War in the Fifth Domain

In the months running up to the 2016 election, the Democratic National Committee was hacked. Documents were leaked, fake news propagated across social media — the hackers, in short, launched a systematic attack on American democracy.

Whether or not that’s war, however, is a matter for debate. In the simplest sense, an act of cyber warfare is defined as an attack by one nation on the digital infrastructure of another.

These threats are what Samuel Woolley, research director of the Digital Intelligence Lab at Institute for the Future, calls “computational propaganda,” which he defines as the spread of disinformation and politically motivated attacks designed using “algorithms, automation, and human curation,” and launched via the internet, particularly social media. In a statement to Futurism, Woolley added that these attacks are “assailing foundational parts of democracy: the press, open civic discourse, the right to privacy, and free elections.”

Attacks like the ones preceding the 2016 election may be a harbinger of what’s to come: We are living in the dawn of an age of digital warfare — more pernicious and less visible than conventional battles, with skirmishes that don’t culminate in confrontations like Pearl Harbor or 9/11.

Our definitions of warfare — its justifications, its tactics — are transforming. Already, there’s a blurry line between threats to a nation’s networks and those that occur on its soil. As Adrienne LaFrance writes in The Atlantic: an act of cyber warfare must be considered an act of war.

A War of 0s and 1s

A little over a decade ago, the United States Cyber Command began developing what would become the world’s first digital weapon: a malicious computer worm known as Stuxnet. It was intended to be used against the government of Iran to stymie its nuclear program, as The New York Times reported. In the true spirit of covert operations and military secrecy, the U.S. government has never publicly taken credit for Stuxnet, nor has the government of Israel, with whom the U.S. reportedly teamed up to unleash it.

Stuxnet’s power is based on its ability to capitalize on software vulnerabilities in the form of a “zero day exploit.” The virus infects a system silently, without requiring the user to do anything, like unwittingly download a malicious file, in order for the worm to take effect. And it didn’t just run rampant through Iran’s nuclear system — the worm spread through Windows systems all over the world. That happened in part because, in order to enter into the system in Iran, the attackers infected computers outside the network (but that were believed to be connected to it) so that they would act as “carriers” of the virus.

As its virulence blossomed, however, analysts began to realize that Stuxnet had become the proverbial first shot in a cyber war.

Like war that takes place in the physical world, cyber warfare targets and exploits vulnerabilities. Nation-states invest a great many resources to gather intelligence about the activities of other nations. They identify a nation’s’ most influential people in government and in general society, which may come in useful when trying to sway public opinion for or against a number of sociopolitical issues.

Gathering nitty-gritty details of another country’s economic insecurities, its health woes, and even its media habits is standard fare in the intelligence game; figuring out where it would “hurt the most” if a country were to launch an attack is probably about efficiency as much as it is efficacy.

Historically, gathering intel was left to spies who risked life and limb to physically infiltrate a building (an agency, an embassy), pilfer documents, files, or hard drives, and escape. The more covert these missions, and the less they could alarm the owners of these targets, the better. Then, it was up to analysts, or sometimes codebreakers, to make sense of the information so that military leaders and strategists could refine their plan of attack to ensure maximum impact.

The internet has made acquiring that kind of information near-instantaneous. If a hacker knows where to look for the databases, can break through digital security measures to access them, and can make sense of the data these systems contain, he or she can acquire years’ worth of intel in just a few hours, or even minutes. The enemy state could start using the sensitive information before anyone realizes that something’s amiss. That kind of efficiency makes James Bond look like a slob.

In 2011, then-Defense Secretary Leon Panetta described the imminent threat of a “cyber Pearl Harbor” in which an enemy state could hack into digital systems to shut down power grids or even go a step beyond and “gain control of critical switches and derail passenger trains, or trains loaded with lethal chemicals.” In 2014, TIME Magazine reported that there were 61,000 cybersecurity breaches in the U.S that year; the then-Director of National Intelligence ranked cybercrime as the number one security threat to the United States that year, according to TIME.

Computer viruses, denial of service (DDS) attacks, even physically damaging a power grid — the strategies for war in the fifth domain are still evolving. Hacking crimes have become fairly common occurrences for banks, hospitals, retailers, and college campuses. But if these epicenters of a functioning society are crippled by even the most “routine” cybercrimes, you can only imagine the devastation that would follow an attack with the resources of an enemy state’s entire military behind it.

Nations are still keeping their cards close to their chest, so no one is really certain which countries are capable of attacks of the largest magnitude. China is a global powerhouse of technology and innovation, so it’s safe to assume its government has the means to launch a large-scale cyber attack. North Korea, too, could have the technology — and, as its relationship with other countries becomes increasingly adversarial, more motivation to refine it. After recent political fallout between North Korea and China, Russia reportedly stepped in to provide North Korea with internet— a move that could signal a powerful alliance is brewing. Russia is the biggest threat as far as the United States is concerned; the country has proven itself to be both a capable and engaged digital assailant.

The Russian influence had a clear impact on the 2016 election, but this type of warfare is still new. There is no Geneva Convention, no treaty, that guides how any nation should interpret these attacks, or react to them. To get that kind of rule, global leaders would need to look at the ramifications for the general population and determine how cyberwar affects citizens.

At present, there is no guiding principle for deciding when (or even if) to act on a perceived act of cyberwarfare. A limbo that is further complicated by the fact that, if those in power have benefitted from, or even orchestrated, the attack itself, then what incentive do they have to retaliate?

If cyber war is still something of a Wild West, it’s clearly citizens who will become the casualties. Our culture, economy, education, healthcare, livelihoods, and communication are inextricably tethered to the internet. If an enemy state wanted a more “traditional” attack (a terrorist bombing or the release of a chemical agent, perhaps) to have maximum impact, why not preface it with a ransomware attack that freezes people out of their bank accounts, shut down hospitals and isolate emergency responders, and assure that citizens didn’t have a way to communicate with their family members in a period of inescapable chaos?

As cybersecurity expert and author Alexander Klimburg explained to Vox, a full-scale cyber attack would result in damage “equivalent to a solar flare in terms of damaging infrastructure.” In short, it would be devastating.

A New Military Strategy

In summer 2016, a group called the Shadow Brokers began leaking highly classified information about the arsenal of cyberweaponry at the National Security Agency (NSA), including cyber weapons actively in development. The agency still doesn’t know whether the leak came from someone within the NSA, or if a foreign faction infiltrated Tailored Access Operations (the NSA’s designated unit for cyber warfare intelligence-gathering).

In any case, the breach of a unit that should have been among the government’s most impervious was unprecedented in American history. Aghast at the gravity of such a breach, Microsoft President Brad Smith compared the situation “to Tomahawk missiles being stolen from the military,” and penned a scathing blog post calling out the U.S. government for its failure to keep the information safe.

The last time such a leak shook the NSA, it was in 2013, when Edward Snowden released classified information about the agency’s surveillance practices. But as experts have pointed out, the information the Shadow Brokers stole is far more damaging. If Snowden released what was effectively battle plans, then the Shadow Brokers released the weapons themselves, as the New York Times analogized,

Earlier this year, a ransomware attack known as “WannaCry” began traversing the web, striking organizations from universities in China to hospitals in England. A similar attack hit IDT Corporation, a telecommunications company based in Newark, New Jersey, in April, when it was spotted by the company’s global chief operations officer, Golan Ben-Oni. As Ben-Oni told the New York Times, he knew at once that this kind of ransomware attack was different than others attempted against his company — it didn’t just steal information from the databases it infiltrated, but rather it stole the credentials required to access those databases. This kind of attack means that hackers could not only take that information undetected, but they could also continuously monitor who accesses that information.

