What does the fox say? In the future, we might have an answer that goes beyond a catchy tune. Scientists are using artificial intelligence (AI) to study how to talk to animals — and it's proving to be a challenging endeavor. Researchers have made progress analyzing and classifying animal sounds, but they can only go so far. They're only human. Computers, however, can sort through massive amounts of data to find patterns that aren't obvious."Communication is just a series of repeated sounds, gestures, signals, etc.," said Jay Thakore, a consulting engineer systems architect at Northrop Grumman. "We learn the sound the letters of the alphabet make to form words or that the wave of a hand means 'hello' using repetition. In that same manner, AI can learn the sounds and gestures of animals and then equate those repeated behaviors to words for humans."Interspecies Communication ChallengesIt's hard enough for people to learn a second language, such as a native English speaker studying Mandarin Chinese. There's also body language to consider, such as eye contact, hand gestures and the subtle meaning behind sighs. What's more, humans are only one of the millions of species of animals on Earth.Humans have taught animals to speak in ways we can understand, such as chimpanzees using American Sign Language or pets ringing a bell when they need to go to the bathroom. However, we haven't gotten very far in understanding animals' perspectives. When we're lucky enough to hear quacks, growls and clicks, humans don't often know what they mean. And some animals "speak" in sounds beyond our ears' frequency range.Animals also communicate in other ways, such as dogs sniffing each other, whales slapping their tales in the water and insects producing chemicals called pheromones. To an untrained ear, critter chatter seems nonsensical and random. But experts can translate languages, and AI is a helpful tool because it's particularly good at classification.AI for AnimalsRecent technological advances have opened new pathways for humans to talk with animals. No matter how dedicated and intelligent they were, previous generations of zoologists were limited by the tools invented before or during their lifetimes. Now, technology — such as miniature microphones and drones — make it easier for scientists to eavesdrop on nature. Notably, the type of AI behind streaming apps, voice-controlled assistants and self-driving cars can extract meaning from plentiful animal communication data."People have been exploring and studying animal behavior for years," says Thakore. "AI is in the forefront of everyone's mind because it's everywhere, and people are getting smarter about how to take advantage of it."For example, a nonprofit organization called the Earth Species Project (ESP) is building machine-learning models to understand crow caws, bat screeches and more. Their technology is based on natural language processing, a branch of AI focused on giving computers the ability to understand spoken and written language.But how do developers write algorithms for machine learning, such as an algorithm to classify animal vocabulary? As Thakore explains, AI algorithms are math- and logic-based and adjust themselves to perform better as they're exposed to more data. For the case of animal vocabulary, the algorithm could be a combination of a regression model that expresses a relationship between an input and the output and a decision tree that catalogs the outputs and can be used to determine which outputs happen most often for a particular input. From there, you can build the animal vocabulary.Now What?We humans have a lot to learn from our furry friends. As we get better at AI for animals, the next step is figuring out what to do with this new information and how to use it ethically. Here are a few potential industry use cases: Engineering: A plethora of innovations have already been inspired by animals — robots emulating the way animals climb, adhesives that copy geckos and astronauts landing with cat-like agility. Imagine what we could do if we knew how to 3D print like spiders or build homes as efficiently as ants dig tunnels. Conservation ecology: Some organizations aim to find out how people are harming the environment so we can stop the damage and repair the planet. Animal husbandry: If farmers could talk to animals, they could optimize their care and make informed decisions about raising livestock. Veterinary care and companionship: Who wouldn't want to know what their dog is thinking? Pets can't communicate how they're feeling or what symptoms they're experiencing, but with the help of AI, we could get a better idea of what they're going through. That could affect how we care for them and improve veterinarian diagnoses and treatments.It's still being determined whether AI will help us communicate with animals soon or if this is a futuristic dream, but Thakore remains hopeful: "As far as actually talking to animals, I think that is still a little in the science-fiction realm. But you never know — if someone were to devote enough resources, time and money, the future might not be that far away."Are you interested in science and innovation? We are, too. Learn more about life at Northrop Grumman, and check out our career opportunities to see how you can participate in this fascinating time of discovery.

