery or allow for yanking them out. The Elbow Lake, Minn., company’s apps have been downloaded more than 4,500 times.

“We really enjoy solving these problems,” said Ben Brutlag, a co-founder of the company who raises corn, soybeans and sugar beets near Wendell, Minn. But, he said, “we’re definitely trying to make some better money at it.”

The democratization of technology is not limited to farming, last year a dutch engineer made big waves by building his own drone that is capable of carrying a human short distances (see above test flight).  In this new age of the commercialization of space and the X-Prize, the sky is no longer the limit.


A Scandinavian inventing shop called Copenhagen Suborbitals is taking the WhatsApp effect to the extreme level by building a single person bullet (drone) aircraft missile at a cost of of $70,000 vs. $350,000,000 (by NASA) to enter the space tourism market.  They are even running successful test-flights over the Baltic Sea with crash-test dummies, proving you do not need a Richard Branson like budget to be successful.  Sit back, relax, as the next decade takes off into the stratosphere.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: