Rethink Robotics shuttered its doors and closed for good on October 4, 2018. For many casual observers the collapse of a much-celebrated company, founded by preeminent artificial intelligence (AI) researcher and minor celebrity Rodney Brooks was a surprise. To others it’s just the latest indication of the trouble in robotics land. According to the company, Rethink Robotics was forced to shut its doors when it couldn’t find additional funding, and in a final attempt to sell the company and/or its assets it couldn’t find a buyer.
Rethink Robotics wasn’t the only robotics company forced to close its doors in the past year. Mayfield Robotics, developers of the social robot Kuri, closed down a few months before Rethink in August 2018, despite just making its debut one year earlier at CES 2017. Prior to that, Jibo, makers of a personal robotics device also shut down even after having raised over $70 Million. These companies shut down despite collectively raising several hundred million dollars in funding and developing their products over many years.
This is particularly perplexing since many AI companies are flush with cash and raising money at increasingly eye-watering levels and valuations. How could it be that these robotics firms, run and operated by some of the most celebrated people in the AI industry could be failing when seemingly less-compelling solutions such as process automation tools and facial recognition applications are raising billions of dollars? Is robotics really that hard or is there something else going on in the industry?
The Venture Capital Disconnect
In the past few years, a staggering amount of venture capital has been raised by companies in the AI, machine learning, and cognitive technology spaces. According to a report by KPMG, over $12 Billion dollars in venture capital was raised in 2017 alone, more than doubling the previous year’s record tally of over $5 Billion. This is a dramatic increase from 2008 when total AI funding was less than $200 million. According to Crunchbase, the average early-stage round for an AI startup in 2010 was about $4.8 million. In 2017, that ballooned to $11.7 million. In 2018, a single company, SenseTime, raised over $1.2 Billion in venture capital, with a rumored additional $1 Billion coming from venture giant Softbank. This is more money than was raised in the entire industry just a few years ago.
So, how can it be in this industry awash with money, where anyone with a half-rational business plan spouting the terms AI and machine learning can raise ridiculous amounts with little market validation while the well runs dry for others run by industry veterans? Is the problem with AI? Is the problem with venture capital? Is the problem with robotics? Or is there something specific happening at each one of these notable robotics failures that bears closer examination?
YOU MAY ALSO LIKE
Robotics is Hard
The oft-quoted refrain in the industry is that “robotics is hard.” It’s hard to make devices made of metal, electronics, and other human-engineered bits function in the same sort of purposeful, elegant way that human bodies can. Getting machines to do seemingly simple tasks like climb stairs, slide down fire poles, assemble intricate components, and exhibit the dexterity that most humans can is an extremely difficult task. If you think engaging Alexa or Siri in a natural conversation is difficult, just try building a robotic humanoid that can function in any capacity similar to a human.
It is for this reason that most robots have been confined to domains where their range of activities and required dexterity and adaptability can be simplified: stationary industrial robots that perform repetitive welding and manipulation operations, semi-autonomous robots that roam hospital hallways delivering supplies, or social bots that provide limited interaction and entertainment. In these roles, robots seem to do well. But these limitations make them less of the intelligent sort of robot with different variations on automation and performing repetitive tasks.
This is where Rethink Robotics’ collaborative robots (cobots) were trying to break the mold. Rather than being purpose-programmed industrial bots operating in isolated environments, they were meant to be flexibly trained and repurposable assistive machines that could operate in close proximity to humans. While Rethink Robotics pioneered the cobot space, they weren’t the only ones to jump into the space. Companies like Universal Robots, ABB, and KUKA jumped into the space once they saw the promise of close-proximity, small form-factor robots that could be easily trained and operate in conjunction with humans.
Confusing Research with Product-Market Fit
Despite the demise of Rethink Robotics, cobot offerings from Universal Robotics and others are going strong. So, is there something else happening in the industry that’s not a sign of a larger scale problem? Perhaps the issue is not with the industry as a whole, or even the specific approach, but rather the go to market approach and strategy of research-driven organizations.
When news hit that Rethink Robotics was shutting down, the response from those actually implementing bots in practice was mixed. Some said that Rethink has bots that were easier to program and could handle a wider range of use cases, but were problematic in other ways including fragility, limited add-ons, and also were more expensive than some of its competitors. While there are arguments to be made from others about advantages that Rethink Robotics had over its competition, it seems that the real issues boiled down to standard competitive issues. Simply put, Rethink Robotics helped to grow and spawn an industry only to find that more nimble competitors outcompeted it. For whatever reasons, the venture investors determined that these market forces were more important than any longer-term vision that the robotics company had and decided not to continue funding it.
Similarly, one can point to the failures of Mayfield, Jibo, and others as indications of providing a solution that few found valuable. There hasn’t yet been intense need for social robots or floor-roaming bots outside of hospital and hotel contexts, and this lack of market demand forced the companies and investors to rethink their investments in the space. In this way, the market is acting rationally just like it would for any product. Regardless of the amount of money being put into these robotics companies, if the need for those products fails to materialize, or in the case of Rethink Robotics, alternative solutions are more appealing, then it becomes harder to justify putting money into an industry that’s particularly cash hungry.
Where to Next?
All this seems to indicate that the robotics industry isn’t going away anytime soon. If anything, the fact that investors are being more critical with their investments, paying more attention to market forces than to visionary-led promises means we’re entering a reality-driven age of investing in AI. The trouble is that this reality phase seems to be limited (so far) to the robotics industry. Tech companies in other corners of AI are still being wooed by investors with deep pockets and more patience than they have for robotics. Will the investments continue at the amounts and valuations currently supporting the industry? Or will these investors also be dragged down to earth by market and competitive realities? All that still remains to be seen. The hope is that the investment does continue, because after all, the quest for the intelligent machine has yet to be fully realized.