Is It Possible to Talk to Animals With AI?

Artificial intelligence (AI) in medicine may evoke images of robots examining patients or even performing surgery — and that would be somewhat accurate. For example, AI technology has already been put to work performing some surgical tasks, according to the American College of Physicians. But AI is also impacting the medical world in other crucial ways, including research tools now being developed and used in brain research.How is AI helping humans understand the brain and, in particular, how the brain ages? According to the Icahn School of Medicine at Mount Sinai, an AI tool called HistoAge is able to determine "brain age" by identifying subtle cellular patterns in a region of the brain called the hippocampus. The tool gets its name from medical histology, which per the National Institutes of Health is the study of biological structures on a microscopic scale.Brain Age AccelerationThe hippocampus region of the brain is known to be associated with deteriorating brain function with age, resulting in conditions such as Alzheimer's disease. As reported by Neuroscience News, by determining the difference between brain age and the chronological age of a patient, brain age acceleration can be determined. This comparison, in turn, yields insights into both normal brain structure and function and also how the brain ages, including age-related decline as well as changes that may indicate the onset of brain disorders.Brain researchers believed that close comparative examination of structures in the hippocampus would yield key insights into how the brain ages, but there was a hitch. The Mount Sinai research team suspected that close examination of sectioned tissue sliced from the hippocampus could be used to determine brain age acceleration, but as noted by Neuroscience News, this was "a task that is impossible for a human observer to perform with any degree of accuracy."The problem is the structural patterns that reveal brain age are both subtle and numerous — meaning more correlations working together and interacting than anyone could hold in mind to make the comparison. This is not unlike the challenge of large-scale language translation, as reported by MIT News. Linguistics researchers sought for years to automate translation by sussing out the underlying grammatical logic that allows different human languages to express ideas. But grammar is just too complicated to fully work out by hand.Then, AI researchers discovered that if they just gave a computer an enormous amount of text to chew over (e.g., millions or billions of pages), they didn't need to fully work out the grammar in advance. The computer could patiently work through the mass of material, find correlations — such as how often one particular word follows some other word — and use all those subtle relationships to determine the best translation for a series of words. And it worked!Multiple Instance LearningThe brain researchers at Mount Sinai use the same approach but with the preserved brains of nearly 700 elderly brain donors instead of a corpus of text. However, this raised a further challenge. Fully detailing all these preserved brains would lead the researchers right back to the problem of too many subtle patterns to follow. The research team could only work with limited information, such as the chronological age and medical diagnoses of the donors at death.The solution to this problem was an AI technique known as multiple instance learning (MIL). This technique is suited to "weakly supervised" machine learning. As described by Jonathan Glaser, "Whereas traditional machine learning techniques rely on feature extraction by domain experts, deep learning algorithms learn high-level features from data on their own."In MIL, as reported at arXiv.org, data is initially sorted into "bags," or assorted collections of individual data instances. A bag may be positive if its data includes at least one instance of a specified characteristic or negative if it has no instances with that characteristic.Repeated passes through the data allowed HistoAge to build up its own detailed model of hippocampus characteristics based on its training data. The process was complex, but the goal was simple: The AI was to determine an estimated brain age at death for each brain studied. And, as with machine translation, it worked. Per Neuroscience News, HistoAge was able to estimate the average age at death within 5.45 years of the donors' actual ages at death.The Future of AI Brain Study?Mount Sinai researchers found that when stacked up against previous existing brain age determination techniques, such as DNA methylation, the age acceleration estimations derived from HistoAge reliably determined brain age and discovered insights into the driving factors of clinical and pathological outcomes related to cerebrovascular disease, cognitive impairment and abnormal levels of Alzheimer's-type degenerative protein aggregation. This offers a preliminary answer to the question, "How is AI helping humans understand the brain?"Note that this line of research into AI in neuroscience is not aimed at the goal of AI examining individual (living) patients but at building up a research model that human brain specialists can then use as a baseline tool for evaluating an individual patient's brain health. This exemplifies the wide range of applications in which AI technology is enabling transformative progress.From a big-picture perspective, HistoAge is a demonstration of AI capabilities: identifying complex and subtle patterns in a large body of data — patterns that humans without AI tools would never be able to detect, let alone work with. And this innovative use of AI could pave the way forward for brain study in the years to come.Interested in solving the toughest problems? Us, too. Learn more about life at Northrop Grumman, or check out our career opportunities to see how you can be a part of defining possible.

