The fear of robots killing us all is a perennial futurist's favourite, but in the digital rather than physical world invisible bots have been waging wars for years.
Beneath the pages of Wikipedia, legions of digital automatons work in silence maintaining the online encyclopedia by checking spelling, undoing vandalism and creating links. Conflicts over the correct edits can lead them to feud for years, according to recent research by the University of Oxford and the Alan Turing Institute published in PLOS ONE.
The contradictions occur when multiple bots believe in different outcomes. One bot may want to change everything on Wikipedia into American English, while another might prefer the British variety, leading the two to fight incessantly over the spelling of the word "colour". At times different bots correct each other thousands of times through several years.
The researchers studied every edit made in 13 language editions of Wikipedia across 10 years since the encyclopedia was launched in 2001. The interactions between bots varied in the different versions.
The Germans were predictably efficient, undoing each other's edits an average of 24 times per bot, the fewest on the list. At the other end were the Portuguese, at an average of 185 reverts over ten years. Rather than confirming any stereotypes, the difference is likely due to linguistic characteristics forcing bots on the Portuguese Wikipedia to make more edits than those on the German edition.
"The fights between bots can be far more persistent than the ones we see between people," researcher Dr Taha Yasseri told the Guardian. "Humans usually cool down after a few days, but the bots might continue for years."
Their conflicts may seem harmless enough but greater peril lurks ahead, as the internet of things gives warring bot factions greater control over our lives.
"We're going to see this problem get much worse as we enter smart homes and smart buildings and smart cities where there are actual real physical objects being controlled that might affect the person's life, and not just a Wikipedia page,” David Moss, the CTO and founder of IoT software company People Power, tells Techworld.
It may be a website today, but tomorrow they could be fighting over the setting of your thermostat, the brightness of your light bulbs, or even the lock on your door.
Negotiating a truce
From its Palo Alto base in Silicon Valley, People Power is bringing peace to the bot wars. The IoT software company has developed a system that enables bots to privately communicate without exposing user information.
"If you're going to build a platform that supports bots you need to be thinking ahead about how do these bots interact with each other. I think there are four things that you need to build into a platform, and I think Wikipedia supports two out of these," says Moss.
"You have to have a platform that enables best practices; some automated testing and an Apple-like review process. Wikipedia has that. They have the ability to register your bots on Wikipedia. It goes through a review committee and that kind of testing period. That's a good start.
"The second thing is accountability and auditability. On Wikipedia when your bot makes changes it actually leaves a comment about what changes were made, so it's auditable. The other two items I believe Wikipedia hasn't solved yet."
They are firstly a system that supports prioritising bots, over decisions such as smart home services, which the provider could set, and secondly the ability for bots to talk directly with each other and coordinate their actions.
"We call this flocking," says Moss. "Bots of a feather flock together. They should all work together to produce the outcome that's meaningful for the end user who's going to experience what those bots can do."
The People Power platform completes the quartet by adding inter-bot communications and a prioritisation system to Wikipedia's combination of best practices of review process and in-built auditability.
Enemies under the same roof
Bots can automatically add intelligence to appliances in the home, but some may have goals that contradict those of others. In the case of lights, one bot might want to save energy and therefore power down your lights while you're away from home.
Another bot responsible for security might want to make it appear as you're home by turning your lights on and off at the times you normally would if you were there. A third bot might be programmed to take of your pets, and leave the lights on for it during its waking hours.
The three bots all want to manage the lights in a different way. People Power would prioritise the needs of a living animal over saving energy and preserving security, and ensure that the bots coordinate their actions to protect this need. They can learn daily habits of users and make decisions based on the patterns of behaviour. A referee bot can observe the user account to resolve conflicts, and ask the customer a question if their opinion is needed.
The company recently launched a white-label home security solution called People Power Pro Security, which integrates AI through bots to help companies rapidly deploy bots to deliver smart home services. It's also provided developer access to both its new Bot Server, a cloud-computing environment in which to run bots, and the Bot Lab, which allows anyone to create bots for their buildings on the People Power platform in Python and run them privately in the company's cloud.
