The cybersecurity company F-Secure has recently created a new AI project that utilizes techniques inspired by “swarm-intelligence”. As AI News reports, F-Secure’s new AI approach makes use of many decentralized AI agents that all collaborate in order to carry out accomplish specific goals.
F-Secure’s new swarm AI is similar in concept to Fetch AI’s earlier take on decentralized AI systems, which have been applied to IoT concepts. However, unlike Fetch AI, F-Secure is aiming to take the concept of decentralized AI and use it in the cybersecurity domain. Specifically, F-Secure is aiming to improve the company’s detection and response capabilities.
As explained by Matti Aksela, the VP of AI at F-Secure, it is commonly believed that AI should aim to copy human intelligence. However, while patterning AI systems after human reasoning and behavior isn’t inherently bad, Aksela explained to AI-News that only patterning AI after human cognition is limiting what we can do with AI. Aksela explained that we can look outside of human cognition and explore other methods of organizing and architecturing AI. A wider range of possible models for AI can augment what people can already accomplish with AI.
Swarm intelligence is a behavior of decentralized systems. It’s a collective behavior that manifests itself in both artificial and natural systems. In terms of biological systems, swarm intelligence is often seen in large colonies of organisms like ants, bees, fish, and birds. For instance, many birds migrate in large flocks and as the flock travels it maintains a consistent formation that fluctuates very little, with the birds only deviating a few inches from one another in their formation. It is thought that flying in such formations reduces the energy that the birds require to fly.
Swarm intelligence has been used for probabilistic routing in telecommunication networks and in the creation of microbots. One example of this concept is the tiny robots created by MicroFactory. The robots are controlled by a circuit board that generates a magnetic field, and the robots themselves are magnets. The robots are also equipped with small manipulation tools that they can use to interact with the environment around them and manipulate objects.
The development of genuinely human-like artificial intelligence, or Artificial General Intelligence, will take some time to be created. Estimates by various AI experts vary, but on average it is thought that it will take around 50 years to succeed in the creation of an AGI. In contrast, the development of distributed autonomous agents like the ones F-Secure should take a significantly shorter time.
According to F-Secure, several years more years of development will be needed to for their distributed intelligence architecture to reach its full potential, but some mechanisms based on the swarm-intelligence model are already in use. F-Scale has used some swarm-intelligence techniques to detect breaches and engineer solutions.
F-Secure’s AI agents are capable of communicating with each other and collaborating.
Swarm intelligence techniques make use of the talents or capabilities of individual agents in the agent pool, and when these skills are networked together there is a robust and flexible system capable of carrying out complex tasks.
“Essentially, you’ll have a colony of fast local AIs adapting to their own environment while working together, instead of one big AI making decisions for everyone,” Aksela explained.
In the specific case of F-Secure the different agents are capable of learning from different networks and hosts, and the agents can thread spread this knowledge through the wider network which joins together different organizations. F-Secure says one of the main benefits of this approach is that it can enable the organization to share sensitive information via the cloud and still remain protected due to superior break and attack detection.
Robots Walk Faster With Newly Developed Flexible Feet
Roboticists at the University of California San Diego have developed flexible feet for robots. The new technology can result in robots walking 40 percent faster on uneven terrains like pebbles and wood chips.
The new development is important for a variety of different applications, especially search-and-rescue missions.
The research will be presented at the RoboSoft conference, which will be virtual and take place between May 15 and July 15, 2020.
Emily Lathrop is a Ph.D. student at the Jacobs School of Engineering at UC San Diego and the first author of the paper.
“Robots need to be able to walk fast and efficiently on natural, uneven terrain so they can go everywhere humans can go, but maybe shouldn’t,” Lathrop said.
Michael T. Tolley is a professor in the Department of Mechanical and Aerospace Engineering at UC San Diego. He is the senior author of the paper.
“Usually, robots are only able to control motion at specific joints,” said Tolley. “In this work, we showed that a robot that can control the stiffness, and hence the shape, of its feet outperforms traditional designs and is able to adapt to a wide variety of terrains.”
Flexible Robotic Feet
The flexible robotic feet consist of a latex membrane that is filled with coffee grounds. The coffee grounds are able to go back and forth between acting as a solid and a liquid. The mechanism that allows granular media, such as the coffee grounds, to act this way is called granular jamming. As a result, the robots can walk faster and have a better grip.
As the robot feet touch the ground, they turn firm and conform to the surface in order to establish solid footing. When they move, the feet unjam and loosen up between steps, and support structures are relied on to help them stay flexible while jammed.
These flexible feet were the first of their kind to be tested on uneven surfaces.
The researchers installed the feet on a hexapod robot, and they designed and built an on-board system. The on-board system is capable of generating negative pressure and positive pressure in order to unjam and jam the feet between each step. In order to jam the feet, a vacuum pump removes air between the coffee grounds. They can also be passively jammed if the weight of the robot forces the air out from between the coffee grounds.
The robot was tested walking on a variety of different surfaces, including flat ground, wood chips, and pebbles, with and without the flexible feet. The findings were that passive jamming is most effective on flat ground and active jamming is best on loose rocks.
With the flexible feet, the robot’s legs were able to grip the ground better, which in turn increased its speed. This was especially true when the robot walked up sloped and uneven terrain.
Nick Gravish is a professor in the UC San Diego Department of Mechanical and Aerospace Engineering and study co-author.
“The natural world is filled with challenging grounds for walking robots — slippery, rocky, and squishy substrates all make walking complicated,” said Gravish. “Feet that can adapt to these different types of ground can help robots improve mobility.”
The researchers will now attempt to incorporate soft sensors on the bottom of the feet, which will allow an electronic control board to be utilized. The electronic control board would identify the type of ground that the robot is going to walk over and if the feet need to be actively or passively jammed. The researchers will also continue to improve design and control algorithms for better efficiency.
New Software Developed to Improve Robotic Prosthetics
New software has been developed by researchers at North Carolina State University in order to improve robotic prosthetics or exoskeletons. The new software is able to be integrated with existing hardware, resulting in safer and more natural walking on different terrains.
The paper is titled “Environmental Context Prediction for Lower Limb Prostheses With Uncertainty Quantification.” It was published in IEEE Transactions on Automation Science and Engineering.
Adapting to Different Terrains
Edgar Lobaton is a co-author of the paper. He is an associate professor of electrical and computer engineering at the university.
“Lower-limb robotic prosthetics need to execute different behaviors based on the terrain users are walking on,” says Lobaton. “The framework we’ve created allows the AI in robotic prostheses to predict the type of terrain users will be stepping on, quantify the uncertainties associated with that prediction, and then incorporate that uncertainty into its decision-making.”
There were six different terrains that the researchers focused on, with each requiring adjustments in the behavior of a robotic prosthetic. They were tile, concrete, brick, grass, “upstairs,” and “downstairs.”
Boxuan Zhong is the lead author of the paper and a Ph.D. graduate from NC State.
“If the degree of uncertainty is too high, the AI isn’t forced to make a questionable decision — it could instead notify the user that it doesn’t have enough confidence in its prediction to act, or it could default to a ‘safe’ mode,” says Zhong.
Incorporation of Hardware and Software Elements
The new framework relies on both hardware and software elements being incorporated together, and it is used with any lower-limb robotic exoskeleton or robotic prosthetic device.
One new aspect of this framework is a camera as another piece of hardware. In the study, cameras were worn on eyeglasses, and they were placed on the lower-limb prosthesis. The researchers then observed how AI was able to utilize computer vision data from the two different types of cameras, first separately and then together.
Helen Huang is a co-author of the paper. She is the Jackson Family Distinguished Professor of Biomedical Engineering in the Joint Department of Biomedical Engineering at NC State and the University of North Carolina at Chapel Hill.
“Incorporating computer vision into control software for wearable robotics is an exciting new area of research,” says Huang. “We found that using both cameras worked well, but required a great deal of computing power and may be cost prohibitive. However, we also found that using only the camera mounted on the lower limb worked pretty well — particularly for near-term predictions, such as what the terrain would be like for the next step or two.”