How Is AI Helping Humans Understand the Brain?

Unmanned aircraft systems, or drones, have become a ubiquitous part of our lives. Smart, microprocessor- and software-driven flying machines now regularly perform military missions tagged as "dull, dirty or dangerous," while local government agencies and private entrepreneurs enlist drones for missions such as aerial surveillance of fires, wedding photography and small-package deliveries. Against this backdrop, however, human test pilots — those daring young men and women in their flying machines of movie legend — remain a central and indispensable part of aircraft design, and the teams and processes that develop, test and field new aircraft systems, both manned and unmanned. "Developmental test pilots serve as the 'bridge' between engineers (who design the plane) and operational pilots (who fly the plane)," explains Bob Hood, a retired U.S. Air Force test pilot and one of 15 active test pilots employed today by Northrop Grumman. "The test pilot must be able to speak 'engineering' to help aircraft developers produce a plane with good flying qualities that can meet its mission requirements. He must also help developers produce a flying experience for pilots that is engaging, intuitive and easy to understand." Start With STEM Many test pilots begin their careers as military pilots with a degree in engineering or another technical degree such as physics or mathematics. "To enter test pilot school, you also have to have substantial flying experience — a minimum of 750 flight hours — in military aircraft," explains Troy Johnson, a retired Air Force test pilot and Northrop Grumman Aerospace Systems' chief test pilot. The U.S. Navy and the Air Force both operate their own test pilot schools; commercial test pilot schools are also available in the U.S. and abroad. Join the Dream Team Ironically, explains Don Weiss, a retired U.S. Marine Corps test pilot and Northrop Grumman's chief B-2 stealth bomber test pilot, "a majority of the contributions of a test pilot are not made while seated in an aircraft." The role of a test pilot actually begins, he explains, "in the 'dreaming' stage of aircraft design as part of a team considering all aspects of aircraft operation and performance." The test pilot contributes to everything from aircraft design criteria; vendor selection for specific subsystems; and the layout of cockpit controls and displays; to helping develop simulators, flight test plans and training programs — and the earlier in the program, the better. "The single most important job of a test pilot," advises Wayne Staley, a retired Northrop Grumman test pilot who began his test pilot career in the Air Force, "is to collect the data that the aircraft's designers need to evaluate the operation and performance of the aircraft." Bring the Right Stuff The most important characteristic of a test pilot? "Integrity," says Weiss. "Both engineers and pilots have to trust, intimately, the real-time findings of the test pilot. If there is an issue with the plane, the test pilot has to be very transparent so that the situation is well understood by designers and operators alike." Good test pilots need to be disciplined, curious, and cool under fire. According to Hood, however, "fear is not part of a test pilot's vocabulary." If something unusual happens during a flight, he says, "You have to operate in the present. It doesn't matter how you got there, only how you're going to fix it." And this advice has paid off. In October, both Hood and Staley were recognized by the Society of Experimental Test Pilots (SETP) for their career-long ability to meet challenging flight test requirements. Both men were honored as members of SETP's 2017 Class of Fellows at the organization's annual meeting, in Anaheim, Calif. Evolve With Technology As aircraft design technology has evolved, so too has the role of the test pilot. "In the 1950s, pilots took planes out to see what they would do. Today we take aircraft out to verify the predictions of our models," says Staley. In an age of sophisticated wind tunnel testing, modeling and simulation tools, he adds, "we don't get into the cockpit until our predicted performance is good." But a new aircraft system doesn't need a cockpit to benefit from the skills of a test pilot. "Increasingly, high speed computers do most of the 'thinking' and management of flight control surfaces for autonomous systems," says Weiss. "But you still need someone who can evaluate the air vehicle's flying qualities, because humans are involved in operating it from the ground." FAA rules require operators of certain classes of unmanned systems to be flight qualified as pilots, and instrument rated. Test pilots work closely with designers of ground control stations, therefore, to ensure that aircraft controls, displays, and methods of presenting information are consistent with how our pilots are trained. Test pilots also help autonomous system designers ensure that an aircraft's avionics and flight control systems can respond appropriately to human air traffic control commands to the air vehicle operator. "The goal is to design autonomous systems so that they are fail-safe and foolproof, but every once in a while, these systems fail and you end up losing an aircraft," observes Johnson. "By including test pilots' expertise in the development of an aircraft, there's a much better chance that a pilot can bring it home safely." Manage the Risks In the end, being a test pilot centers on risk management, whether you're testing a manned or an autonomous system. "The test pilot and the test team spend a lot of time discussing and mitigating risks before a plane ever takes off," explains Staley. "These discussions involve the people who understand best how the aircraft works. Our shared goal is to identify possible problems and minimize our exposure to those problems. That way we'll always have a safe escape route if unexpected situations arise." Enjoy the Ride For all the discipline, teamwork, and mission planning, however, test pilots still go to work to do what they love most — fly high performance aircraft. "One of my most memorable experience was testing the weapons delivery capabilities of a military jet over the ocean," recalls Weiss. "On test day, there was a marine layer of clouds. We needed to have clear conditions to complete the test properly. We spotted a hole in the clouds that was just large enough to drop in, release the weapons, document it and get out. That was fun, and it felt good to accomplish the test under challenging conditions. Most importantly, we got the data. I call that a pretty good day at the office."

