AI Learns To Predict Human Behavior From Videos

 

As humans we are able to pick up on the many versions of body language and process whether it is good or bad, and quickly come up with a reaction to it. Some of the body language I'm speaking of a proper hello, handshake or even a wave. We may not know right away what another human will use but because of our quick processing and awareness of the surrounding area, we are able to read the situation and respond appropriately. Recently, researchers at Columbia Engineering have brought to light a computer vision technique that allows machines a intuitive sense for projecting next moves by capitalizing higher-level correlations between people, animals, and objects. According to Carl Vondrick, Assistant Professor of Computer Science at Columbia, "Our Algorithm is a step toward machines being able to make predictions about human behavior, and thus better coordinate their actions with ours". He further stated, "Our results open a number of possibilities for human-robot collaboration, autonomous vehicles, and assistive technology". The researchers claimed this to be the most accurate method that will predict live video events up to seven minutes in the future. In their study, they analyzed thousands of videos and one in particular like The Office. The system processed and predicted various activities that ranged from handshaking to fist bumping. In the past, predictive machine learning ran by other teams have concentrated on prediction of single actions at a time. This would then categorize the action as a hug, high five, handshake, or a non-action. Columbia Engineering with this knowledge chose to approach this by looking at the longer-range prediction problem from a different viewpoint. At the end, the team concluded that the new algorithm would refine accuracy of future predictions on significant tasks than in previous times. However, the next step to ensure this success is to verify that it will work in real world application and not just the lab. If the system can handle various settings, then we can be assured many opportunities await to deploy machines and robots that will greatly improve our safety, health, and security. 

Swarm of Autonomous Tiny Drones Can Localize Gas Leaks

         

        Accidents happen all the time. It is sometimes preventable and sometimes not. On average, there are 4,200 home structure fires per year started with the ignition of natural gas (nfpa.org). When a leak is created, human firefighters are called upon to figure it out. These firefighters will use gas sensing instruments to enter either the residence or industrial site to locate the leak which often takes a significant amount of time. Along with that, they are increasing the risk to their lives. Fortunately, researchers at TU Delft, University of Barcelona, and Harvard University have created the first prototype of a very small swarm of very safe drones that can automatically identify and centralize gas sources in congested environments indoors. During their research, the biggest trouble they discovered was designing artificial intelligence (AI) that could be placed very carefully in the drones' tight computational and memory parameters. To solve this challenge the researchers introduced bio-spired navigation and search strategies.

        Autonomous identification of a gas leak is complex. The reason is that artificial gas sensors are at this time, less than capable compared to animal noses to detect small amounts of gas while staying alert to rapid changes in gas concentration. Furthermore, gas will spread within an environment making it more complex. The information from the researchers focused the majority on solo robots that quickly identify a gas source in a small, obstacle-free environment that allows sourcing to be easier. According to Guido de Croon, Full Professor at the Micro Air Vehicle Laboratory of TU Delft, "We are convinced that swarms of tony drones are a promising avenue of autonomous gas source localization. The drones' tiny size makes them very safe to any humans and property still in the building, while their flying capability will allow them to eventually search for the source in three dimensions."

Source: https://www.sciencedaily.com/releases/2021/07/210714110540.htm

Computer-Assisted Biology: Decoding Noisy Data to Predict Cell Growth

             In natural science, we have seen biology utilize mathematical models to process data and predict future outcomes. However, this is more difficult when dealing with the dynamics of living systems. Most of the data rely upon simplifying assumptions that do not result in the organic biology process that is true. Researchers at The University of Tokyo have created a way to deal with the complexity of living systems. They developed a machine-learning algorithm that uses the size of one cell over the course of time to make future predictions about their size. The computer can quickly identify patterns in the data set, unlike the boundaries of conventional methods. Recently, data was collected from either an Escherichia coli bacterium or a Schizosaccharomyces pombe yeast cell that was positioned in various temperatures in a microfluidic channel. What the researchers noticed was over time the size appeared to look like a "sawtooth" as growth was halted by division events. Normally, human biologists utilize a "sizer" model based on the exact size of the cell, the increase in size since birth, to anticipate when cellular division will happen. Fortunately, the algorithm discovered supporting efforts for the absolute size principle. Senior author, Tetsuya Kobayashi stated, "Our deep-learning neural network can effectively separate the history-dependent deterministic factors from the noise in given data." The team noted that this method could be utilized in other areas of biology and not just predicting cell size. The takeaway from this is that possibly soon life science can be more influenced by artificial intelligence than human models. In doing so, this will produce a greater control over microorganisms that are used to ferment products and produce drugs.