WannaCry and the IDT attack both relied upon the cyber weapons stolen and released by the Shadow Brokers, effectively using them against the government that developed them. WannaCry featured EternalBlue, which used unpatched Microsoft servers to spread malware (North Korea used it to spread the ransomware to 200,000 global servers in just 24 hours). The attack on IDT also used EternalBlue, but added to it another weapon called DoublePulsar, which penetrates systems without tripping their security measures. These weapons had been designed to be damaging and silent. They spread rapidly and unchecked, going undetected by antivirus software all over the world.

The weapons were powerful and relentless, just as the NSA intended. Of course, what the NSA had not intended was that the U.S. would wind up at their mercy. As Ben-Oni lamented to the New York Times, “You can’t catch it, and it’s happening right under our noses.”

“The world isn’t ready for this,” he said.

The Best Defense

The average global citizen may feel disenfranchised by their government’s apparent lack of preparedness, but defending against the carnage of cyber warfare really begins with us: starting with a long overdue reality check concerning our relationship with the internet. Even if the federal agencies aren’t as digitally secure as some critics might like, the average citizen can still protect herself.

“The first and most important point is to be aware that this is a real threat, that this potentially could happen,” cybersecurity expert Dr. Eric Cole told Futurism. Cole added that, for lay people, the best defense is knowing where your information is being stored electronically and making local backups of anything critical. Even services like cloud storage, which are often touted as being safer, wouldn’t be immune to targeted attacks that destroy the supportive infrastructure — or the power grids that keep that framework up and running.

“We often love going and giving out tons of information and doing everything electronic,” Cole told Futurism, “but you might want to ask yourself: Do I really want to provide this information?”

 Some experts, however, argue that your run-of-the-mill cyber attack against American businesses and citizens should not be considered an act of war. The term “war” comes with certain trappings — governments get involved, resources are diverted, and the whole situation escalates overall, Thomas Rid, professor and author, recently told The Boston Globe. That kind of intensity might, in fact, be counterproductive for small-scale attacks, ones where local authorities might be the ones best equipped to neutralize a threat.

As humans evolve, so too do the methods with which we seek to destroy each other. The advent of the internet allows for a new kind of warfare — a much quieter one. One that is fought remotely, in real time, that’s decentralized and anonymized. One in which robots and drones take the heat and do our bidding, or where artificial intelligence tells us when it’s time to go to war.

Cyber warfare isn’t unlike nuclear weapons — countries develop them in  }and, should they be deployed, it would be citizens that suffer more than their leaders. “Mutually assured destruction” would be a near guarantee. Treaties mandating transparency have worked to keep nuclear weapons in their stockpiles and away from deployment. Perhaps the same could work for digital warfare?

We may be able to foretell what scientific and technological developments are on the horizon, but we can only guess at what humanity will do with them.

Humans made airplanes. These allowed them to fly above the clouds…and they used it to drop bombs on each other.

No More Heart Transplants

Around the world, lists of patients in need of an organ transplant are often longer than the lists of those willing (and able) to donate — in part because some of the most in-demand organs for transplant can only be donated after a person has died. By way of example, recent data from the British Heart Foundation (BHF) showed that the number of patients waiting for a heart transplant in the United Kingdom has grown by 162 percent in the last ten years.

Now, 50 years after the first successful heart transplant, experts believe we may be nearing an era where organ transplantation will no longer be necessary. “I think within ten years we won’t see any more heart transplants, except for people with congenital heart damage, where only a new heart will do,” Stephen Westaby, from the John Radcliffe Hospital in Oxford, told The Telegraph.

Westaby didn’t want to seem ungrateful for all the human lives saved by organ transplants, of course. On the contrary, he said that he’s a “great supporter of cardiac transplantation.” However, recent technological developments in medicine may well offer alternatives that could save more time, money, and lives.

“I think the combination of heart pumps and stem cells has the potential to be a good alternative which could help far more people,” Westaby told The Telegraph.