Aircraft Design in the Age of Autonomous Systems: Test Pilots Still Play Essential Role

Key Takeaways:  Dogs take a "total body" package approach rather than focusing on the face for social recognition. This social behavior preference is associated with anatomical differences in rod and cone density in the retina. Facial-focused species include cats and sheep, as well as humans and primates.What do dogs see when they look at humans? With their wide eyes and steady gazes, it's easy to anthropomorphize our furry friends and imagine they're seeing the same things we are — the precious and particular facial features of a trusted companion. However, according to recent research, that's not entirely accurate. While humans typically take a face-first approach to social interactions, dogs prefer the total body package.The Kin ConsiderationFor dogs, seeing and recognizing familiar faces is a subset of the larger kin and mate selection process. In effect, dogs become familiar with their "pack" — human or otherwise — by creating a mental map of key physical characteristics that include "non-facial bodily cues, acoustic or chemical signs," according to Attila Andics, co-author of the recent study from the Society of Neuroscience.While research from Andics and his colleagues found that dogs are perfectly capable of recognizing human faces, this information isn't their top priority. Using fMRI scans, the research team discovered no difference in dogs' mental activity when looking at human faces or the back of human heads. This suggests that dogs lack the specific face recognition functions found in the brains of humans and other primates. While dogs did show species preferences through greater brain activity when looking at pictures of other dogs compared to humans, they're simply not built with the same facial framework.This isn't to say that your dog doesn't like your face or doesn't enjoy gazing into your eyes. Andics notes, "It is amazing that, despite apparently not having a specialized neural machinery to process faces, dogs nevertheless excel at eye contact, following gaze, reading emotions from our face, and they can even recognize their owner by the face." In fact, research suggests that dogs can differentiate between happy and angry human faces and are capable of recognizing the human smile — often a sign of aggression in other species — as a positive social cue in human/canine interactions. In other words, dogs have evolved their ability to recognize and react to human faces and emotions as part of people-focused packs.The Vision VariancePart of the disconnect between dog and human facial recognition stems from differing volumes of light receptors known as "cones" and "rods". Compared with canine eyes, human eyes have a higher abundance of cone receptors, which are used to distinguish colors. This allows us to detect three distinct colors — red, green and blue — along with variations and combinations of these colors. Lower cone volumes in dogs' eyes mean they can only see two colors, although the jury is out on exactly which ones are part of the canine color experience.Meanwhile, when it comes to the volume of rods, canine vision vastly outpaces that of humans, allowing our four-legged friends to see much better in the dark. Dogs also have an extra layer of eye tissue known as the "tapetum lucidum," which further enhances light reflection into the retina. The result is a visual system that's designed to see shapes in any light condition but isn't well suited to recognizing the specifics and subtleties of human facial features.The Interspecies IntersectionIf dogs can recognize human faces but don't prioritize this physical characteristic, where does that leave us? Are