Source: https://www.sciencedaily.com/releases/2021/07/210709104218.htm

Insect-Sized Robot Navigates Mazes With The Agility Of A Cheetah

 

        Growing up we quickly learn that a majority of insects (and spiders) possess the ability to move about rapidly up walls and remain on ceilings by walking upside down. This is contributed by their footpads being sticky with a specialized chemical so that no surface is limiting their movements across them. Recently, the University of California formed a team of engineers to understand and apply the same values behind a spider's ability to stick. They called it electrostatic adhesion. Their goal was to create an insect-scale robot that has the ability to maneuver with agility like you find in a cheetah while also being able to maneuver difficult terrain with little to no effort and avoid obstacles quickly. The engineers created the structure of the robot from thin, layered material that allows various bending and contracts when an electric voltage is used. In 2019, the same team showed the same demonstration when a cockroach-sized robot was created. It possessed the ability to hurry across a floor at a sustained rate of 20 body lengths per second. That translates out to be 1.5 miles per hour. What was interesting about this is the team was able to match up the cockroach-sized robot to the same speed as an organic cockroach. By applying two additional electrostatic footpads to the robot, adding an increase of voltage to either footpad increases electrostatic between footpads and surface. The end result produces a stickier footpad that bonds to the surface more so and causes the other portions of the robot to pivot around the foot. The trajectory of the robot gives operators full command because of the two footpads, allowing sudden turns using centripetal acceleration that can exceed other insects.

Source: https://www.sciencedaily.com/releases/2021/07/210702154253.htm

New Algorithm Helps Autonomous Vehicles Find Themselves, Summer or Winter

         

            Vehicles are being equipped with plenty of new technologies as the automotive industry adapts to the rapid technological advances. A big part of technology that is used daily is GPS. GPS Is important for autonomous vehicles and without it they would lose themselves easily and possibly not even function to move.  However, good news has recently been discovered. A team at Caltech has created a functioning algorithm that allows autonomous vehicles to identify their location by observing the terrain surrounding the vehicles. To make things even more interesting, they further discovered that the technology is effective regardless of seasonal changes. In the 1960s a general process called visual terrain-relative navigation (VTRN) was created. This process took satellite images that have been collected and compared them with nearby terrain which helped the autonomous find themselves. However, with this process, there was one major downfall. The current VTRN requires both the terrain and images provided by the databases to be closely matched up. For example, fallen leaves would cause image distortion and throw errors to the system. Fortunately, the team at Caltech created a solution to overcome this problem. They utilized deep learning and artificial intelligence (AI) to filter seasonal content that would throw errors to the current VRTN systems. This process is known as “self-supervised learning”. The majority of computer-vision strategies require human interactions who vigilantly create large data sets that teach an algorithm to understand what it is seeing, but this one allows the algorithm to educate itself. The AI scans for patterns by focusing on smaller details and features that human interaction would more than likely miss. This doesn't stop just at everyday vehicles; it goes beyond Earth boundaries into space missions. The entry, descent, and landing (EDL) system uses VTRN for the first time on Mars to assist in landing at the Jezero Crater. This area was known to be too dangerous for safe entry. The team utilized Martian polar regions on Mars because they have similar intense seasonal changes like Earth as there's no GPS on Mars. The new system would improve navigation to assist in scientific objectives like water.  