People Power also offers the opportunity to send new bots for review to ensure they conform to People Power's best practices. Once published, other users can add the bot to their account. After the user provides permission the bot listens for data from devices such as thermostats and lights and the developer can no longer access the information.
"This bot is now running 24/7 in the background of your life, listening to these real-time streams of data and influencing outcomes," says Moss.
From apps to bots
People Power was founded by Moss and CEO Gene Wang in 2009. The company started working on bots around five years ago, and currently supports services for a number of global consumer service brands, including the world's largest telecom company China Mobile.
In the near future Moss envisages bots playing a vital role in managing energy, security, and senior care. He believes that mobile apps are on a decisive decline and that People Power are in prime position to capitalise on their coming replacement by bots by providing a platform to help companies quickly spin up new bot services.
"What we're seeing now in the case of Wikipedia is an example of a much bigger problem that is to come as the entire market transitions from apps into bots," he says.
"We believe that the internet of things is really not about things. That's the biggest secret of this market. The value of the internet of things is not in the things, and it's really not even in the connectivity to those things.
"As a software company, we understand very deeply that the real value of this internet of things market is in the services and the experiences that are created by having access to lots of data. We've found directly that creating a bot and promotionally launching a bot service is 16 times faster than creating a mobile app. The reason why is because bots don't have a fancy user interface on top. it goes beyond your screens. This is something that can exist on devices that don't have a screen and really operate on behalf of those devices.
As the market transitions from apps into bots, Moss anticipates that independent developers will use the People Power platform to create their own services. The company’s IoT suite helps them quickly create their own smart home products.
"Developers have this opportunity now to come in and create little microservices adding features and services to products that the manufacturer never imagined," he says. "This is one of the biggest opportunities in the market today as developers worldwide are moving away from mobile apps onto something bigger.
"In the past when you'd download an app to your phone it's sort of like adding a feature to your phone, and you may pay 99 cents for example for a new feature on your phone. But bots are looking at the real-time data streams 24/7 and providing services, and this should be really interesting for developers because it justifies more of a service fee and not a one-time download fee."
It will also offer the opportunity to democratise development in an open ecosystem that allows independent operators to find their own solutions. An example of a service that would benefit from such an approach is the first learning thermostat developed by Nest.
"A lot of times I hear people who say well it's really not all that intelligent," says Moss. "In fact the thermostat sort of fights with them, it has a little bit of an arrogant personality in that it thinks that it knows your temperature better than you do.
"What if a developer could make a better learning thermostat algorithm, and democratise this so that there's a bunch to choose from. Now instead of just having to settle for whatever algorithm Nest created for you, you get to choose the one you want. Do you want the Oxford algorithm, do you want the Stanford or Berkeley algorithm, do you want the algorithm made by a company focused on energy efficiency and saving the environment?"
People Power's bot services are designed to transform the current market "from the smart home to the conscious home".
"Smart homes aren't good enough," says Moss. "We want to add thoughts to things. That's the biggest opportunity in this market right now, and you add thoughts to things through bots. Bots are the vehicle that delivers artificial intelligence and machine learning solutions into this market.
"In order to cross the chasm into mass adoption, you need something more intuitive. Having the ability to select an outcome that's delivered by a bot and turn it on and off and have the bot do the rest of the work for you is very important in this market. That's going to shift us from basic smart homes of the past to homes that are conscious and predict what you want your home to feel and act like, without having to be bogged down with these clumsy binary 'if' statements like you've had to in the past.
"With the data flowing from the internet of connected things today, we actually have the ability to learn a person's patterns and behaviours and know if those patterns are changing unexpectedly. For example, senior care is a domain for that. You want to know if the senior has fallen down or not. You can't do that with an ''if' statement and you can't do that without adding bots to all of your things. Your home has to become people-conscious."