Talking to Machines

All technological advancements came about as the result of a single human desire: to be lazy. Fire helped humans chew easier. The wheel helped humans lift with less effort. The abacus helped humans compute with more speed. These traditional technological inventions were all designed with one goal in mind: to act as tools that enable the human towards some action. Being ‘tools,’ machines have no rights or volition, freedom or intentionality. They exist for our purposes and our needs only. But what happens when our tools — machines — become intelligent? And, what are the implications of our relationships with these intelligent machines?

What is an intelligent machine?

Amazon.com

Before we dig into what makes a machine intelligent, let’s first take a look at one that is not. Take a look at the object above — you might list a variety of reasons for why it’s not intelligent, including:

• It is dependent: the human needs to manually make every selection for it to work
• It is static: it does not learn from previous interactions with the human
• It is limited: it can only do as much as the human instructs it to do, and no more
• It is predictable: there is a limited number of possible actions and outcomes

In other words, there is no symbiosis between the fax machine and the human given that one is completely dependent on the other. The machine is simply a tool — one with which we interact, in a unidirectional fashion.

Without machine intelligence: human —defined inputs — computation — output

Taking the flip side of the characteristics that make a machine not intelligent, we then have the following criteria for intelligent machines:

• It is independent: it understands the human intention of the computation being performed and can act with uncertain (nonspecific) inputs
• It is dynamic: it possesses the ability to incorporate uncertainties and variability during computation
• It learns: it remembers past computations, and learn from previous inputs and results
• Unlimited outputs: it can make predictions and recommendations based on its understandings of human intentions, and makes changes based on feedback.

With machine intelligence: human — defined & undefined inputs — computation and prediction — output — feedback & learning

An example of an intelligent system that teaches itself to recognize images based on human inputs and continuous feedback:

Implications for design

What differentiates an intelligent machine from one that is not rests in its ability to be dynamic and emphatic as illustrated above. To enable these characteristics of intelligent machines, our relationships with machines have changed as well. With intelligent machines, our interactions with them are no longer unidirectional. There is a continuous feedback loop between the machine and its user, with both benefitting from continuous interaction.

As designers in the era of intelligent systems, we can no longer create with the unidirectional mentality that dominated our relationships with machines. Our designs must take into account the ceaseless cycles of learning — of the machine trying to perfect itself — and use this advantage of customizability to the fullest advantage for our stakeholders.