Can AI-powered robots have a point of view?


Artificial intelligence has advanced leaps and bounds in just the last few years. We’re now in an age of Amazon Alexas, Google Assistants and all sorts of computer interfaces that are designed to communicate in a pseudo-human way. All of these technologies are fascinating and advanced, but for now, humans still have one key advantage over robots: the ability to improvise.

For example, Piotr Mirowski is a computer scientist living in London approaching this idea of robot improvisation in a very literal way. He and his colleague are the creators of A.L.Ex, an improv-performing AI system. He hosts a two-“person” show starring himself and A.L.Ex performing improv together.

Mirowski programmed A.L.Ex using a dataset comprised of millions of pieces of dialogue from movies, films, novels and TV. It picked up different phrases and language and learned how to piece together a response from the millions of previously recorded responses contained in its “memory.” It learns from every interaction it has during its improv shows and continues to hone its algorithm to more accurately replicate human speech.

I’ve watched some videos of their show, Human-Machine Live!, and well, as an improv performer, I can say the improv is terrible. A lot of what A.L.Ex says in response to Piotr’s lines are non-sequiturs that don’t advance the scene or resemble any sort of basic understanding of “Yes, and.” But that’s not really the point. As with all software, this program will eventually get better and will be able to have an interesting, life-like conversation.

But at that point, can we really call that improvisation? The way this software operates is not exactly in line with how human actors hear a line of dialogue, feel an emotion and then say the most logical response. Can we really call what A.L.Ex is doing the same (or better) as two humans improvising a scene together?

Point of View

In its simplest form, a “point of view” is a set of perspectives and deep-seated beliefs that together help to form a person’s opinions. Even if you’re not an improv performer, your “point of view” is visible every time you have a conversation.

When it comes to communicating, thinking about one’s point of view is a helpful way to prepare a person to speak about a topic, especially if he or she doesn't have a lot of time to prepare a script or plan remarks. Having a firm grasp of one’s opinions and perspectives on a topic will help guide a speaker to communicate his or her thoughts in an organized way.

This brings us back to the question at hand: Can robots have a point of view? The answer is unclear. Some scientists are looking into ways to program certain ethics or values into an AI robot’s coding. The idea is that if AI shares our values, the decisions it makes autonomously will have been generated through a process that takes into consideration the same things that a human would care about.

Anca Dragan, an assistant professor of computer science at UC Berkeley, was recently interviewed about AI values and had some interesting thoughts on the topic:

“Robots aren’t going to try to revolt against humanity, they’ll just try to optimize whatever we tell them to do. So we need to make sure to tell them to optimize for the world we actually want.”

The idea of robot values is something straight out of science fiction (see Asimov’s Laws for an example). There are countless movies (2001 and iRobot are two that jump to mind) where robots decide to wipe out humans because of a misdirected sense of purpose. Many data scientists today are thinking about this exact challenge. How can we make sure the AI systems we create will have the same values as the humans they’re designed to help?

All this brings us back to the idea of point of view. Can AI have an opinion? And would that mean a robot could have a personality that drives it to make surprising or unexpected choices?

These are deep questions that we don’t know the answer to right now. We can, however, begin to examine our own deeply held assumptions and gain confidence in our ability to communicate our values as humans.

Exercise: Point of View Questions

The next time you’re preparing to speak on a topic, whether it’s a small team check-in meeting or a large keynote speech, try asking yourself the follow questions. The answers to each of these will help you refine and hone your message:

  • Why you? Why does your expertise make you the most qualified person to explain this content?

  • Why them? What are the audience’s expectations and how will you satisfy or subvert them?

  • Why do you care? What about this topic ignites your passion? What will get your audience excited?

  • What action do you want to inspire? What is the one big takeaway that you must convey to the audience?

Hopefully, these rhetorical questions will help you clarify your point of view in order to give a clearer and more interesting presentation. Perhaps someday we’ll have a similar set of questions we can ask our AI colleagues, but for now, we can all sleep a little easier knowing that that day is still a few years away.