Taking a Whole-Body Approach to AI and Robotics
July 01, 2019      
Josh Ong

How do you build a robot that’s truly useful? Obviously, the answer depends on your definition of what a “useful” robot does; but in this case, I’m referring to robots that can perform helpful tasks, such as giving someone directions, or making a cup of coffee, which previously only humans could carry out.

The approach most scientists have taken thus far is to focus on perfecting a certain type of function or movement – say, clutching an object, or recognizing speech – instead of developing multiple capabilities simultaneously that could then be integrated into a robotic whole.

We’re calling this different approach the whole-body approach to developing robots. Essentially, you can view the human body as a chain of connected parts and functions, and use that as a framework for developing the corresponding AI parts. For example, the robotic equivalent of the brain would be its hardware chip combined with the algorithms that enable it to think, learn, and function. Computer vision and image recognition would be the robot’s eyes, its microphone and language recognition abilities its ears, text-to-speech capabilities its voice, and so on.

Very few AI companies have focused on developing the entire breadth of that chain. Some focus specifically on computer vision, others on text-to-speech, still others on developing robotic arms and legs that mimic human movement. However, if you look at the bigger picture, it becomes easier to develop products that will be successful in their objectives, and consequently have a more impactful on people’s lives.

Beyond isolation

Instead of focusing on perfecting each aspect in isolation and moving on to the next, it can be more efficient to create a platform that encompasses the different senses, skills, and movement of an entire body. Not only does this broader approach widen the applications, it also makes it easier to figure out how different pieces of machinery can work together in tandem. It also makes it easier to bring new products to market, since the underlying technology is already there; it just needs to be assembled.

For example, at OrionStar, we are creating robots that can both see and move – or one that can grab objects and think – this makes it possible to use them in multiple situations. Its latest release, CM Translator, uses a “brain,” “ears,” and “mouth” to listen to conversations and accurately translate them into six languages.

Not about humanoid robots

To be clear: this process is not about making the most human-like robot. Taking a holistic approach to robotic development does require understanding how humans move, speak, listen, and interact, but it does not mean you have to completely mimic human function. Instead, companies should focus on developing machines capable of using those skills to carry out certain tasks. It’s helpful to start with a holistic vision, but when it comes time to build products and bring them to the market, it’s better to narrow down the functions to better meet customers’ needs.

Qualcomm RB3 announcement article Adobe StockWhen it comes to robots, the most important aspect is utility. You should always be looking to find the real world impact – for example, does it help people do their jobs better? Does having a robot ably perform certain tasks free people up to do more creative, complex work? Will people experience more job satisfaction if they no longer have to perform the same tedious, repetitive tasks all day long? On a more granular level, what are the specific tasks that can be automated in order to make people’s lives easier?

Even though each link in the whole-body chain is forged individually, the fact that each component can be seen as part of a greater framework ensures that we never lose sight of the bigger picture. However, it’s important to avoid the trap of reacting to trends or scrambling to catch up with competitors – the best way to do this is by perfecting the technology before thinking about how to incorporate it into a product. Once you’ve finally decided on the suitable application for a certain technology, think long and hard about the role that it will fulfill, as well as how it will interact with other people and the space around it. Not only do you have to make sure the robot has the knowledge needed to carry out its function, you also need to ensure that it fits in with its environment, and has the capabilities needed to perform its duty.

Taking a whole-body, whole-mind approach forces you to think on a grander scale. It also requires companies to spend more time on the research and development side before rolling out their products to the mass-market, which is likely why so many companies have been reluctant to apply this model.

But at the same time, the goal should be to create a new generation of robots, not just the next generation of robot arms or robot legs. By developing a platform that can bring together the skills, mobility, and senses of a person’s body, it is possible to create tools that will have significant impact, and make the world a smarter place.

About the author: Josh Ong is the director of marketing and communications for Cheetah Mobile, the fourth-largest app developer in the world.