Artificial Innovation: Teaching Computers to Speculate

When most people discuss robotics and artificial intelligence, the discussion of tasks and capabilities tends to lean towards the mundane. How can we teach robots to assemble parts? What kind of a recommendation engine can we build for movie watchers? How can we automate basic household tasks?

However, as our understanding of deep learning algorithms and cognitive computing continues to evolve, the limitations of our AI capabilities will very quickly become not our programming skills, but our imaginations. If we continue to only expect and ask for the mundane from designers of AI systems, we will continue to get the mundane.

But let’s give credit where credit is due – IBM has been playing with their Watson system in some interesting and exploratory ways. My personal favorite experiment is Chef Watson, a system that proposes non-intuitive recopies for new meals and cocktails based on the input of ingredients. The system works by deconstructing a sort of olfactory taxonomy of most ingredients – knowing the basic chemistry and flavor profile of just about anything you could throw its way – and, by understanding what flavor combinations we humans find pleasing, Watson proposes new recipes that may sound crazy on paper, but from a scientific perspective, should taste good (or at least interesting). Whether it’schicken in cocktails, blood sausage porridge, orchocolate burritos, IBM has found a unique way of proving that AI can do more than simply replicate and optimize human tasks; it can even out-innovate us.

And if we tear down the Watson Chef system to what it really is – a library of different inputs and an understanding of objectively “good” combinations – suddenly our eyes are opened to a very different world of AI. In this world, computer systems do not simply replicate and optimize the tasks that we do, but are able to explore the near infinite combinatorial possibilities of complex systems much like humans, only without the intuitive constraints of our humanity.

The lack of human intuition has long been touted by naysayers as one of the key shortcomings of robotics and AI, however, it may also be the thing that ultimately makes them better than us at tasks like speculative thinking, future exploration, and scenario planning. Without being limited by basic human logic and norms, these systems can explore ideas that our own instinct-driven minds wouldn’t even approach, for whatever deep-seeded constraints our humanity throws in the way. While 99% of these ideas will be completely random and likely terrible, if even a tiny fraction of them are stumbling across novel ideas with plausible value, we may very well find the holy grail of innovation.

Normally, this is the part of the article where I would write about all of the amazing things we could dream up with such a system, but the whole point of this article is that… I can’t. I’m too human to push the true boundaries of innovation and, unfortunately, if you want to keep finding chicken in your whiskey or sweets in your burrito, you’ll likely need to turn to either hard drugs or exploratory cognitive systems.

So, where does your industry turn to for impossible breakthroughs?

Shane Saunderson is the VP of IC/things at Idea Couture.