As if that wasn’t bad enough, machines have gained consciousness and superintelligence, and — against our will — they’ve taken over the world. With cold calculation, our AI overlords decide humans had their chance and that it’s about time to get rid of us before we do more damage.
Now rewind to June 2017, when delegates from around the world met in Geneva for a United Nations-hosted summit to design AI for global good. The goal wasn’t just to develop friendly AI but to devise ways to use the technology to make the world better for everyone. Naturally, there were plenty of cautionary tales about technology run rampant and how AI could make the world worse if we aren’t careful. But the overall message was one of hope.
It’s true: Humanity is facing more problems than it we can probably fix on our own. Without some drastic and immediate changes, we’re sure to usher in a dystopian future. But we may also be able to solve these problems — or at least minimize their negative impacts — with the help of AI. Here are some of the ways how.
Protecting our ocean by policing ourselves
It’s easy for us land-dwellers to forget just how vital the oceans are. They cover about 71 percent of the Earth’s surface and account for 91 percent of its living space. The oceans are where life began and our species has been linked to it ever since.
And yet, we’re doing a pretty poor job protecting this resource. The Great Barrier Reef isn’t yet dead but it’s dying off at a dangerous pace. Once vibrant and thriving communities of coral are turning into bleached graveyards. Despite regulations on the capture and sale of certain marine species, illegal fishing operations is still widespread.
Organizations like The Nature Conservancy (TNC) are now leveraging facial recognition software to fight overfishing in a bid to save the ocean. In November of last year it launched a contest that challenged software developers to create a system to monitor footage from fishing boats. The goal was to identify protected species so that inspectors can review the tape and make sure the fish are handled correctly and returned to the ocean.
This system is expected to drastically reduce the time spent policing fisheries. Inspectors usually spend some six hours analyzing every ten hours of tape, according to The Guardian. With an AI system tagging the minute mark where the suspected fish is on the film, that time could be cut by 40 percent.
“The end result is an incredible first step in moving us beyond what was currently thought to be impossible.”
“The winning team used computer vision and machine learning technology similar to what’s used for facial recognition,” Matt Merrifield, TNC’s chief technology officer, told Digital Trends. “The first layer of the models identify the region in the video that is most likely to have a fish present. The next layer actually identifies the species of the fish which requires training and deep learning with a more generic model. The end result is an incredible first step in moving us beyond what was currently thought to be impossible to an era of inevitable for using AI in fisheries monitoring.”
Other initiatives are already under way using AI to monitor illegal fishing activity. The website Global Fishing Watch tracks the fishing vessels around the world using data from nonprofit environmental watchdog SkyTruth, which mines satellite data to monitor the movements of big ships. An AI platform developed by Global Fishing Watch has identified over 86,000 cases in which fishing vessels performed potentially illegal actions at sea.
Predicting natural disasters
One of the best steps toward minimizing the impact of a natural disaster is predicting the event in the first place. It turns out that’s easier said than done.
For decades, scientists from a range of fields have tried and failed to reliably predict earthquakes with enough notice for the public to prepare. In the eighties and nineties, some even used machine learning, but couldn’t establish a reliable enough system, according to Scientific American. But AI has come a long way in the past few decades and today’s supercomputers allow scientists to crunch more data faster than ever before.
Scientists are now going back to machine learning to better understand earthquakes and predict when they’ll occur. If successful, the method could save hundreds of thousands of lives.
Researchers like Paul Johnson and Chris Marone, geophysicists at Los Alamos National Laboratory and the Pennsylvania State University respectively, have renewed interest in the potential for AI to predict earthquakes and they’re hoping it can help save lives.
“If we had tried this ten years ago, we would not have been able to do it,” Johnson told Scientific American. He is not only applying AI but is also approaching the problem of quake prediction differently.
“Hopefully decision makers of the future would be using these tools since they were children.”
Rather than using standard “earthquake catalogues,” which contain data only about magnitudes, locations, and times, Johnson and his team use huge datasets of measurements collected from artificial earthquakes that are constantly being simulated in a Penn State lab. The algorithms are tasked with analyzing this raw data — much of which seems superfluous — searching for patterns that might help predict a simulated quake.
The algorithms have already revealed that certain acoustic signals coincide with upcoming quakes. Within the simulator, tectonic plates creak like wooden floors as they slide over each other, and the system identified a particular change in that sound before the temblors occur. Although these sounds haven’t yet been observed in the natural world, Johnson and his team are listening closely.
“Not only could the algorithm tell us when an event might take place within very fine time bounds—it actually told us about physics of the system that we were not paying attention to,” he said. “In retrospect it was obvious, but we had managed to overlook it for years because we were focused on the processed data.”