we the only species with dedicated facial processing functions? Not quite. As NPR notes, other primates also have groups of specialized neurons designed to facilitate facial recognition. As our closest evolutionary cousins, this makes sense since many primate species demonstrate human-like interactions within social groups.And primates aren't the only ones with a face focus — research suggests that cats have specific facial expressions that can be decoded by humans with enough observation. The existence of these expressions also suggests that other cats could potentially recognize them and, in turn, observe and interpret human expressions. However, similar to how cats recognize their human-given names but don't respond unless they feel like it, they simply may not care to do so with humans.Also making the list of facial-focused species? Sheep. As pack animals with relatively large brains, it makes sense that sheep learn to quickly recognize common companions. But studies also found a predisposition for facial recognition — even having only seen their human handlers in person, sheep were quickly able to recognize their photograph. In addition, they were able to recognize facial images both straight-on and from an angle, and with similar precision as humans.The Eyes Have ItWhat do dogs see when they look at us? Trusted kin and familiar pack mates. And although canine vision may use a broader data set to ensure species safety, dogs' natural tendency toward empathy and engagement means that, while your face isn't the biggest priority, they're always happy to take a look.Find out how AI can help you communicate with animals.

What Do Dogs See? It's Rough, But Faces Aren't First Priority

 Key Takeaways
 
 
 

    Biomimicry is the study of systems in nature to find solutions in human design.
    Aviation design can benefit from biomimicry — studying birds can help us make aircraft flight and design more efficient.
    AI can help tackle and refine the huge data sets generated by biomimicry research, building algorithms to mimic bird flight or analyzing wing shape, for example.

 
 
 
Who hasn't stared up at a hawk or vulture circling lazily in the sky and wondered how they stay aloft so long? Researchers are now able to tackle this question more systematically using biomimicry, the process of mirroring nature's systems to solve complex human problems.
 
Their goal is to develop artificial intelligence (AI) algorithms that will allow gliders to mimic the flight behavior and endurance of birds and understand wing structural changes that facilitate silent flight, among other projects. Soon, AI in aviation will transform the types of missions that gliders perform and substantially reduce noise in flight.
 
 
 
Flying Like Birds
 
 
 
AI and biomimicry are already helping aviation researchers understand how gliders can function more efficiently. Gautam Reddy, a graduate student in physics, and a team of his colleagues at the University of California, San Diego, described their biomimicry research in a 2018 article in Nature.
 
 
 
The team equipped a two-meter-wingspan, radio-controlled glider with a flight controller and instruments that could measure vertical wind accelerations along the length of the glider and across its wings. Their working theory was that birds use similar sensory information to identify and navigate thermals.
 
 
 
Between 2015 and 2017, they conducted more than 500 flights near Poway, California, each one lasting 20 to 30 minutes. The aim was to train the glider in a concept called "reinforcement" to navigate thermals autonomously. They used their measurements to build a simple algorithm that would steer the glider toward zones of higher lift.
 