An Era of Artificial Organs

Foremost among these medical advances, and one that while controversial has continued to demonstrate potential, is the use of stem cells. Granted, applications for stem cells are somewhat limited, though that’s down more to ethical considerations more than scientific limitations. Still, the studies that have been done with stem cells have proven that it is possible to grow organs in a lab, which could then be implanted.

Science has also made it possible to produce artificial organs using another technological marvel, 3D printing. When applied to medicine, the technique is referred to as 3D bioprinting — and the achievements in the emerging technique have already been quite remarkable.

Thus far, scientists have successfully 3D-bioprinted several organs, including  a thyroid gland, a tibia replacement that’s already been implanted into a patient, as well as a patch of heart cells that actually beat. All of these organs were made possible by refinements to the type of bioink; one of many improvements to the process we can expect to see in the years to come, as there’s now an institution dedicated to advancing 3D bioprinting techniques.

Other technologies that are making it possible to produce synthetic organs include a method for growing bioartificial kidneys, the result of a study in 2016.

For his part, Westaby is involved in several projects working to continue improving the process: one uses stem cells to reverse the scarring of heart tissue, which could improve the quality of life for patients undergoing coronary bypass. Westaby is also working on developing better hardware for these types of surgical procedures, including inexpensive titanium mechanical heart pumps.

Together with 3D bioprinting such innovations could well become the answer to donor shortages. The future of regenerative medicine is synthetic organs that could easily, affordably, and reliably be printed for patients on demand.

Quark Quirks Called Tetraquarks

Everything in the universe is made up of atoms — except, of course, atoms themselves. They’re made up of subatomic particles, namely, protons, neutrons, and electrons. While electrons are classified as leptons, protons and neutrons are in a class of particles known as quarks. Though, “known” may be a bit misleading: there is a lot more theoretical physicists don’t know about the particles than they do with any degree of certainty.

As far as we know, quarks are the fundamental particle of the universe. You can’t break a quark down into any smaller particles. Imagining them as being uniformly minuscule is not quite accurate, however: while they are tiny, they are not all the same size. Some quarks are larger than others, and they can also join together and create mesons (1 quark + 1 antiquark) or baryons (3 quarks of various flavors).

In terms of possible quark flavors, which are respective to their position, we’ve identified six: up, down, top, bottom, charm, and strange. As mentioned, they usually pair up either in quark-antiquark pairs or a quark threesome — so long as the charges ( ⅔, ⅔, and ⅓ ) all add up to positive 1.

The so-called tetraquark pairing has long-eluded scientists; a hadron which would require 2 quark-antiquark pairs, held together by the strong force. Now, it’s not enough for them to simply pair off and only interact with their partner. To be a true tetraquark, all four quarks would need to interact with one another; behaving as quantum swingers, if you will.

“Quarky” Swingers

It might seem like a pretty straightforward concept: throw four quarks together and they’re bound to interact, right? Well, not necessarily. And that would be assuming they’d pair off stably in the first place, which isn’t a given. As Marek Karliner of Tel Aviv University explained to LiveScience, two quarks aren’t any more likely to pair off in a stable union than two random people you throw into an apartment together. When it comes to both people and quarks, close proximity doesn’t ensure chemistry.

“The big open question had been whether such combinations would be stable,
or would they instantly disintegrate into two quark-antiquark mesons,” Karliner told Futurism. “Many years of experimental searches came up empty-handed, and no one knew for sure whether stable tetraquarks exist.”

Most discussions of tetraquarks up until recently involved those “ad-hoc” tetraquarks; the ones where four quarks were paired off, but not interacting. Finding the bona-fide quark clique has been the “holy grail” of theoretical physics for years – and we’re agonizingly close.

Recalling that quarks are not something we can actually see, it probably goes without saying that predicting the existence of such an arrangement would be incredibly hard to do. The very laws of physics dictate that it would be impossible for four quarks to come together and form a stable hadron. But two physicists found a way to simplify (as much as you can “simplify” quantum mechanics) the approach to the search for tetraquarks.

