Norwegian A.I. Retreat
Last week I took a group of friends and colleagues over to the Norwegian Fjords for a 3 day retreat. We’d all been working hard the past year, and were feeling pretty burnt out. So everybody jumped at the chance to breath in the clean Nordic air, marvel and the beautiful surroundings, and get a sense of inspiration and perspective.
We’d booked the wonderful Juvet Hotel, a place I’ve wanted to stay ever since seeing it used as the location for Ex Machina. We needed an area of focus for the retreat, and considering the location, Artificial Intelligence seemed like the obvious choice.
Once the preserve of Science Fiction, A.I. is starting to weave its way into our lives. At the moment there’s a lot of hype and speculation; over active marketing teams trying to convince us that adding the letters “A" and “I" to a product instantaneously make it better. While that isn’t always the case, it's helping lots of start-ups attract more investment and increase their valuation.
At the same time there is also a lot of fear, uncertainty and doubt being generated by the media. Rarely a day goes by without some news story proclaiming an end of days scenario, only for that same publication to dismiss the fears as irrational a week later. It seems that if A.I is good for one thing, it’s boosting circulation.
With opinions fluctuating between utopia and dystopia, I was struggling to get an accurate view of the field. As such, one goal of the retreat was to gain a more realistic understanding of A.I. away from the media hyperbole. We did this by starting the retreat with a simple domain mapping exercise, to ensure that we all had a shared vocabulary and understood the general direction of travel.
My other hope was to bring more diversity of thought to the conversation, and break away from the usual circle of computer scientists and technocrats. Rather than domain experts, we were a group of interested parties, comprised of people from the arts, humanities, academia and design. We weren’t necessarily the people inventing this brave new world, but we would be the type of people called upon to make it palatable to the general public through story telling and design.
At the outset of most technological revolutions, the focus is understandably on the benefits it can bring. For industrial advances, these benefits are often in the form of reduced costs, increased productivity, and increased shareholder value. It’s only later that the social effects become clear.
As a group of user-centered designers and humanist technologists, it became evident that our interest lay in the effects A.I. would have on people and society as a whole, rather than the more immediate and obvious benefits to productivity and commerce. Over the course of around a dozen unconference style conversations, from brainstorming dystopian futures to discussing robot ethics, a pattern of concerns started to emerge. We’ve tried to capture the outputs of these discussions as a series of open questions, which we hope to share soon.
One big topic of discussion was the fear that A.I. and robotics may bring about large scale under-employment. Past technological revolutions have sparked similar fears, and humanity has always been able to adapt. Engines surpassed human power, production line technology automated mundane and repetitive tasks, and computers allowed people to outsource data storage and processing. Each time this has happened, we’ve be able to find new and meaningful work to replace the lost jobs, driving productivity ever forward. However could A.I. be difference? If we can finally outsource human cognition to the machine, is there anywhere left to go?
A related problem is the nature of the work we’ll end up doing as a result. A.I. has the ability to remove mundane tasks and let us focus on the fun and creative parts of our work. It also has the ability to create a generation of workers who’s sole job is to babysit machines, only stepping in when some sort of exception is thrown up. While this may be efficient, it’s not a great route to job satisfaction. As a result A.I. could very well eat into the middle of the jobs market, pushing some people up the skills ladder, and others down.
Another big problem was the realisation that as systems become more and more sophisticated, they become more difficult to understand. It’s no longer just a case of viewing source and checking the code, but also understanding the training data. If the training data is biased, because society can be biased, the results may be skewed and difficult to detect. This could result in new jobs like A.I. trainers and data bias consultants to ensure that new A.I.s are being fair with their decisions.
We briefly talked about robot skeuomorphism; how a lot of household robots are currently designed to look vaguely humanoid. This has certain benefits, such as signalling the robots capabilities to their users. If the robot has eyes, you presume it can see, if it has ears you presume it can hear, and if it has legs you assume it can walk. A lot of robots also seem to demonstrate rather childlike capabilities like big eyes and short stature, partly to communicate a level of simplicity and demonstrate that they aren’t a threat. At the moment the form is largely a consequence of engineers trying to create robots that can do similar tasks to humans by duplicating their movements. However over time I believe that robots will move away from the humanoid form and develop shapes which are better designed and more suited to their particular tasks.
We also touched on the area of ethics and morals. For instance should A.I.s be forced to adopt a human moral code, and if so, what would that actually be? If we did manage to create some form of super intelligence, would we have to grant them human-like rights, or could we still consider them a utility like a car or a toaster. If they were treated like utilities, wound’t that raise some rather uncomfortable questions?
On a slightly more mundane, but possibly more near-term scenario, should we encourage people to be polite towards A.I.s in the same way we are polite to people? Obviously the current crop of A.I.s wouldn’t really care if we say please or thank you. However by ignoring these common social behaviours, we may be baking in problems for the future. One of the attendees mentioned how they accidentally started talking to their partner like they were talking to their Alexa, while several parents noted their kids had started to behave in similar ways. Imagine a future where robots pets and helpers were common. Could we imagine a scenario where mistreating a robot pet change the way we treated actual animals? If so, would we eventually have to consider legislation to prevent robot abuse?
Obviously a lot of these questions are just thought experiments, and are a long way off at the moment. However I truly believe that when I’m old enough to need some form of home care, there’s a good chance I’ll be looked after in part by a robot. So many of these challenges may will be here In our lifetimes, and a few may arrive sooner than we think.
Of course the trip wasn’t just pondering theoretical questions. I was as much a holiday as anything else. So as well as lots of stimulating conversations over good food and slightly overpriced beer, we had plenty of time to enjoy our surroundings. This included a lovely forrest walk along King Olav’s Way, a hike up a mountain to visit a Glacier, and even a chance to see the Northern Lights. These shared experiences helped us bond as a group, while the beautiful scenery helped put things in perspective and provide the space to think.
Considering the short amount of time we had at our disposal, It’s amazing how much ground we managed to cover and how productive we felt by the end. We had started the retreat with the 20 of us sitting around discussing what we’d hoped to get out of the next two days. We finished on a similar note, explaining what we’d all taken away. Everybody had a different story to tell, but we all left with new connections, deeper friendships, a better understanding of the emerging field of A.I. and a newfound love of the Norwegian Fjords.
We captured these discussions as a series of questions about the future of AI, which we've called The Juvet Agenda. If this is a space that interestes you, it's well worth having a poke around.