Human in the Loop

Picture a robot that has never been told how its own body works, yet watches itself move and gradually learns to understand its physical form through vision alone. No embedded sensors, no pre-programmed models, no expensive hardware—just a single camera and the computational power to make sense of what it sees. This isn't science fiction; it's the reality emerging from MIT's Computer Science and Artificial Intelligence Laboratory, where researchers have developed a system that could fundamentally change how we think about robotic control.

When Robots Learn to Know Themselves

The traditional approach to robotic control reads like an engineering manual written in advance of the machine it describes. Engineers meticulously map every joint, calculate precise kinematics, and embed sensors throughout the robot's body to track position, velocity, and force. It's a process that works, but it's also expensive, complex, and fundamentally limited to robots whose behaviour can be predicted and modelled beforehand.

Neural Jacobian Fields represent a radical departure from this paradigm. Instead of telling a robot how its body works, the system allows the machine to figure it out by watching itself move. The approach eliminates the need for embedded sensors entirely, relying instead on a single external camera to provide all the visual feedback necessary for sophisticated control.

The implications extend far beyond mere cost savings. Traditional sensor-based systems struggle with robots made from soft materials, bio-inspired designs, or multi-material constructions where the physics become too complex to model accurately. These machines—which might include everything from flexible grippers to biomimetic swimmers—have remained largely out of reach for precise control systems. Neural Jacobian Fields change that equation entirely.

Researchers at MIT CSAIL have demonstrated that their vision-based system can learn to control diverse robots without any prior knowledge of their mechanical properties. The robot essentially builds its own internal model of how it moves by observing the relationship between motor commands and the resulting visual changes captured by the camera. The system enables robots to develop what researchers describe as a form of self-awareness through visual observation—a type of embodied understanding that emerges naturally from watching and learning.

The breakthrough represents a fundamental shift from model-based to learning-based control. Rather than creating precise, often brittle mathematical models of robots, the focus moves towards data-driven approaches where robots learn their own control policies through interaction and observation. This mirrors a broader trend in robotics where adaptability and learning play increasingly central roles in determining behaviour.

The technology also highlights the growing importance of computer vision in robotics. As cameras become cheaper and more capable, and as machine learning approaches become more sophisticated, vision-based approaches are becoming viable alternatives to traditional sensor modalities. This trend extends beyond robotics into autonomous vehicles, drones, and smart home systems.

The Mathematics of Self-Discovery

At the heart of this breakthrough lies a concept called the visuomotor Jacobian field—an adaptive representation that directly connects what a robot sees to how it should move. In traditional robotics, Jacobian matrices describe the relationship between joint velocities and end-effector motion, requiring detailed knowledge of the robot's kinematic structure. The Neural Jacobian Field approach inverts this process, inferring these relationships purely from visual observation.

The system works by learning to predict how small changes in motor commands will affect what the camera sees. Over time, this builds up a comprehensive understanding of the robot's capabilities and limitations, all without requiring any explicit knowledge of joint angles, link lengths, or material properties. It's a form of self-modelling that emerges naturally from the interaction between action and observation.

This control map becomes remarkably sophisticated. The system can understand not just how the robot moves, but how different parts of its body interact and how to execute complex movements through space. The robot develops a form of physical self-perception, understanding its own capabilities through empirical observation rather than theoretical calculation. This self-knowledge extends to understanding the robot's workspace boundaries, the effects of gravity on different parts of its structure, and even how wear or damage might affect its movement patterns.

The computational approach builds on recent advances in deep learning, particularly in the area of implicit neural representations. Rather than storing explicit models of the robot's geometry or dynamics, the system learns a continuous function that can be queried at any point to understand the local relationship between motor commands and visual feedback. This allows the approach to scale to robots of varying complexity without requiring fundamental changes to the underlying approach.

The neural network architecture that enables this learning represents a sophisticated integration of computer vision and control theory. The system must simultaneously process high-dimensional visual data and learn the complex mappings between motor commands and their visual consequences. This requires networks capable of handling both spatial and temporal relationships, understanding not just what the robot looks like at any given moment, but how its appearance changes in response to different actions.