Several years ago, Karliner and his research partner, Jonathan Rosner of the University of Chicago, set out to establish the theory that if you want to know the mass and binding energy of rare hadrons, you can start by comparing them to the common hadrons you already know the measurements for. In their research they looked at charm quarks; the measurements for which are known and understood (to quantum physicists, at least).

Based on these comparisons, they proposed that a doubly-charged baryon should have a mass of 3,627 MeV, +/- 12 MeV. The next step was to convince CERN to go tetraquark-hunting, using their math as a map.

Smashing Atoms

For all the complex work it undertakes, the vast majority of which is nothing detectable by the human eye, The Large Hadron Collider is exactly what the name implies: it’s a massive particle accelerator that smashes atoms together, revealing their inner quarks. If you’re out to prove the existence of a very tiny theoretical particle, the LHC is where you want to start — though there’s no way to know how long it will be before, if ever, the particles you seek appear.

It took several years, but in the summer of 2017, the LHC detected a new baryon: one with a single up quark and two heavy charm quarks — the kind of doubly-charged baryon Karliner and Rosner were hoping for. The mass of the baryon was 3,621 MeV, give or take 1 MeV, which was extremely close to the measurement Karliner and Rosner had predicted. Prior to this observation physicists had speculated about — but never detected — more than one heavy quark in a baryon. In terms of the hunt for the tetraquark, this was an important piece of evidence: that more robust bottom quark could be just what a baryon needs to form a stable tetraquark.

The perpetual frustration of studying particles is that they don’t stay around long. These baryons, in particular, disappear faster than “blink-and-you’ll-miss-it” speed; one 10/trillionth of a second, to be exact. Of course, in the world of quantum physics, that’s actually plenty of time to establish existence, thanks to the LHC.

The great quantum qualm within the LHC, however, is one that presents a significant challenge in the search for tetraquarks: heavier particles are less likely to show up, and while this is all happening on an infinitesimal level, as far as the quantum scale is concerned, bottom quarks are behemoths.

The next question for Rosner and Karliner, then, was did it make more sense to try to build a tetraquark, rather than wait around for one to show up? You’d need to generate two bottom quarks close enough together that they’d hook up, then throw in a pair of lighter antiquarks — then do it again and again, successfully, enough times to satisfy the scientific method.

“Our paper uses the data from recently discovered double-charmed baryon to point, for the first time, that a stable tetraquark *must* exist,” Karliner told Futurism, adding that there’s “a very good chance” the LHCb at CERN would succeed in observing the phenomenon experimentally.

That, of course, is still a theoretical proposition, but should anyone undertake it, the LHC would keep on smashing in the meantime — and perhaps the combination would arise on its own. As Karliner reminded LiveScience, for years the assumption has been that tetraquarks are impossible. At the very least, they’re profoundly at odds with the Standard Model of Physics. But that assumption is certainly being challenged. “The tetraquark is a truly new form of strongly-interacting matter,” Karliner told Futurism,”in addition to ordinary baryons and mesons.”

If tetraquarks are not impossible, or even particularly improbable, thanks to the Karliner and Rosner’s calculations, at least now we have a better sense of what we’re looking for — and where it might pop up.

Where there’s smoke there’s fire, as they say, and while the mind-boggling realm of quantum mechanics may feel more like smoke and mirrors to us, theoretical physicists aren’t giving up just yet. Where there’s a 2-bottom quark, there could be tetraquarks.

The Perfect Pet

The Aibo is the perfect family dog. It’s attentive and engages eagerly with its owners, happy to follow wherever you go. It never makes a mess in the house. It sings and dances on request, and even greets you with a pleasant “good morning.”

That’s because the Aibo isn’t some exotic breed; it’s a type of robotic dog, manufactured by Sony.

However, its body of metal and plastic, rather than bones and fur, doesn’t change how Aibo owners connect with them. As illustrated in a New York Times mini-documentary, when Sony stopped manufacturing parts for the Aibo in 2014, owners were genuinely distressed that it meant the impending “death” of their pets — even going so far as to hold a funeral ceremony for them.

