HOW SMART ARE YOUR MACHINES?

Each day we’re introduced to new “Smart Devices” in our day-to-day life. We ask SIRI to play a song, our GPS suggests the most efficient route to a place we’ve never been or our virus protection software protects us from a hacker while we sleep. Is this artificial intelligence?

While sophisticated, these examples are usually ALGORITHMIC. This means they are engineered with a sophisticated flowchart of “IF-THIS-THEN-DO-THIS” conditions that govern the actions they take. For instance:

These algorithms are rarely as simple as the above example.  As the tree of nested conditions and actions becomes more complex a machine may appear to have some intelligence because it is returning useful outcomes.  But is the machine, strictly speaking, smart?  

How does Artificial Intelligence work?

Artificial intelligence is an attempt to emulate the teleological processes humans and many animals use to figure things out.  

Here’s a very relatable example:  Say, you’re looking for a light switch in the dark, in a room you’ve never been in before.  You have some hypothetical ideal of where the light switch will be located. It’s likely on the inside of the room close to the door.  You reach out your hand and get a handful of wall.  You touch the space above that.  Nothing. You touch the space below the original space.  You find the switch and turn it on.  You mentally note the location of the light switch for next time.  

What if the target is or the data is changing over time?  Consider a more sophisticated example.  

In the case of a heat-seeking missile seeking a target, the pilot has taken aim on another aircraft.  Immediately after launching the missile the location of the target has changed.  The missile must detect the new location using infrared (heat) and adjust its heading accordingly and then immediately note the new location of the target relative to itself.  The missile continues to refine its trajectory until it finally hits the target.  (https://en.wikipedia.org/wiki/Infrared_homing)

In both the above examples there are some fundamental elements to note:

  • There is a target outcome
  • There is some hypothetical or seed data the system starts with 
  • There is a feedback system that captures data related to the target outcome
  • The system has instructions for evaluating the data 
  • The system corrects its process to refer to in future iterations.

The most sophisticated model of this kind of problem-solving can be observed as a baby learns to feed itself.  The spoon hits the check.  The child tries again.  The spoon hits the other cheek.  The child tries again.  Closer to a direct hit.  The spoon goes in.  Now we just have to figure out how to eat peas…   

In a true artificial intelligence system problem solving doesn’t need to be hampered by any fixed assumptions.   The system can challenge or retry,  optimize or respond to changes in any part of the above fundamental elements list as the shifting situation requires.  

For example: Suppose a machine is performing in an optimal way.  One day some environmental change such as temperature or humidity changes characteristics of the material hampers the range of motion of on one part of the system.  A true artificial intelligence system would be able to adapt to the changing conditions and potentially correlate the need to adapt to the observed change in conditions, so it doesn’t have to “figure out” this adaptation the next time the conditions change in the same way.

There are infinite applications for this technology in manufacturing.  True artificial intelligence is poised to revolutionize performance and efficiency manufacturing.  It’s much easier to stay ahead than to catch up.   Give us a call today to discuss how your operation can be on the leading edge of this exciting technology.

When we talk about true artificial intelligence, we are describing a system that may start with some primitive task flow or data like the above example, but improves its performance over time.  A machine governed by an algorithm runs best on its first day but requires maintenance and human attention to maintain or optimize performance.  A machine governed by true artificial intelligence performs better as it runs.