Google’s Re-Pivot Robotics Launch is a Smart Machine Doers story

In In 2013, Google started an ambitious, flashy effort to create robots. That didn’t work out so well, so now, in 2019, it’s off to a fresh start. Hardware is seen as the bottleneck of deep learning’s evolution; indeed, the smartest AI company on Earth recognizes that fewer humanoid robots with more sophisticated software is the way to go!

The IoT is coming, 5G is coming and we’ll need robots for that. Remember Andy Rubin? Yeah, that guy! Google has been quietly retooling an ambitious but troubled robotics program (NYT) that was once led by an executive who left the company amid accusations of sexual harassment.

Alphabet is incredibly like a microcosm of the world. A bizarre world where the American POTUS must sit down with Google to get reassurance it’s on our side. Trump previously said the company, along with Facebook and Twitter, was “treading on very, very troubled territory.”

Google caters to the U.S. Military, but in exploring a censored search product in China, it might even be playing both sides. The future of robotics certainly involves both sides, whatever happens with a Trade deal or partnerships with “strategic rivals”.

The Smart Machine “Doers” are Coming
Robots have been able to pick up and move objects for decades, but only if they were programmed with exact instructions to complete the task. Neural networks have opened up new frontiers in robotics research. Now obviously Google wants to be at the head of the pack.

Hardware and software are converging. AI chips made by NVIDIA and the G-MAFIA giants themselves are changing the future of how robots will scale. We’re getting closer to the fabled 4th industrial revolution acceleration of the smart city. Alphabet won’t just leave that to Waymo, it needs a coordinated attack beyond just deep learning and autonomous vehicles.

As specialized AI chips are the future, and chip makers are scrambling to figure out which designs will prevail, functional robots are getting smarter and better at performing actual tasks in the real world.

In recent times, Google AI researchers have built a simple robot arm that understands the physics of tossing objects, which is more impressive than you might assume. When Google bought and then sold Boston Dynamics (later bought by the Japanese conglomerate SoftBank), which is still working on robots that move like humans or animals, Google decided to go down a different path, a robot of things path.
Now the new director of Robotics at Google is Vincent Vanhoucke, who helped Google launch its ongoing AI research department.
In collaboration with researchers at Princeton, Columbia, and MIT to develop a picker robot they dubbed TossBot, which learns to grasp and throw objects into boxes outside the confines of its “natural range.” Pretty impressive right?

Google has regrouped, reconsidering its focus on the mechanics of complex robots. The era of smart machines is coming, and companies like Amazon, Google, NVIDIA and Apple will likely be in the mix. “Doer” robots comes well before consumer robots really are smart and cheap enough to be viable. The doer robots will be cooks at fast food places, operate “Cloud Kitchens” and be a barista at your local coffee joint; doer robots that will be making pizzas in high volume businesses, better and faster than a human could.

Google
Google’s also designing these new robots to be good at their assigned tasks and not to look like humans or animals. This essentially means Google’s robotics approach now is tethered to an internet-of-things IoT approach, much in the way of a Huawei or a Xiaomi. Smart robotics and BioTech are likely waves of technology that will be dominated sooner rather than later by China.

The problem is that this puts Google head-on in the military robots debate. A lot of these things have implications for DARPA, like utilization of R&D funds. When the CEO of Google strongly states that he’s committed — like totally committed — to the U.S. Military, that’s not such a popular thing to say necessarily within Google with its own employees. To a greater or lesser extent, Amazon, Microsoft and Alphabet are becoming big providers for the U.S. Military and national defense, for killing and national rivalries rather than making the world a better place.

The Post Rubin Era
Google is thinking along new lines regarding how machine learning and robotics work together. It has been rebuilding its program for the last few years, with robots that are much simpler than the humanoid-shaped machines of its glorious, if faulty, robotics past.

Several contractors and robotics researchers told the Times that they’ve found machine learning AI valuable in their projects, which involve things like training a robotic hand to grasp delicate objects. A not-so-stealthy Google! But obviously up to something useful.

The new lab called “Robotics at Google,” however, focuses more on software: more precisely, in using machine learning to develop robots useful in the real world.
Atlas can pick up a box and move it to another location, but the TossingBot can sort through a whole bin of different objects in the same amount of time. It’s not clear if Google will throw enough money at this to keep up with market leaders, though. Alphabet would have to go on another robotics acquisition binge, no doubt.

Robotics at Google is working on an R2-D2-like rolling robot that uses machine learning to navigate new spaces. Researchers are also training a set of mechanical hands with fingers to manipulate objects by pushing, pulling and spinning them. Okay, cool enough!

Robots Drunk on Machine Learning Could become more Independent
Robotics at Google recently moved into a new lab on Google’s main campus in Mountain View.
Google researchers believe that the subtly more advanced technology inside them gives them more potential in the real world.
The company is developing ways for these robots to learn skills on their own.
This isn’t so dry at it sounds, it’s the holy grail of the future of robotics.

Let’s be clear though, Google isn’t a robotics company. Although Google is a pioneer in AI research, its efforts in robotics have produced no commercial successes to date after many years of efforts.

Google has always loved robots, but never really known what to do with them, until 2019 it seems. Google’s new lab is indicative of a broader effort to bring so-called machine learning to robotics. Google remains the top AI-engineer talent hub on the planet and needs to leverage this now, before that title changes hands. It’s a closing window of opportunity for a company that makes the bulk of its revenue simply off of Ads.

Can Google Make smart machines in an ethical way?

The robotics companies Google bought in 2013 were focused on ambitious legged robots, like the one above designed by Schaft. Credit: DARPA
Google has offered no long-term road map for its robotics research. Nor has Google given much indication it knows how to scale robotic products that align with stringent ethical guidelines. Despite drafting some guiding principles in AI, Google has in the past two years made a bunch of PR errors relating to its business strategy that actually point to diminishing executive ownership of values for social good.

Google opened an artificial intelligence lab in Beijing in 2017. Earlier this month, Gen Dunford said the lab “indirectly benefits the Chinese military”. Google was (and likely still is) creating a censorship search product for China. If corporations want in both ways, what sorts of ethics will surround the smart machines of tomorrow? That’s not clear, but it’s still early days for the future of robotics.

It’s entirely possible OpenAI, the artificial intelligence lab founded by the Silicon Valley kingpins Elon Musk and Sam Altman, will become more relevant to the future of robotics. Google has a lot of side projects, but realistically, most of them will fail. That’s not to say that Google isn’t smart at what it does, just to say that for 4th industrial revolution tech there will more than likely be new winners.

The AI chip revolution will transform the future or robotics in ways we can’t truly appreciate in 2019. Many believe that machine learning — not extravagant new devices — will be the key to developing robotics for manufacturing, warehouse automation, transportation and many other tasks. Google is arguably the global leader in machine learning.

Robots are already used in warehouses and on factory floors, but they can handle only specific tasks, like picking up a particular object or turning a screw. The tasks robots are able to do will only increase. The ratio of robots to human workers will only increase. Smart machines in a world of automation are the future, a world in which machine learning is embedded into everything — and that sounds a lot like being on the perpetual cusp of an internet-of-things: robots, algorithms, and then humans.