Why can’t we help but connect with robo-pets, even when we know they’re not alive?

“It’s a very interesting question, and the research on very young children suggests that it’s not a learned behavior,” said Gail Melson, a psychologist and Professor Emerita at Purdue University, who has studied human-robot interactions and blogs about our connection with wildlife for Psychology Today. Melson told Futurism that while we haven’t identified a brain mechanism for this anthropomorphism, we can speculate that there’s an evolutionary basis to the bond.

“We are inherently social creatures,” Melson explained. “Because of that we have evolved to be attuned to other life forms, and not only other human life forms. We are predisposed to see the characteristics of life.”

Melson’s research has examined how children, ranging from age 4 to 15, interact with the AIBO robot dog, finding most treat the robotic pet differently from a real dog. However, most do not behave as if it were an inanimate object or a toy. Younger children, in particular, often ascribed emotions and thoughts to AIBO. Intriguingly, children of all ages placed the robo-animals in the moral dimension, with most expressing that it would be wrong to harm the AIBO dog or throw it out.

“What’s happening in our age is the emergence of new categories, that haven’t existed before,” said Melson, noting that this is particularly the case for children who have lived with computer technology since birth. “We’ve divided the world, up to now, into things that are alive, or were once alive and now are dead, or never were alive. But now we have, thanks to this technology, these hybrid categories.”

Good Dog, Bad Dog, Malfunctioning Pets?

Just as there has long been discomfort and concern over humanoid robots, and the ethics of their existence, these robotic animals and their uncertain categorization too raise ethical and societal questions.

On the one hand, robotic pets have shown growing therapeutic values. Artificially furry friends like the the Joy for All Companion, Hasbro’s line of reactive robot dogs and cats, and Paro, a robotic seal made for therapy applications, have been used successfully for dementia patients, who often experience anxiety and distress. The service that these animals provide are similar to those given by an actual animal, cutting the isolation and sadness caused by their condition with companionship and affection—without the feeding and care demands of living pets.

“In general, people respond to a pet robot like they would to an animal, by patting and cuddling it and speaking to it like its an animal,” said Elizabeth Broadbent, an associate professor at the University of Auckland researching human-robot interactions in health contexts. She noted that, unlike proposed robotic caretakers that are modeled after humans, humans “don’t expect much of a response except some animal noises and movements,” making them simple and effective in their design and execution.

A 2016 study compared how 61 dementia patients fared when given a robotic pet(specifically, the Paro seal) three times a week for 20 minutes, as opposed to a control group who received the usual standard of care. The results were notable: the group that spent time with Paro showed a decreased pulse rate and higher blood oxygen levels (a sign of decreased stress), a lower rating on scales for depression and anxiety, and a decreased need for both pain and behavior medication.

One small study also showed that children with autism engaged more with an AIBO robot dog than with a simple mechanical toy dog, displaying the verbal engagement and authentic, reciprocal interaction that autistic children often lack.

For allergy sufferers or those without the time or money to care for pets, a robotic version might also be a better and more ethical option. Those trying to be eco-friendly might also be attracted to a robot’s smaller carbon paw-print.

Yet developmental psychologists in particular have raised concerns: that humans exposed primarily to robotic animals, and not to living ones, might lack in the social or emotional connections provided by living creatures.

“We [already] see concerns about children using other technologies like iPads and cell phones,” says Broadbent. “One of the fears is that children grow up more isolated and lonely because they do not form the same close friendships with other children through social media sites as they can form through face to face social contact.” The same concern applies, she says, to robotic companions. 

Melson added: “That question has given people pause […] are we going to diminish treatment of living animals, and people, because of the greater and greater presence of robots that seem to be good substitutes?” She cited the example of robotic pets in nursing homes, wondering if a decision to use only robots, and never real animals, might diminish the potential therapy benefit.