Source: https://www.sciencedaily.com/releases/2021/06/210623141655.htm

Humans Are Ready to Take Advantage of Benevolent AI

This article goes over a study that the University of London and other sources conducted on the relationship between human and artificial intelligence (AI). Picture yourself in the near future coming up to an intersection and to your right another car pulls up. That other car is unique because it is one of the self-driving cars with no passengers inside. How would you handle this? Would assert dominance and be the first one to proceed through the intersection, or be considerate and give way to allow the self-driving car to proceed? These were the questions for the international team of researchers from UL. They used behavioral game theory, a method to conduct a large-scale online study to understand and possibly conclude whether humans would cooperate with AI systems just as much as they do with humans. As humans, we know that for a society to thrive and do well, cooperation is crucial. It requires understanding with others to find a middle ground and accept the associated risk that other humans can potentially let us down. During the study, an observation that occurred earlier found that the level of trust humans has for other humans is the same level of trust humans have toward AI. There were some who expected to be an individual ready to work together. Further in the study, the researchers notice how much humans are not willing to accommodate machines as it presents a new challenge for the future interaction of humans and AI. According to Dr. Jurgis Karpus, a behavioral game theorist and a philosopher for the study, said "People expected artificial intelligence to be as a cooperate as fellow humans. However, they did not return their benevolence as much and exploited the AI more than humans." The researchers discovered that "algorithm exploitation" is a booming reality. This was evident when they duplicated substantial findings from nine experiments with 2,000 human volunteers. With the idea of AI being more benevolent to humans, this will naturally show the result of less cooperation between the two. Compromising is crucial to the success of society.

Source:https://www.sciencedaily.com/releases/2021/06/210610135534.htm

Hacking and Loss of Driving Skills Are Concerns For Self-Driving Cars

 

As the years go on, technology and technological advances become more intertwined into our daily lives and movements. One fascinating futuristic concept we see in the movies is now becoming more of a reality in the year 2021. Autonomous vehicles. These types of vehicles are being developed rapidly worldwide which prompted the researchers at the University of Kent to investigate the question of "WHY?" Why might the market for these vehicles continue to grow and why aren't consumers deterred from purchasing? Based on the researchers' findings, both qualitative and quantitative, they drew the conclusion on consumer decision making about autonomous vehicles.

        The cons to these vehicles included performance risks to the AI and sensor systems, loss of competencies by driving public, and security breach of privacy. With these cons, it has and will continually be a challenge especially road and passenger safety. Automotive companies must continue to be innovative in their marketing of the products. Marketers have been doing their job by advertising improvements to the product's technology with the hope that it will reduce safety concerns. However, the hurdle of overcoming the loss of driving ability and privacy breach has proved to be an uphill battle.

        In a positive light, the researchers saw from their observations not only cons but benefits autonomous vehicles present. These benefits were more time available (giving more time to the consumer which normally would have been used for operating the vehicle to their destination), the removal of human error (fewer accidents caused by human drivers), and outperforming the human factor meaning the AI would improve routes, traffic predictions, and handling speed.

        According to Ben Lowe, Professor of Marketing at Univerisity of Kent and co-author of the study said, "The results of this study illustrate the perceived benefits of autonomous vehicles for consumers and how marketers can appeal to consumers in the growing market. However, we will now see how the manufacturers respond to the concerns of these key perceived risks as they are most factors in the decision-making of consumers, with the safety of the vehicles' performance the greatest priority. Our method used in this study will help clarify for manufacturers and marketers that, second to the issue of online account security, they will now have to address concerns that their product is reducing the autonomy of the consumer." 

