Wednesday, June 15, 2016

Robots Aren't Just For Cost Savings

A new burger place opened in my town and I was very excited to try it.  I went for lunch and had one of the best burgers I have ever eaten.  The bun was perfectly buttered and toasted on the grill so it crunched just a little with each bite, the meat was perfectly warm, pink and juicy in the middle and the special sauce, lettuce tomato took it over the top.

I gathered the rest of my family and headed back a few days later, telling them this place was amazing.  Unfortunately, my bun was not perfectly toasted, my meat was not perfectly cooked, it wasn't warm, and the special sauce couldn't save the overall poor execution on the rest of the burger.

I blame humans for this problem. A robot would have been able to provide a more consistent level of product quality by using precise measurements for cooking temperature and time, understanding all the items that are related in an order to bring the entire order together with precision.  A simple survey after the meal would allow the machine to use machine learning for continuous improvement.

A lot of people talk about robots and automation as a way to reduce cost for large corporations by replacing people with machines.  But, there is another, potentially more important upside to robots and automation and that is service consistency, product quality and topline revenue impacts. I don't think our burger place is about to buy a Momentum Machines burger robot, but it illustrates the point.

A few technology trends are compounding to make automation and robots more broadly applicable and these trends could very likely cause an explosion in real world applications of robots.  The first is the dramatic increase in computing power and speed, the second is the dramatic reduction in the cost of storing data.  Both of these trends are true in small form factors and at massive scale.

A very small robot can house terabytes of data and process billions of instructions per second.  A consolidated cloud environment can house limitless data at pennies per gigabyte and has limitless, scale-out computing power to analyze all that data, instantly.

This allows a robot to do things that a human can't.  It can analyze more complicated interdependencies in realtime.  It can execute with more perfect precision than a human can.  And, crazily enough, it can learn and continually improve, called machine learning.

Machine learning has a couple of modes.  The first is supervised machine learning.  That's where you tell the machine what the organizing principle or goal is, and the machine categorizes things into the model defined by a human.  The other is where you just give a machine a bunch of data points and the machine categorizes them into groups and can continually refine the categorization. (Interested in more on machine learning, check this out)

The more data we can give a machine about the situation at the time of an action and the more data we can give the machine about the output of that action, the faster the machine will learn and improve.  At some point, humans can't take any more data about the situation or output and so we continually repeat sub-optimal actions.  Robots typically won't do that - they will learn.

Robots are coming. They are coming soon.  And they will be able to make our lives better and reduce the cost of just about everything.  That's not really a question.  The real question we need to wrestle with is how our society will adjust to a world where we can automate almost every aspect of our lives and economy.

No comments:

Post a Comment