The visuomotor Jacobian field effectively replaces the analytically derived Jacobian matrix used in classical robotics. This movement model becomes a continuous function that maps the robot's configuration to the visual changes produced by its motor commands. The elegance of this approach lies in its generality—the same fundamental mechanism can work across different robot designs, from articulated arms to soft manipulators to swimming robots.

Beyond the Laboratory: Real-World Applications

The practical implications of this technology extend across numerous domains where traditional robotic control has proven challenging or prohibitively expensive. In manufacturing, the ability to control robots without embedded sensors could dramatically reduce the cost of automation, making robotic solutions viable for smaller-scale operations that couldn't previously justify the investment. Small manufacturers, artisan workshops, and developing economies could potentially find sophisticated robotic assistance within their reach.

Soft robotics represents perhaps the most immediate beneficiary of this approach. Robots made from flexible materials, pneumatic actuators, or bio-inspired designs have traditionally been extremely difficult to control precisely because their behaviour is hard to model mathematically. The Neural Jacobian Field approach sidesteps this problem entirely, allowing these machines to learn their own capabilities through observation. MIT researchers have successfully demonstrated the system controlling a soft robotic hand to grasp objects, showing how flexible systems can learn to adapt their compliant fingers to different shapes and develop strategies that would be nearly impossible to program explicitly.

These soft systems have shown great promise for applications requiring safe interaction with humans or navigation through confined spaces. However, their control has remained challenging precisely because their behaviour is difficult to model mathematically. Vision-based control could unlock the potential of these systems by allowing them to learn their own complex dynamics through observation. The approach might enable new forms of bio-inspired robotics, where engineers can focus on replicating the mechanical properties of biological systems without worrying about how to sense and control them.

The technology also opens new possibilities for field robotics, where robots must operate in unstructured environments far from technical support. A robot that can adapt its control strategy based on visual feedback could potentially learn to operate in new configurations without requiring extensive reprogramming or recalibration. This could prove valuable for exploration robots, agricultural machines, or disaster response systems that need to function reliably in unpredictable conditions.

Medical robotics presents another compelling application area. Surgical robots and rehabilitation devices often require extremely precise control, but they also need to adapt to the unique characteristics of each patient or procedure. A vision-based control system could potentially learn to optimise its behaviour for specific tasks, improving both precision and effectiveness. Rehabilitation robots, for example, could adapt their assistance patterns based on observing a patient's progress and changing needs over time.

The approach could potentially benefit prosthetics and assistive devices. Current prosthetic limbs often require extensive training for users to learn complex control interfaces. A vision-based system could potentially observe the user's intended movements and adapt its control strategy accordingly, creating more intuitive and responsive artificial limbs. The system could learn to interpret visual cues about the user's intentions, making the prosthetic feel more like a natural extension of the body.

The Technical Architecture

The Neural Jacobian Field system represents a sophisticated integration of computer vision, machine learning, and control theory. The architecture begins with a standard camera that observes the robot from an external vantage point, capturing the full range of the machine's motion in real-time. This camera serves as the robot's only source of feedback about its own state and movement, replacing arrays of expensive sensors with a single, relatively inexpensive visual system.

The visual input feeds into a deep neural network trained to understand the relationship between pixel-level changes in the camera image and the motor commands that caused them. This network learns to encode a continuous field that maps every point in the robot's workspace to a local Jacobian matrix, describing how small movements in that region will affect what the camera sees. The network processes not just static images, but the dynamic visual flow that reveals how actions translate into change.

The training process requires the robot to execute a diverse range of movements while the system observes the results. Initially, these movements explore the robot's capabilities, allowing the system to build a comprehensive understanding of how the machine responds to different commands. The robot might reach in various directions, manipulate objects, or simply move its joints through their full range of motion. Over time, the internal model becomes sufficiently accurate to enable sophisticated control tasks, from precise positioning to complex manipulation.

One of the notable aspects of the system is its ability to work across different robot configurations. The neural network architecture can learn to control robots with varying mechanical designs without fundamental modifications. This generality stems from the approach's focus on visual feedback rather than specific mechanical models. The system learns principles about how visual changes relate to movement that can apply across different robot designs.

The control loop operates in real-time, with the camera providing continuous feedback about the robot's current state and the neural network computing appropriate motor commands to achieve desired movements. The system can handle both position control, where the robot needs to reach specific locations, and trajectory following, where it must execute complex paths through space. The visual feedback allows for immediate correction of errors, enabling the robot to adapt to unexpected obstacles or changes in its environment.