According to Lobaton, the work is applicable to any type of deep-learning system.
“We came up with a better way to teach deep-learning systems how to evaluate and quantify uncertainty in a way that allows the system to incorporate uncertainty into its decision making,” Lobaton says. “This is certainly relevant for robotic prosthetics, but our work here could be applied to any type of deep-learning system.”
Training the AI System
In order to train the AI system, the cameras were placed on able-bodied participants, who then moved through different indoor and outdoor environments. The next step was to have an individual with lower-limb amputation navigate the same environments while wearing the cameras.
“We found that the model can be appropriately transferred so the system can operate with subjects from different populations,” Lobaton says. “That means that the AI worked well even though it was trained by one group of people and used by somebody different.”
The next step is to test the framework in a robotic device.
“We are excited to incorporate the framework into the control system for working robotic prosthetics — that’s the next step,” Huang says.
The team will also work on making the system more efficient, by requiring less visual data input and data processing.
Anthony Tayoun, Co-founder & COO of Dexai Robotics – Interview Series
Anthony is the co-founder and COO of Dexai Robotics, a startup that automates activities in commercial kitchens using flexible robot arms. Prior to Dexai, Anthony worked as a consultant with the Boston Consulting Group, focusing on growth strategies. Anthony holds a MBA from Harvard Business School, and a B.E. in Mechanical Engineering and a B.S. in Mathematics from the American University of Beirut. Outside of work, Anthony enjoys chasing soccer balls and exploring sunken sea treasures.
What is it that attracted you to robotics initially?
I’m amazed by our ability, as humans, to develop “complex tools” out of simple components to improve our standard of living. At the same time, we’re living in a period during which many enabling technologies are being improved by an order of magnitude. Just look back at the past two decades: collaborative robots were created and became affordable for commercial applications, control theory advanced substantially, computer vision is arguably at the super human level, machine learning is enabling very rapid decision making, and the internet infrastructure improved enough to connect all of this together. Right now is really the most exciting time for robotics; for the first time in history, robot performance is soon going to exceed our expectations.
You have a very diverse background including being an Associate for the Boston Consulting Group (BCG). One of your projects was designing a prediction tool to detect illicit activity using advanced statistical methods and big data analysis. Could you talk about this project?
At a high level, that project involved analyzing a very large dataset, comprising demographic and behavioral data for commercial establishments, to unearth predictive behavior. We used advanced statistical modeling techniques, such as binomial regression, to compute the probability of illicit activity based on past non-related data. The results were staggering: from data such as types of licenses owned or historical financial performance, we were able to make predictions an order of magnitude more accurate than the baseline.
Can you discuss how you transitioned away from being an Associate of BCG, to launching Dexai Robotics?
My BCG experience enriched my business knowledge tremendously, as I helped companies navigate various strategic and managerial topics. During this experience, I realized that the projects I enjoy the most are those related to market entry or helping clients set up businesses from the ground up, which pushed me in the entrepreneurial direction. I decided to pursue a Master of Business Administration, and joined Harvard Business School. At HBS, I focused on entrepreneurship and related classes, and had the fortune to experiment with a few ideas at the school’s innovation lab. Midway through the MBA, I met Dave Johnson (now Dexai’s co-founder), and together we started developing business plans to commercialize technology that he and others at Harvard and MIT were developing. A few business competitions and tens of customer calls later, Dexai was born!
Dexai Robotics features AIfred a robot that automates activities in commercial kitchens and the food industry. What are the tasks that AIfred is capable of?
Alfred is currently capable of end-to-end meal assembly for a variety of recipes. Alfred can use regular utensils such as tongs, dishers (scoops), spoons, and ladles to pick and/or scoop almost any ingredient. It takes Alfred ~1 day to “learn” a new ingredient, as long as it can be manipulated using the mentioned utensils. Alfred can also “see” and identify different ingredients in the workspace, pass bowls around, and perform simple tasks such as opening a rice cooker or an oven door. In the future, Alfred will learn additional tasks such as operating kitchen equipment (e.g., fryer, grill), and perform ingredients preparation tasks (e.g., cutting, slicing).
