I expected to feel a little out of place at the swanky university event for bioethics. My wife, a professor with expertise in the field, had invited me. But when I introduced myself as a software engineer to the attendees, many wanted to talk about one thing -- artificial intelligence (AI). How would it affect society? What are the goods and the bads?
Artificial intelligence has been popping up everywhere. At a recent family holiday event, a relative held a long conversation with my brother and I about how our companies --- Google and Agolo --- were destroying jobs by automating human tasks.
Artificial intelligence has entered the national conversation on a broad scale. 2017 could be the year that we start to see AI becoming part of our lives and causing some disruptive change.
Before we speculate on the potential downsides of increased automation due to AI, let’s consider some of the positives:
The U.S. Department of Transportation estimates that self-driving technology could save up to 94% of the 35,000 yearly vehicle deaths caused by human error. That number is more than the number of U.S. military casualties in Iraq and Afghanistan, and victims of domestic terrorism.
New consumer products that utilize AI can make us more efficient and enable us to potentially have more impact. Services such as a personal assistant or research assistant, which previously were only available to people in higher positions, are now available for a small price due to AI-powered services (see: X.ai and agolo.com).
AI may be able to shed light into problems that have been unsolvable - cancer treatment, virus control, how to optimize education, etc. An anecdote from the Go match between Fen Hui and Google’s AlphaGo engine illustrates this.
After playing a close match, AlphaGo played a move that surprised many of the people in attendance, including Fen Hui. “It’s not a human move. So beautiful.”
In the case of the game of Go, the machine is given a simple goal; win the match. But what about when AI faces complex decisions and trade-offs? That brings us to some of the concerns about AI - that despite the complexity of the algorithms, the biases and short-sightedness of the engineers can cause unforeseen ramifications.
If we look over the past year, we can see a few examples of this. Microsoft released an AI-powered chat bot, “Tay.” However, within a few hours Twitter users had figured out how to get the bot to repeat Nazi propaganda and images. It seems that the engineers hadn’t considered this “edge-case,” and the bot wasn’t prepared for the malicious attack.
Tesla released their autonomous driving technology, insisting that it was not meant to be fully autonomous. However, the reliance of the technology on light images may have caused the death of a driver in a fatal accident. Newer models are relying on radar technology as well as light images.
Besides the engineering errors that can seep through AI technology, there are also human concerns. What about the jobs that AI will replace? The self-driving truck startup Otto threatens to disrupt the jobs of 10’s of 1000’s. What will these people do from now on?
Every technical disruption has a darker side. The industrial revolution improved manufacturing efficiency 50-fold, but it also created an oversupply of workers, causing many to lose their livelihood. That’s actually where the term “luddite” comes from - Luddites were textile workers who revolted against the new technology.
Many have begun proposing solutions to the eventual automation onslaught caused by AI. Anti-technologists would like to prevent autonomous technologies from being adopted in order to protect jobs. Some technologists like Elon Musk have said that governments might have to establish a universal basic income to support displaced workers.
Just like complex problems faced by AI, our national attitude towards AI involves trade-offs. If we accept the technological progress, we also have to confront the job losses and need for re-training. If we shut out AI, we risk dropping behind as a leader in innovation and industry. The trick will be getting AI to play for us, not against us.
This article by Tom Goldenberg is part of the LinkedIn Top Voices list, a collection of the must-read writers of the year.