Google Parent Alphabet Introduces Intrinsic: A New Robotics Software Company To Train Industrial Robots With AI


The company Alphabet has recently announced an interesting new venture named ‘Intrinsic’ that would focus primarily on understanding the architecture of the industrial robots and make manufacturing more effortless, and subsequently lowers their prices for better affordability. This venture, if successful, could prove to be a leap further for all the industries that would be able to exploit the benefits of artificial intelligence and unlock new boundaries with the potential of robotics.

A Brief Look Into the venture

This new venture is a spinoff of Alphabet’s X research lab which works on multiple projects simultaneously. The lab’s working is relatively simple; they continue to work under one umbrella until one project starts showing significant promise. It is then categorised into an independent unit that works in a startup like manner. The only difference is that even though it is an independent unit, it is still under the Alphabet corporate umbrella. The formation of Intrinsic also happened similarly. The researchers worked on this project in the lab for five and a half years before being given the status of an independent unit.

The researchers have developed a prototype version of the technology, and now that the new venture has been spun out, the prototype will further be developed into a fine product. Once complete, this technology would then be commercialised for the industries to benefit.

The Objective

The primary aim of Intrinsic is to widen the usage of industrial robots. For example, the focus could be on robotic arms, making them more flexible and accessible. The direction in which this new venture is working is to unlock both the creative as well as the economic potential of the industrial robots for numerous businesses, entrepreneurs and developers. Moreover, the software tools that the company plans to unveil will work on a broad range of robots, right from ones making solar panels to the ones manufacturing cars.

Working of the Technology

The researchers at Intrinsic are working in a manner where they can reduce the maximum possible costs of the robots, and the key area there is programming. The robots, even the ones performing the simplest of the tasks, require tonnes of coding. Therefore, for industries, more automation would mean more software engineering work which could be pretty expensive. The approach taken by Intrinsic is making use of artificial intelligence to develop Software that would train the robots in a much easier manner and require only minimal coding. Intrinsic claims that the Software is equipped with multiple AI approaches under the umbrellas of deep learning (most complex and sophisticated AI algorithms) and reinforcement learning (a form of neural network that performs a plethora of tasks by using trial and error).

Two robots use perception, force control, and multi-robot planning to assemble a simple piece of furniture (video sped up) Source:

The Projects Undertaken for the Development of the Software

Intrinsic undertook an experiment wherein the Software so developed was used to train a robot in connecting USB cables to a panel. As claimed by Intrinsic, the results were excellent as the Software successfully trained the robot in merely 2 hours. The researchers also partnered with a team at Swiss University ETH Zurich to manufacture an architectural installation that was sustainable as well as successful. Four industrial robots powered by Intrinsic Software were used to assemble the parts for the installation. Not only this, Intrinsic is continuing to partner with more companies in the automobile industry, electronics industry and more to enhance the technology further and make it available for usage.

This is one project in the loop; other than this, Facebook Inc is also putting its foot in the robotics software research arena. On the other hand, Google is already quite ahead in the race of artificial intelligence and has an entire unit of researchers developing new machine learning techniques.