The computational requirements, while significant, remain within the capabilities of modern hardware. The system can run on standard graphics processing units, making it accessible to research groups and companies that might not have access to specialised robotic hardware. This accessibility is important for the technology's potential to make advanced robotic control more widely available.

The approach represents a trend moving away from reliance on internal, proprioceptive sensors towards using rich, external visual data as the primary source of feedback for robotic control. Neural Jacobian Fields exemplify this shift, demonstrating that sophisticated control can emerge from careful observation of the relationship between actions and their visual consequences.

Democratising Robotic Intelligence

Perhaps one of the most significant long-term impacts of Neural Jacobian Fields lies in their potential to make sophisticated robotic control more accessible. Traditional robotics has been dominated by large institutions and corporations with the resources to develop complex sensor systems and mathematical models. The barrier to entry has remained stubbornly high, limiting innovation to well-funded research groups and established companies.

Vision-based control systems could change this dynamic. A single camera and appropriate software could potentially replace substantial investments in embedded sensors, making advanced robotic control more accessible to smaller research groups, educational institutions, and individual inventors. While the approach still requires technical expertise in machine learning and robotics, it eliminates the need for detailed kinematic modelling and complex sensor integration.

This increased accessibility could accelerate innovation in unexpected directions. Researchers working on problems in biology, materials science, or environmental monitoring might find robotic solutions more within their reach, leading to applications that traditional robotics companies might never have considered. The history of computing suggests that transformative innovations often come from unexpected quarters once the underlying technology becomes more accessible.

Educational applications represent another significant opportunity. Students learning robotics could focus on high-level concepts and applications while still engaging with the mathematical foundations of control theory. This could help train a new generation of roboticists with a more intuitive understanding of how machines move and interact with their environment. Universities with limited budgets could potentially offer hands-on robotics courses without investing in expensive sensor arrays and specialised hardware.

The democratisation extends beyond formal education to maker spaces, hobbyist communities, and entrepreneurial ventures. Individuals with creative ideas for robotic applications could prototype and test their concepts without the traditional barriers of sensor integration and control system development. This could lead to innovation in niche applications, artistic installations, and novel robotic designs that push the boundaries of what we consider possible.

Small businesses and developing economies could particularly benefit from this accessibility. Manufacturing operations that could never justify the cost of traditional robotic systems might find vision-based robots within their reach. This could help level the playing field in global manufacturing, allowing smaller operations to compete with larger, more automated facilities.

The potential economic implications extend beyond the robotics industry itself. By reducing the cost and complexity of robotic control, the technology could accelerate automation in sectors that have previously found robotics economically unviable. Small-scale manufacturing, agriculture, and service industries could all benefit from more accessible robotic solutions.

Challenges and Limitations

Despite its promise, the Neural Jacobian Field approach faces several significant challenges that will need to be addressed before it can achieve widespread adoption. The most fundamental limitation lies in the quality and positioning of the external camera. Unlike embedded sensors that can provide precise measurements regardless of environmental conditions, vision-based systems remain vulnerable to lighting changes, occlusion, and camera movement.

Lighting conditions present a particular challenge. The system must maintain accurate control across different illumination levels, from bright sunlight to dim indoor environments. Shadows, reflections, and changing light sources can all affect the visual feedback that the system relies upon. While modern computer vision techniques can handle many of these variations, they add complexity and potential failure modes that don't exist with traditional sensors.

The learning process itself requires substantial computational resources and training time. While the system can eventually control robots without embedded sensors, it needs significant amounts of training data to build accurate models. This could limit its applicability in situations where robots need to begin operating immediately or where training time is severely constrained. The robot must essentially learn to walk before it can run, requiring a period of exploration and experimentation that might not be practical in all applications.

Robustness represents another ongoing challenge. Traditional sensor-based systems can often detect and respond to unexpected situations through direct measurement of forces, positions, or velocities. Vision-based systems must infer these quantities from camera images, potentially missing subtle but important changes in the robot's state or environment. A loose joint, worn component, or unexpected obstacle might not be immediately apparent from visual observation alone.

