"Toothbrush Test" did not allow Google to produce industrial robots

This week, Google has published the results of a robot self-learning experiment. About this experience article and on Geektimes was published. As it turned out, self-study for robots is a very real process, and this is not about the individual training of each system, but about the “kindergarten” for robots. The company showed the results of this work, recording the training progress on video.

The experiment was not conducted for the sake of the experiment itself. Google planned to sell self-learning manipulators to various companies, retailers, owners of huge warehouses. But these plans did not come true: after Larry Page’s decree on the mass production of self-learning robots had to be forgotten. And the reason is simple - the robots did not pass the “toothbrush test”.

This is a company's requirement for its products. Management believes that Google should only sell products that will be used by billions of people. Yes, just like toothbrushes. The management did not make this decision now, but at the end of 2015, just before the robots department moved from Google to X, a division of Alphabet holding. This unit is engaged in projects such as, for example, robotic cars. Alphabet X has something to show the public, but it’s not ready to sell its products yet.
')
“This is still a prototype, but with tremendous benefits,” says James Kuffner. Now this specialist holds a senior position at the Toyota Research Institute. Previously, he led a division that developed robotic systems at Google. “The team has worked well. If it depended on me, I would have already started selling. But no, ”complains Kaffner.

Google spokesman Jason Freudenfelds (Jason Freidenfelds) said that plans to sell these robots yet. “We use them for basic research related to machine learning. Now we are trying to teach robots to be more accurate, and something happens, but we cannot talk about sales, ”says Freudenfelds. Courtney Hohne, spokeswoman for Alphabet X, declined to comment on the robots.

Now Google has already 50 robotic manipulators capable of lifting a weight of up to 4.5 kg. These manipulators were manufactured by Meka Robotics, a startup absorbed by the corporation in 2013. The cost of manufacturing such manipulators is not too high, so the unit was going to sell them at a relatively low price. But, as mentioned above, these plans had to be canceled. Moreover, the project participants were asked not to comment on the company's “internal kitchen”.

Similar models from another company, Universal Robots, are priced at $ 20,000.

The current situation may indicate that the company has difficulties with some high-tech divisions. They are not without leadership, but there is no understanding of what to do and what to do next. Moreover, the company plans to sell some units. First of all, we are talking about Boston Dynamics. At the beginning of the year, the corporation announced the possibility of selling Boston Dynamics, but so far there has been no buyer.

Google is rumored to often have problems with positioning and selling new products. Larry Page convinces employees not to be afraid of creating products that can be risky, but should have a strong effect (in their own field, for example). Often, this results in such products simply gathering dust on the shelves, remaining intact for many years.

Kaffner compares the situation in Google with Amazon, which began selling its Echo smart speaker two years ago. “When this gadget was first introduced, it was pathetic. He was unreliable, ”said Kaffner. “But they started selling them. This allowed Amazon to collect user opinions about its product and quickly fix it. And now they sell a large number of devices. ”


Interestingly, the other day, Google introduced its own wireless speaker, which works on the same principle as the Echo. According to Kaffner, the situation with the delay in the release of new Google products is largely due to the need for the company to protect its brand. This is not so easy to do, because the product should not be associated with something that is not up to par. But nevertheless, it is necessary to do something, because the situation with robotic self-learning manipulators shows that the company detains even good products, which are not equal.

The positive thing is that Alphabet has not completely abandoned its project. It works, although robots are not for sale. Already, robots are gradually learning to cope with new conditions for themselves. Joint training helped them, for example, learn how to perform the most difficult tasks of opening various doors and locks. The final experiment showed the effectiveness of this type of training: robots were able to open the door and lock, which they had not yet encountered.

Perhaps when robots are taught to perform more than the current number of operations, they will become useful in everyday life. And then the "toothbrush test" will be passed, after which the company will sell its robots.