Thoughts on Emotional AI

Leonardo da Vinci capturing the depths of human emotion.

Emotion can drag us through life, scraping every part our bodies against a hard, gray and despairing world. Emotion can also lift us to the highest peaks, refracting the neutral light of life into every vivid and radiant color imaginable.

Is it worth it? I don’t know, but I believe so. To me, experiencing the serene beauty of a fawn take its first steps in a blossoming spring meadow is worth feeling the heavy sadness of its inevitable death only a few years later. It’s taken many years and a lot of Dostoyevsky to come to this conclusion, but suffering and joy are interconnected, and I would rather know both than neither at all.

As you can see, emotion is complex. Part of me thinks it will always elude our understanding. The reasoning for this is twofold. As a rational being, I do not trust our ability to ever fully perceive reality. How can we map out complex territory when we do not even know what we cannot see? As an irrational being, I do not want to ‘solve’ emotion. I fear the emptiness resulting from catching the thing I have so adamantly chased, only to hold it cold and lifeless in my hands.

However, the deeper I dive into machine learning, the more I question the unassailable nature of human emotion. Having built models to classify emotion from human speech as well as from faces detected in video, I’ve had to reluctantly accept that emotional perception can be reduced to 0’s and 1’s. It does not feel good to accept the reductionality of a deeply human phenomona, but fruit grows best on a heavily pruned tree.

I should preface, I did not enter the world of emotional AI as a technological enthusiast. I’ve always favored the richness of the analog world and trusted it over its digital representation. For example, half of my tiny living room is taken up by an old upright piano. It cramps an otherwise open space. It weighs hundreds of pounds. It needs to be tuned. The middle F key sticks, and the wooden frame has given me splinters. An electric keyboard would solve all of these problems and if I’m being honest, I probably could not tell their sounds apart in a blind test. However, I want know the vibrations I feel are from a mallet striking a string and nothing more. This feels honest and true in a rapidly changing world.

My natural inclination is to sit by a wood stove far from the city reading Tolstoy, Jung, Nietzsche, Kierkegaard, and all my other favorite authors plunging as deeply as one can into the human condition, unafraid of using fact or fiction to explore ideas. I lived this way for awhile, but I always heard a nagging voice telling me that I was hiding from this modern world. I wanted to pick up the torch from these great philosophers and theorists, but they already crossed the frontier of understanding human depth. Our society needed them then, but we need something else now. I knew that if I truly wanted to carry on their work, I would have to embrace the next frontier of understanding, human-machine interaction.

That is why I am here, learning to program computers and studying machine-learning with a focus on emotional AI. We are emotional creatures and all of our commincation is drenched in emotion. Emotion leads to deeply human things like beauty, art, love, nuance, and humor. As interacting with machines is becoming just as common as interacting with other humans, I don’t want us to sacrifice our capacity for emotion. I don’t want a large portion of our daily comminication to be non-emotive and stale, life is too short for that.

I cannot quite tell what the future of emotional AI will look like. While the visions I see are still grainy, I know that there will be unimaginable benefits and harrowing dangers we will have to navigate. In building emotional AI, we will learn so much about ourselves and our lives will grow richer. We’ll have personalized coaches in our pockets asking just the right questions to lead us to more meaningful lives. We’ll have cars that can tell when we’re tired and take over the wheel, or accurately predict when we’re angry and cue a guided meditation. We’ll have 24/7 access to individualized therapists to talk those at-risk away from becoming another tragic opioid overdose.

Unlike humans, these machines will always listen and never judge. As a result, our companion applications will be so gratifying that we will become addicted to them. Our social skills will atrophy at a rate far faster than what we’ve already seen, especially in youth. Interacting with other humans is hard, so why risk rejection and discomfort when your phone stimulates your desire to be loved and known?

Not only this, but throwing emotional perception into the manipulative toolkit of the already powerful attention economy will only make putting the phone away even more difficult. We will buy more things we do not need, we’ll do more things we regret later, and we’ll waste more of our lives away. Even worse, we will remain unaware, stuck in a prison of empty pleasure. Orwell often takes the prize of the most prophetic dystopian author, but Huxley’s Brave New World is far more likely.

It is because this unavoidable dualism that I work on these projects. As we begin to interact with emotive, non-organic ‘beings’, this technology will force us to grapple with what it means to be human. It will make our lives better in practical and revolutionary ways. It will also cause catastrophes that we’ll be too blind to see until it’s too late. Through all this, I intend to have one ear listening to technological futurism, the other tuned to ethics and individual empowerment, and eyes looking straight ahead to navigate our society towards an ideal future.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s