The approach also requires careful consideration of safety, particularly in applications where robot malfunction could cause injury or damage. While the system has shown impressive performance in laboratory settings, proving its reliability in safety-critical applications will require extensive testing and validation. The lack of direct force feedback could be particularly problematic in applications involving human interaction or delicate manipulation tasks.

Occlusion presents another significant challenge. If parts of the robot become hidden from the camera's view, the system loses crucial feedback about those components. This could happen due to the robot's own movements, environmental obstacles, or the presence of humans or other objects in the workspace. Developing strategies to handle partial occlusion or to use multiple cameras effectively remains an active area of research.

The computational demands of real-time visual processing and neural network inference can be substantial, particularly for complex robots or high-resolution cameras. While modern hardware can handle these requirements, the energy consumption and processing power needed might limit deployment in battery-powered or resource-constrained applications.

The Learning Process and Adaptation

One of the most fascinating aspects of Neural Jacobian Fields is how they learn. Unlike traditional machine learning systems that are trained on large datasets and then deployed, these systems learn continuously through interaction with their environment. The robot's understanding of its own capabilities evolves over time as it gains more experience with different movements and situations.

This continuous learning process means that the robot's performance can improve over its operational lifetime. Small changes in the robot's physical configuration, whether due to wear, maintenance, or intentional modifications, can be accommodated automatically as the system observes their effects on movement. A robot might learn to compensate for a slightly loose joint or adapt to the addition of new tools or attachments.

The robot's learning follows recognisable stages. Initially, movements are exploratory and somewhat random as the system builds its basic understanding of cause and effect. Gradually, more purposeful movements emerge as the robot learns to predict the consequences of its actions. Eventually, the system develops the ability to plan complex movements and execute them with precision.

This learning process is robust to different starting conditions. Robots with different mechanical designs can learn effective control strategies using the same basic approach. The system discovers the unique characteristics of each robot through observation, adapting its strategies to work with whatever physical capabilities are available.

The continuous nature of the learning also means that robots can adapt to changing conditions over time. Environmental changes, wear and tear, or modifications to the robot's structure can all be accommodated as the system observes their effects and adjusts accordingly. This adaptability could prove crucial for long-term deployment in real-world applications where conditions are never perfectly stable.

The approach enables a form of learning that mirrors biological development, where motor skills emerge through exploration and practice rather than explicit instruction. This parallel suggests that vision-based motor learning may reflect fundamental principles of how intelligent systems acquire physical capabilities.

Scaling and Generalisation

The ability of Neural Jacobian Fields to work across different robot configurations is one of their most impressive characteristics. The same basic approach can learn to control robots with different mechanical designs, from articulated arms to flexible swimmers to legged walkers. This generality suggests that the approach captures something fundamental about the relationship between vision and movement.

This generalisation capability could be important for practical deployment. Rather than requiring custom control systems for each robot design, manufacturers could potentially use the same basic software framework across multiple product lines. This could reduce development costs and accelerate the introduction of new robot designs. The approach might enable more standardised robotics where new mechanical designs can be controlled effectively without extensive software development.

The system's ability to work with compliant robots is particularly noteworthy. These machines, made from flexible materials that can bend, stretch, and deform, have shown great promise for applications requiring safe interaction with humans or navigation through confined spaces. However, their control has remained challenging precisely because their behaviour is difficult to model mathematically. Vision-based control could unlock the potential of these systems by allowing them to learn their own complex dynamics through observation.

The approach might also enable new forms of modular robotics, where individual components can be combined in different configurations without requiring extensive recalibration or reprogramming. If a robot can learn to understand its own body through observation, it might be able to adapt to changes in its physical configuration automatically. This could lead to more flexible and adaptable robotic systems that can be reconfigured for different tasks.

The generalisation extends beyond just different robot designs to different tasks and environments. A robot that has learned to control itself in one setting can often adapt to new situations relatively quickly, building on its existing understanding of its own capabilities. This transfer learning could make robots more versatile and reduce the time needed to deploy them in new applications.

The success of the approach across diverse robot types suggests that it captures principles about motor control that apply regardless of specific mechanical implementation. This universality could be key to developing more general robotic intelligence that isn't tied to particular hardware configurations.