 

Source: https://www.sciencedaily.com/releases/2021/05/210526093059.htm

Researchers Develop AI That Can Detect Sarcasm in Social Media

        

        Social Media has exploded in growth very quickly in recent years and being labeled as a powerful form for individuals to communicate, and organizations to reach further to consumers to market and sell their products or services. With this immense communication tool, it is important for each of us to understand how to properly respond to the feedback from customers on social media platforms for there to be a success, however, this involves greatly the demand for labor. For this success to be achieved, a natural language processing technique called Sentimental Analysis takes center stage. What this means is that a piece of text is computationally identified and classified to determine a writer's overall attitude toward a product, service, etc. is either positive, negative, or neutral in some cases. We know that artificial intelligence correlates to logical data analysis and their response, while sentiment analysis solely focuses on emotional communication identifying. At the University of Central Florida (UCF) a team of experts created a procedure that identifies all forms of sarcasm in a social media text. Initially, the team taught the computer model to observe patterns that involved the usage of sarcasm while also instructing the program to identify clue words that would be most likely be in relation to sarcasm. According to Assistant Professor of Engineering, Ivan Garibay, he stated, “The presence of sarcasm in the text is the main hindrance in the performance of sentiment analysis.” Garibay further states, “Sarcasm is always easy to identify in conversation, so you can imagine it’s pretty challenging for a computer program to do it and do it well.” The UCF team created an intelligible deep learning model that uses various head self-attention and gated recurrent units. What this means is that the various head self-attention module helps in pinpointing essential clue words that relate to sarcasm from input, and the recurrent units master far-reaching responsibilities between these words to better analyze the text.  

Source: https://www.sciencedaily.com/releases/2021/05/210507112040.html 

Is AI the Future of the Insurance Industry?

 

        Hello, in this week’s post we will cover the future of the insurance industry and how artificial intelligence (AI) ties into that. All of whom own a vehicle and house must purchase an insurance policy. It is essential to survive when accidents or trauma occur. It is meant to provide safety assurance for personal or professional needs. However, what we do not realize is that these companies who we rely on are determining their policies, and products on outdated, incorrect data. Long-term effects could result in enormous amounts of money spent and damage to the consumer. This article details that the insurance companies which collect data points are still relying on demographic information over 40 years old. It is 2021 and as a consumer, this needs to be rectified to more relevant to modern time. With integration, of further AI, it can process over 4,000 data points within minutes, and want to know something even better? It can also process mortality, demographic, health, and government trends dating back 20 years! This results in excellent algorithm creation which means better more effective decision making. Tools that improve this process are Data Science-as-a-Service (DSaaS) and Insurance-as-a-Platform (IaaP). The benefits of these platforms include reducing risk more accurately, reducing cost, creating products that are more tailored to the clients, and improved customer relations. Going further, this can also predict future understandings of consumers who are ready to buy and establishing the correct risk on their policies. Result—more profit. These data streams policies become more adjusted to the consumer’s needs whether they are a new or existing client. In the past, new products were created based on existing ones which meant they did not fully analyze what was more effective and where true adjustments are needed. Now, with AI and DSaaS, these platforms are allowing carriers to produce better products based on demographic and risk data. With the help of the human factor (industry leaders), they can assist by deploying agents to monitor and design products to truly meet the policyholder’s needs. Although these technological advances can be an enormous jump for a company, it is still vital to have human agents. Agents are crucial in developing trust with the clients and being the gateway between big data and personal connections. Consumers still want to know there are individuals with a pulse that care and are knowledgeable when they need help.

Source:https://insidebigdata.com/2021/05/25/is-ai-the-future-of-the-insurance-industry/

Topic Submission: Artifical Intelligence with an Emphasis on Data Mining and Machine Learning

     

        Hello, during the Summer 2021 semester, I am enrolled in LIS 4022 Writing for Information Professions. I will cover Artificial Intelligence while emphasizing Data Mining and Machine Learning. Both are used by AI, however, in different manners for different outcomes. This post is designed to have you ponder about AI with both Data Mining usage and Machine Learning usage. Now, I know that both these fall under the same science, but it will be important to acknowledge their differences and see their impact on AI in the present time and toward the future. Some key points of discussion will include programmed control systems for Data Mining and algorithms that Machine Learning uses to complete desired tasks. Artificial Intelligence has continued to spark my interests since I was young. Being a part of the IT major, I am curious to see correlations between Data Mining and Machine Learning and what the future could possibly hold for these two categories on AI.