Lets play God
(Thoughts on design in the age of Machine Learning and Artificial Intelligence)
“What is the difference between man’s design and God’s design?”
This was the interview question that I had to face from Prof.Kirti Trivedi while seeking admission to IIT Bombay for Master of Design program. I’m not sure whether I understood the profoundness of that question at that time, but I tried to answer based on my observation of the natural world with all its complexity and its self-driven behavior. It went something like this : “God doesn’t create end-products but designs a system that finally ends up in the complex things we see.”
What makes our design different from what we see in nature is the fact that we always design and specify the end products. Our blueprints are really large sheets of paper that explains how machines should look and act like. But nature does it in a different way, instead of creating a blueprint of what a tree should look like in the end, it encodes the rules of growth into a tiny seed which then turns into a tree eventually.
We don’t design for all
How do we design now? We start by dividing people into ‘User groups’ or ‘target segments’. We say something like “this is the time of the ‘Millennials’ and we are designing for them”. We study them thoroughly, understand their needs and pain-points, empathize with them, and deduce some great insights specific to their context. And then we go ahead and create a few Personas who are the representatives of our particular ‘user groups’. We say “this is John Smith, and that’s Jane Doe; and I’m going to design for these averages, for their lifestyle, for their aspirations, their needs”. And in the process, we end up forgetting the rest of the people. We forget everyone except John Smith and Jane Doe!
But we don’t have a choice but to do this becuase when we do the math, designing for everyone is neither profitable nor viable. As a result, none of our users get the right solution in their hands. Instead, what they get is a compromise good enough.
All that we can do now is to hope that things will change; if not today, tomorrow.
We can’t design for every individual, but there’s hope
It’s only very recently that Machine Learning and Artificial Intelligence have entered our everyday life; enabling us to search from our thousands of mobile phone photographs just by describing its contents, providing quality translations, recognizing our handwriting with greater accuracy and being helpful in a lot more everyday stuff.
That’s because our computers have been getting this great “ability to learn without being explicitly programmed.” They have been learning from examples.
They are not only getting better at learning from what we teach them, but also by figuring out better ways of doing something all by themselves. Be it getting better at a game of Go by playing it so many times with itself, or inventing an intermediary language while translating between unknown languages.
Probably, someday, these machines will be able to understand our every need and adapt accordingly. Actually, many of the software products that we use everyday have this element of adaptations built into them already. We know very well how our Facebook and Twitter streams learn so quickly and adapt to our taste.
So, there’s hope. May be all that we need to do is to teach these machines how to make good design choices and let them design for each and every one.
Designers of tomorrow · Designing designing machines
Creating design blueprints for the average persona is not really the God-way of designing. Instead, we are here to design great seeds that when planted, grow gracefully, swaying in the direction where sunlight is more, gripping the roots if the air is windy, shedding its leaves when it’s winter and blooming when the spring shines.
Most certainly, in the near future, we are going to teach machines how to design. What we do in our design schools today to sensitize our students, we’re going to have to do that to our computers tomorrow. We will train them to have a ‘feeling’ for what’s more aesthetically pleasing. We will sensitize the machines about proportions, rhythm, and harmony. But not by dissecting what’s beautiful and finding out what constitutes beauty and then feeding it into the system. We won’t have to do that anymore. Instead, we will show our machines what’s good and let the machines get a hang of what’s beautiful, a hang of what’s good, and a hang of what’s right. We will help machines make intuitive decisions on what’s best. Much like an experienced human mind taking design decisions.
Afterall, these machines would know so much about every individual user, their preferences, and context, far better than any designer would ever know about their users, they would make the best design choice, every single time.
What next?
The time is not so distant for designers to shift from the current user-group centric design practices. When we have the capability to help everyone, we won’t have to stick with the profitable majority anymore.
Of course, designers will be out of their current jobs, and will take up something different instead. Rather than creating Interface guidelines and cropped image assets, designers are going to deal with something much more valuable… examples! We will be creating and curating great examples for AI to look at and learn from.
Let’s be ready to create things that get self-tailored for individuals; things that adapt themselves to specific situations.
Let’s teach machines how to design.
Let’s design designing machines.
Lets play God.
Image credits
Tree icon: Ani
Eiffel Tower icon: Atif ArshadFootnote
Well, when we understand evolution in its right sense, it’s clear that we don’t really need an intelligent designer-guy up there to make all these living beings we see around us. However, I believe it’s okay to use the God analogy for explaining the point about designing for AI.
