Monica Porteanu has written her third installment in our Emerging Fellows program. Here, she questions the effects of artificial intelligence on society. The views expressed are those of the author and not necessarily those of the APF or its other members.
“Will AI take over the world?” is a common question across many news outlets these days. “Artificial Intelligence will best humans at everything by 2060, experts say,” predicts one of them. “More than 70% of US fears robots taking over our lives, survey finds,” describes another. Most of all, “how long will it take for your job to be automated?” seems to be the question on everyone’s mind. Opposing views are also present, arguing about “The great tech panic: robots won’t take all our jobs.” How do we reconcile these views into what Artificial Intelligence is and can be?
The term “Artificial Intelligence” was coined in the 1950s, intending to describe the ability of machines to perform tasks at a human intelligence level. Today, the definition encompasses more nuanced meanings, especially when considering the level of human cognition. In this regard, there seem to be four categories: (1) automation; (2) machine learning using artificial neural networks; (3) deep learning; (4) and beyond.
Automation represents a low cognitive process that is repeatable, having well-defined sequences of actions that are pre-programmed into machine behaviour. The machine is a passive executor of what is being instructed to accomplish. Its ability to complete complex computations fast and without error is superior to humans. Automation can be applied on a large scale, with numerous examples from manufacturing production lines, to, more recently, interactions with customers, such as onboarding operations. It has the most concrete social impact, as it does take away jobs as we know them today. However, it also opens the opportunity for humans to do what they are better at than machines are: empathy, critical thinking, and creativity. The key to staying ahead of automation is, as Garry Kasparov puts it, “human ambition.”
Machine learning using artificial neural networks requires a more sophisticated, yet still moderate level of cognition. The machine can mimic repeatable but personalized activities, while learning from each interaction, and utilizing increasing amounts of data. It reacts to events based on what was instructed to be accomplished. In other words, it can present a solution to a problem as posed, recommend tasks, or take simple actions. For example, it can automatically set up preferences at home, adjust ambient environment parameters based on these preferences, turn appliances on/off, or keep track of our grocery list. This stage has developed in leaps and bounds during the last decade or so, achieving results in recognition and even digitization of image, face, or speech. However, the machine still has difficulty perceiving at a level comparable to a human. Although we are still irritated by recommendations gone wrong or irrelevant comments coming from the chat box, we allow this type of artificial intelligence into our lives, without yet understanding its concrete positive and negative impacts.
The leap to deep learning is the phase that debuted only a few years ago. With big visions at the forefront, deep learning aims to build capacity for a machine to solve problems without being told how. Such machines mimic the brain, through layers of artificial neurons that connect with and send signals to each other in the network. Initial results are astounding. For example, the machine has been able to beat humans at Go, the complex ancient Chinese game, whose number of alternative positions surpasses the atoms in the universe. However, it seems we have yet to uncover what is happening inside these deep neural networks. Scientists are currently investigating adversarial examples, in which the difference between what the human and machine sees is extreme (e.g., turtle versus gun).
Beyond deep learning is yet an area for even bigger dreams in which, perhaps, machines will surpass the human brain capacity, being able to create symbol systems (e.g., language, money, time, religion, governance) and with that, structurally alter every aspect of the life as we know it.
It seems we are now somewhere during the development of the second category, machine learning, and in the early stages of the third one, deep learning.
We have been warned that “Artificial Intelligence will best humans at everything by 2060.” With the many and contradicting opinions though, one could wonder, what will human capacity be in 2060? How will our brain functions evolve, and with that, where will our creativity, empathy, ambition, and critical thinking take us?
© Monica Porteanu 2018