 
 
Another study published in the Flow, Turbulence and Combustion journal, also looked at soaring flight, studying behavior in white storks for clues to how the birds stay aloft so effortlessly. The researchers took data observed during stork flight and created a biomimetic computational model that analyzed the way the wing tip changed shape during flight. The study also looked closely at the airflow features around the wing tip feathers during flight. Similar work in pigeons could also aid wing design. For example, Popular Mechanics points to PigeonBot: a highly sophisticated bird-like robot that researchers are using to study how pigeons change the shape of their wings during flight to gain efficiency.
 
 
 
Taming Unpredictability
 
 
 
Though much valuable information can be gleaned from the concept of biomimicry, field studies can be problematic. Atmospheric conditions vary so much hour to hour and day to day, even in the same location, that building a robust AI algorithm for glider aviation using field measurements alone is quite challenging.
 
 
 
"AI algorithms require a lot of data," Reddy said. "One way to add fidelity to the algorithm would be to use numerical simulation and a process called model-based learning."
 
 
 
This model-based approach, which his team did not use, would involve "flying" a simulated glider repeatedly through a simulated atmospheric environment containing realistic thermals, convective currents and turbulence. Atmospheric data and the glider's response to those conditions could be collected on each simulated flight and then used to systematically update and refine the AI algorithm.
 
 
 
A better strategy, Reddy proposes, would be to extract visual and other sensor information from gliders.
 
 
 
"State-of-the-art gliders actually have better aerodynamic properties than birds," he said. "We can also use instruments to gather higher accuracy measurements of temperature, wind velocity, humidity, barometric pressure and other relevant data."
 
 
 
New Rules of Engagement
 
 
 
AI in aviation certainly offers tantalizing prospects.
 
 
 
"If we can come up with a glider that's as efficient as a bird, it could travel across continents," said Reddy. "That capability could be extremely useful for surveillance or remote sensing missions." The one limitation, he admitted, is that gliders rely on thermals, which are normally present during daylight hours only.
 
 
 
AI-controlled gliders could be a perfect complement to high-altitude surveillance systems, such as the U.S. Air Force's Global Hawk unmanned aircraft.
 
 
 
"Gliders are very small, unpowered and virtually silent," said Reddy. "So, they would not need to fly at high altitude. I think they could provide our military commanders with some very interesting options."
 
 
 
AI in the Cockpit and Beyond
 
 
 
Reddy suggests we try out AI as a navigational aid for pilots to enhance their abilities in the air.
 
 
 
"Pilots would still fly as before," he said, "but would use the AI algorithm to suggest the fastest, most efficient route to their destinations, taking advantage of the fastest-flowing streets (thermals) along the route."
 
 
 
AI and biomimicry could also aid other areas of in-flight performance, creating aircraft that fly silently, similar to an owl swooping on its prey. According to Popular Mechanics, owls can inspire quieter aircraft and turbine blades. Analysis of the wing's trailing edge has found that the orientation of the feathers changes the way the air flows as the owl flies.
 
 
 
In a similar way, hummingbird flight dynamics could help drone flight. For example, Penn State University describes novel modeling that's allowing researchers to simulate hummingbird flight, and Science Daily notes that the data from these studies could refine flying robots.
 
 
 
And it's not just birds. The Biomimicry Institute describes AI coupled with slime mold growth can revolutionize materials design. Along with studying bone growth, researchers are developing new approaches that could make aircraft lighter and use less fuel. Combining generative design, AI and 3D printing can even help us create lightweight components, such as cabin partitions that are strong but use less raw material.
 
These are just a few examples of how looking to AI and ecosystems can transform aviation systems. It's clear the future holds some exciting advancements in technology still to come, and nature may play a pivotal role in realizing those innovations.
 
 
 
Northrop Grumman is a leader in the development of autonomous aerial reconnaissance systems. If you'd like to learn more about what we do, check out our work in AI and machine learning.
 
 
 

Interested in a career with Northrop Grumman? Check out our open positions.