“We certainly don’t have robotics at the level to reproduce the smell, the feel, the response of even the crankiest living dog,” she said. “It would be a great diminishment of the experience to envision this, and yet people are short staffed, people are looking to save money, you can see how robots would be ‘good enough.'”

However, Melson is optimistic that our “biophilia,” humans’ hypothesized attraction to life and nature, will prevent us from replacing living animals altogether. While researching the AIBO, she brought one of the little dogs home to test out its presence in her own home. “I have to say I was struck by the limitations rather than the possibilities,” Melson said.

However, she added: “One would have to look at the increasing levels of sophistication and understand the different applications. We’re not jumping to say, let’s think of this as a substitute for living animals. I think that they have their own place.”

An AI That Can Build AI

In May 2017, researchers at Google Brain announced the creation of AutoML, an artificial intelligence (AI) that’s capable of generating its own AIs. More recently, they decided to present AutoML with its biggest challenge to date, and the AI that can build AI created a “child” that outperformed all of its human-made counterparts.

The Google researchers automated the design of machine learning models using an approach called reinforcement learning. AutoML acts as a controller neural network that develops a child AI network for a specific task. For this particular child AI, which the researchers called NASNet, the task was recognizing objects — people, cars, traffic lights, handbags, backpacks, etc. — in a video in real-time.

AutoML would evaluate NASNet’s performance and use that information to improve its child AI, repeating the process thousands of times. When tested on the ImageNet image classification and COCO object detection data sets, which the Google researchers call “two of the most respected large-scale academic data sets in computer vision,” NASNet outperformed all other computer vision systems.

According to the researchers, NASNet was 82.7 percent accurate at predicting images on ImageNet’s validation set. This is 1.2 percent better than any previously published results, and the system is also 4 percent more efficient, with a 43.1 percent mean Average Precision (mAP). Additionally, a less computationally demanding version of NASNet outperformed the best similarly sized models for mobile platforms by 3.1 percent.

A View of the Future

Machine learning is what gives many AI systems their ability to perform specific tasks. Although the concept behind it is fairly simple — an algorithm learns by being fed a ton of data — the process requires a huge amount of time and effort. By automating the process of creating accurate, efficient AI systems, an AI that can build AI takes on the brunt of that work. Ultimately, that means AutoML could open up the field of machine learning and AI to non-experts.

As for NASNet specifically, accurate, efficient computer vision algorithms are highly sought after due to the number of potential applications. They could be used to create sophisticated, AI-powered robots or to help visually impaired people regain sight, as one researcher suggested. They could also help designers improve self-driving vehicle technologies. The faster an autonomous vehicle can recognize objects in its path, the faster it can react to them, thereby increasing the safety of such vehicles.

The Google researchers acknowledge that NASNet could prove useful for a wide range of applications and have open-sourced the AI for inference on image classification and object detection. “We hope that the larger machine learning community will be able to build on these models to address multitudes of computer vision problems we have not yet imagined,” they wrote in their blog post.

Though the applications for NASNet and AutoML are plentiful, the creation of an AI that can build AI does raise some concerns. For instance, what’s to prevent the parent from passing down unwanted biases to its child? What if AutoML creates systems so fast that society can’t keep up? It’s not very difficult to see how NASNet could be employed in automated surveillance systems in the near future, perhaps sooner than regulations could be put in place to control such systems.

Thankfully, world leaders are working fast to ensure such systems don’t lead to any sort of dystopian future.

Amazon, Facebook, Apple, and several others are all members of the Partnership on AI to Benefit People and Society, an organization focused on the responsible development of AI. The Institute of Electrical and Electronics Engineers (IEE) has proposed ethical standards for AI, and DeepMind, a research company owned by Google’s parent company Alphabet, recently announced the creation of group focused on the moral and ethical implications of AI.

Various governments are also working on regulations to prevent the use of AI for dangerous purposes, such as autonomous weapons, and so long as humans maintain control of the overall direction of AI development, the benefits of having an AI that can build AI should far outweigh any potential pitfalls.