There is still plenty of work to be done before scientists can reliably predict quakes but Johnson is now using real-world data with his algorithms. If the method works, he thinks experts could use it to make earthquake predictions months or years in advance.
Feeding the future
When it comes to feeding the globe, we’re facing a daunting task. The UN hopes to end hunger and all forms of malnutrition by 2030, which is optimistic considering that the world’s population is nearing the eight billion mark, and it’s expected to keep growing at least until 2050.
Even today we struggle feed everyone — one in nine people go to bed with an empty stomach each night, according to the World Food Programme.
But scientists at Carnegie Mellon University are developing a system called FarmView, which incorporates robotics and artificial intelligence to predict crop yield and hopefully make our food system more efficient.
FarmView works by mobilizing an autonomous ground robot that can take visual surveys of crops at different times of the season, including using computer vision and machine learning to predict crop yields. An algorithm then analyzes a particular plant and instructs the robot to clip away leaves or thin out fruit in order to facilitate a ratio for more optimal growth. Going one step further, the CMU researchers think AI could help geneticists identify and select for beneficial traits. In this way, AI would work together with breeders to produce more productive crops.
“If we had tried this ten years ago, we would not have been able to do it.”
“We’re not doing this to replace people,” said CMU system scientist George Kantor. “What we’re doing is to introduce new technologies that can make farmers more efficient at what they do, and allow them to use fewer resources to do it. The scenario we envision doesn’t involve using fewer people; it involves using robotics and other technologies to carry out tasks that humans aren’t currently doing.”
The main goal here is not just to produce more food but to use existing resources as efficiently as possible.
“The way we produce food right now is very resource intensive, and the resources that are available are being used up,” Kantor said. “We have to increase the amount of food we produce, as well as the quality, but do so in a way that doesn’t assume we have unlimited resources.”
An end to conflict?
One of the most ambitious plan for AI to save humanity comes from the mind of Timo Honkela, a professor at the University of Helsinki in Finland, who thinks technologies like machine learning and natural language processing could actually help eliminate conflict. He calls his concept the “Peace Machine” and it’s less farfetched than it sounds.
From Honkela’s point of view, there are three things we humans should really work on: our own emotions, our communication with others, and equality in society as a whole.
“We live in a complex world and we live complex lives that are culturally oriented and individually grounded in our experience,” he told Digital Trends. “So far, machines have been developed in a very rigid way. What’s not becoming possible is to make these systems to be more humanlike. My statement for a long time has been, ‘It’s better that we make machines to be humanlike because the other option is that we humans need to be machine-like in order to use these powerful tools.’”
Rather than claiming that AI can suddenly bring about world peace, Honkela thinks the technology can help in small ways that would have an emergent effect. For example, advances in machine translation can facilitate better communication between individuals from different backgrounds, minimizing misunderstanding and their subsequent conflicts, no matter how trite. From the bigger picture, all these resolved small conflicts would have an overall effect of creating a more agreeable society.
“The hypothesis is, if we have this situation in which we can understand each other better, that kind of naturally leaps in an emergent way to more peaceful relations overall,” Honkela said.
One of Honkela’s main points is that words are bound by meaning and context, which are not always clear. The phrases “My shirt is blue,” “I’m feeling blue”, and “I’m blue in the face,” each mean very different things that are difficult for a non-native English speaker to distinguish.
Of course, no wars have been fought over the word blue, but Honkela thinks this same system could be applied to every facet of communication.
“The further away people are in their experience of life, education, or cultural background, the more risk there is for miscommunication,” he said. “Even the words we use can mean different things to two different people.”
In the end, Honkela thinks everyone from school children to world leaders could have some sort of AI agent that could make sure they’re understanding correctly and speaking clearly.
“The basic idea is to use a device like a smartphone, whatever we have at hand, and it could say, ‘Christian what you just said would be understand quite differently than what you intend to mean,’” he said.
These devices may also be used to help people make more rational decisions calling out bias and emotional whim – a feature that would be ideal in today’s political climate. “Hopefully decision makers of the future would be using these tools since they were children,” Honkela said, so they will be better suited to address important issues without digressing into an emotional rant.
An end to war is still a distant dream. Indeed, some would argue that conflict is inherent – or even essential – to human nature. But perhaps AI can make these altercations more constructive by helping humans better understand each other. Maybe rather than wiping out humanity in some dystopian purge, AI will usher us into a new future in which we live together in harmony. That’s a future we’ll have to create ourselves.
