36 year-old Michael Sayre is quite the handy man.

With a job in high school repairing electronics like TV’s, VCR’s, and all kinds of strange, esoteric equipment, Michael was fascinated by robotics from a young age. Graduating from university with a Bachelor’s in Engineering, this interest in robotics found itself transitioning from a childhood hobby to a full-time career.

“I knew I wanted to work with robots after I graduated!” said Michael.

Originally from the US, he moved to Singapore in 2007 to pursue a rather different American dream in the Lion City. Securing his first job working with ROVs (they are underwater vehicles that are remotely operated), he has seen many things in his 10 years of deploying robotic systems (even giant squid!).

Venturing once again into the world of start-ups, Michael is the founder of 4-month old Cognicept — a service that provides robotics-based human-in-the-loop (HITL) error handling systems. With a vision to bridge the limitations of artificial intelligence, what is he aiming to build?

The Limitations of Artificial Intelligence

Being exposed to science fiction as a child, I was always fascinated by the high efficiency and ability of artificial intelligence (AI) in movies like WALL-E and Star Trek. With the invention of machine learning and data labelling, machines can now learn from the existence of data itself and act with autonomy and accuracy. AI is now a reality, where life is made easier with its presence.

But, does it really?

“Cognicept solves a problem that I have personally encountered,” Michael says.

“I realized many others have faced the same difficulties from listening to my colleagues and reading about “horror stories” in the press.”

The recent UBER accident involving Elaine Herzberg in Arizona, USA, raised countless eyebrows regarding the limitations of AI. The mother of 2 was hit by a self-driving car when she attempted to cross the road with her bicycle — reason being the sensors’ failure to recognise her as a “real object”.

While not referencing this incident directly, Michael says that “such problems need to be solved if the promises of robotics are going to be realized on a large scale.”

And with bigger, better promises each year, it becomes even more pertinent to provide a solution to alleviate the Achilles’ heels involved.

The Human Side of AI

Cognicept is a service that resolves reliability issues for robotic systems in unstructured or unpredictable applications, such as in a warehouse, hotel, or even delivery services. Such widely variable environments, where one size does not fit all, results in a lot of machine intelligence failures.

How then, does Cognicept solve this problem?

Cognicept uses telepresence technology, which allows one to experience being in another location (think of it as Virtual Reality). Capitalising on that, they develop platform agnostic human-in-the-loop (HITL) error handling systems, giving them the opportunity to correct an error the moment it occurs.

“You could think of it as downtime insurance.” laughs Michael.

Borne out of his experience deploying robotics in a rapidly expanding e-commerce warehouse, he shares how the dynamic and chaotic environment there led to an important realization: a hybrid AI solution that allowed us to inject human intelligence on demand needed to be created.

In similar situations involving established MNCs, the robots act as components in a larger machine called a factory, and in tightly controlled environments such as semiconductor manufacturing, the AI in robots would work smoothly without a hitch. In a dynamic, unpredictable application however — it fails miserably.

Finding a way to ensure that both robotics and AI can cope with the ever-evolving changes in a start-up then became the challenge that Michael transformed into his brainchild. By incorporating a human element to solve problems immediately, it reduces the downtime drastically for machines, ensuring that work goes on as smoothly as it can for such complex environments.

A Brighter Future

With technological progress occurring at an unprecedented rate, breakthroughs in the tech scene are becoming a dime a dozen. However, not everyone has the time and resources to train their machines, one by one, function by function, to perform at the peak of their potential.

Michael has a way to do so.

“We can create AI training sets using our technology.” he says. “It isn’t economical to create training sets in a lab environment for every application of AI.”

Self-driving cars, for example, would be ground-breaking given that they work safely and effectively. The millions of dollars developers invest into collecting miles and miles of data seems worth it when you think of the many benefits such technology could have.

Spending endless hours in a lab training a robotic arm to pull things out of boxes is harder to justify.

“Our training sets are based on the resolutions our pilots have performed for systems that are already put in place. That way, we can incrementally improve systems that are not yet perfect.” shares Michael.

Aiming to close the gap between the physical and the virtual world might seem slightly farfetched, but with human operators behind the wheel, Michael believes robotics can do what couldn’t be done before.

“We have already seen massive changes as a result of computers and the internet. Robotics essentially allow those technologies to interact with the physical world more directly.”

And with Cognicept, we can now bridge the limitations of AI and the demands of robotics applications. Robotics has never seemed so accessible.

__

Want to hear Michael share more? Sign-up for our next Pizza and Chill {Code} on 18 July to bridge that virtual gap!

If you enjoyed this article, join the Wantedly Community on Facebook for more stories. Of course, sign up here to VISIT great companies!