Is there a learning curve for a restaurant operator who wishes to install AIfred in their commercial kitchen?
There is a slight learning curve, in-line with most other kitchen appliances. The initial setup consists of entering supported recipes into Dexai’s software, specifying ingredient portions, and connecting Alfred to the point-of-sale system. After that, Alfred runs pretty much on its own, with restaurant operators only needing to periodically refill food bins with fresh ingredients. Alfred is designed to simplify the lives of restaurant workers: we made a conscious choice to solve the “difficult” problem ourselves, so that our customers don’t have to worry about that. Alfred’s camera, combined with Dexai’s proprietary AI software, allows for seamless adaptation to the majority of layouts and processes. Further, Alfred can adapt to changes in the environment, such as moving a bowl around, or swapping ingredients, to maximize the operator’s flexibility.
What’s the initial reaction from restaurateurs that initially test the AIfred robot?
That’s a very interesting question because the reaction progresses very quickly. The universal initial reaction is to take out your phone and start snapping pictures and videos. There’s something really magical about a robotic arm smoothly moving around in a purposeful manner. Maybe it’s because popular culture has us expecting clunky, abrupt motions, similar to when someone makes a “robot impression”. Contrast that with the robot moving very smoothly, picking up utensils, and scooping food the same way a person would do, and your reaction dramatically changes.
Are there any brand names or large restaurants that are currently using AIfred or trialing AIfred?
We deployed a couple of successful trials to test the system, and had to pause due to concerns for our employee safety related to COVID-19. Our customer names are all still confidential and our initial focus is on salads and bowls. Later this year, we will have our first customer-facing deployment, so stay tuned!
One of your earliest robotic projects was the Mule Robot which assisted users with transporting everyday merchandise. How did this early experience influence your thinking on robotics?
My biggest learning from the Mule Robot project was that solving the technical problem is a necessary but insufficient requirement for success. Without customer focus and a robust business model, even the most elegant technical solution won’t leave the research lab. For the Mule Robot, we developed a solution for residential applications, but struggled to take the project forward. Alternatively, thinking about the same problem with a more commercial lens: transporting merchandise inside a building is perfect for “room service” applications in hospitality. Today, a Chicago hotel uses two robots automating room service, made by a startup that successfully commercialized a similar project.
What do you believe a commercial kitchen of the future will look like? How will robots cooperate or in some cases replacing kitchen staff?
I believe that kitchen staff will always be needed; hospitality is incomplete without a human touch. Regarding the kitchen of the future, the answer really depends on how far in the future we’re looking. In the short and medium term, we’ll see dramatic efficiency increases in different areas of the kitchen, either through automated single-use equipment such as sushi rollers and vegetable slicers, or through end-to-end flexible automation such as ingredient assembly through Dexai’s Alfred. Longer term, in 10 years or so, the commercial kitchen will capitalize on efficiencies by combining all these solutions, and will feature novel cooking techniques instead of only efficiency gains. To illustrate this point, imagine a circular, vertically stacked serving counter operated by a robot at the center which can reach inside the oven and make changes to the meal while it cooks. Eventually, the target is to get from raw ingredients to prepared meals through the smallest and most efficient operation.
Is there anything else that you would like to share about Dexai Robotics or AIfred?
We’re really excited to have Alfred’s first public appearance this year. Especially given the health crisis that our world is suffering from, securing the access to prepared food is a necessity. We look forward to a future where everyone has access to affordable, healthy foods!
Thank you for the fantastic interview. I look forward to the day when we see different version of AIfred in commercial kitchens everywhere. Anyone who wishes to learn more should visit Dexai Robotics.
- Robots Walk Faster With Newly Developed Flexible Feet
- Bradford Newman, Chair of North America Trade Secrets Practice – Interview Series
- Eugene Terekhov, CEO of AiBUY – Interview Series
- How Riiid! is Helping to Bring in New Era of AI-Education
- Vincent Scesa – Autonomous Vehicle Program Manager, EasyMile – Interview Series