Amazon states that fully automated shipping warehouses are in at least ten years

The future of Amazon's logistics network will undoubtedly involve artificial intelligence and robotics, but the question of when AI machines will do most of the work will be open. According to Scott Anderson, director of the company's robotic solutions, the point where an Amazon warehouse is fully automated from start to finish is in at least 10 years. Anderson's comments, reported today by Reuters, highlight the current pace of automation, even in environments that are conducive to robotic work, such as an Amazon warehouse.

In the current state of things, robots in the job market are mainly mastering specific and repeatable tasks for which they are accurately programmed. Bringing the robot to do something else requires expensive and tedious reprogramming. And robots capable of performing many different tasks and operating in dynamic environments that require the robot to see and understand what surrounds it are still in the field of research and experimental testing . Even the simple process of identifying an object and picking it up without ever having seen it before requires a complex and sophisticated set of software and hardware that do not exist commercially yet. .

Thus, even if a robot can manufacture a microchip and the bodywork of a Tesla motor vehicle, it is not able to perform the human tasks required by the work in the warehouses. In Amazon's facilities and the execution centers of other companies, most of the work is still done by humans, because it is difficult to train robots so they can see the world and use robotic forceps with the dexterity of human workers.

However, as part of the ongoing in-depth learning revolution that has accelerated the progress of AI research over the last decade, robots are beginning to gain levels of vision and knowledge. engine control that approach levels of human sophistication. Amazon is one of the pioneers in this field, and every year it organizes a picking challenge, which is to pick up an object to move it to another part of the supply chain, to promote breakthroughs in the field. ground.

Several other companies and research laboratories have also made progress in this area. UC Berkeley has a robotics laboratory that has made significant progress in the field and its new low-cost robot, a pair of humanoid arms controlled by a central system called Blue, can perform complex manual tasks such as folding a napkin motorized vision system. Similarly, the OpenAI research laboratory uses an AI training technique called reinforcement to teach a robotic hand more precise and elegant movements, types of movements that would be needed for a robot to replicate a human in a warehouse. . Kindred, a San Francisco-based start-up, is manufacturing a robotic arm called Kindred Sort, which is deployed in warehouses for the Gap retailer, which uses a mix of human control and automation to make a dynamic product selection.

according to Reuters, Amazon has 110 warehouses in the United States, 45 sorting centers and approximately 50 delivery stations, all of which employ more than 125,000 full-time warehouse workers. But only a fraction of this work is done by robots. At the present time, robots are just too imprecise and clumsy and require too much training to be able to be deployed in factories, apart from very narrow use cases.

For example, Amazon uses small Roomba-style robots, simply called "readers," mainly to deliver large piles of products to human workers, following defined paths around the warehouse. "In the current form, the technology is very limited. The technology is very far from the fully automated workstation we would need, "Anderson said. Reuters, who visited an Amazon warehouse in Baltimore earlier in the day.