Expanding Applications and Future Possibilities

The Neural Jacobian Field approach represents a convergence of several technological trends that have been developing independently for years. Computer vision has reached a level of sophistication where single cameras can extract remarkably detailed information about three-dimensional scenes. Machine learning approaches have become powerful enough to find complex patterns in high-dimensional data. Computing hardware has become fast enough to process this information in real-time.

The combination of these capabilities creates opportunities that were simply not feasible even a few years ago. The ability to control sophisticated robots using only visual feedback represents a qualitative leap in what's possible with relatively simple hardware configurations. This technological convergence also suggests that similar breakthroughs may be possible in other domains where complex systems need to be controlled or understood.

The principles underlying Neural Jacobian Fields could potentially be applied to problems in autonomous vehicles, manufacturing processes, or even biological systems where direct measurement is difficult or impossible. The core insight—that complex control can emerge from careful observation of the relationship between actions and their visual consequences—has applications beyond robotics.

In autonomous vehicles, similar approaches might enable cars to learn about their own handling characteristics through visual observation of their movement through the environment. Manufacturing systems could potentially optimise their operations by observing the visual consequences of different process parameters. Even in biology, researchers might use similar techniques to understand how organisms control their movement by observing the relationship between neural activity and resulting motion.

The technology might also enable new forms of robot evolution, where successful control strategies learned by one robot could be transferred to others with similar capabilities. This could create a form of collective learning where the robotics community as a whole benefits from the experiences of individual systems. Robots could share their control maps, accelerating the development of new capabilities across populations of machines.

The success of Neural Jacobian Fields opens numerous avenues for future research and development. One promising direction involves extending the approach to multi-robot systems, where teams of machines could learn to coordinate their movements through shared visual feedback. This could enable new forms of collaborative robotics that would be extremely difficult to achieve through traditional control methods.

Another area of investigation involves combining vision-based control with other sensory modalities. While the current approach relies solely on visual feedback, incorporating information from audio, tactile, or other sensors could enhance the system's capabilities and robustness. The challenge lies in maintaining the simplicity and generality that make the vision-only approach so appealing.

Implications for Human-Robot Interaction

As robots become more capable of understanding their own bodies through vision, they may also become better at understanding and interacting with humans. The same visual processing capabilities that allow a robot to model its own movement could potentially be applied to understanding human gestures, predicting human intentions, or adapting robot behaviour to human preferences.

This could lead to more intuitive forms of human-robot collaboration, where people can communicate with machines through natural movements and gestures rather than explicit commands or programming. The robot's ability to learn and adapt could make these interactions more fluid and responsive over time. A robot working alongside a human might learn to anticipate their partner's needs based on visual cues, creating more seamless collaboration.

The technology might also enable new forms of robot personalisation, where machines adapt their behaviour to individual users based on visual observation of preferences and patterns. This could be particularly valuable in healthcare, education, or domestic applications where robots need to work closely with specific individuals over extended periods. A care robot, for instance, might learn to recognise the subtle signs that indicate when a patient needs assistance, adapting its behaviour to provide help before being asked.

The potential for shared learning between humans and robots is particularly intriguing. If robots can learn through visual observation, they might be able to watch humans perform tasks and learn to replicate or assist with those activities. This could create new forms of robot training where machines learn by example rather than through explicit programming.

The visual nature of the feedback also makes the robot's learning process more transparent to human observers. People can see what the robot is looking at and understand how it's learning to move. This transparency could build trust and make human-robot collaboration more comfortable and effective.

Economic and Industrial Impact

For established robotics companies, the technology presents both opportunities and challenges. While it could reduce manufacturing costs and enable new applications, it might also change competitive dynamics in the industry. Companies will need to adapt their strategies to remain relevant in a world where sophisticated control capabilities become more widely accessible.

The approach could also enable new business models in robotics, where companies focus on software and learning systems rather than hardware sensors and mechanical design. This could lead to more rapid innovation cycles and greater specialisation within the industry. Companies might develop expertise in particular types of learning or specific application domains, creating a more diverse and competitive marketplace.

The democratisation of robotic control could also have broader economic implications. Regions that have been excluded from the robotics revolution due to cost or complexity barriers might find these technologies more accessible. This could help reduce global inequalities in manufacturing capability and create new opportunities for economic development.

The technology might also change the nature of work in manufacturing and other industries. As robots become more accessible and easier to deploy, the focus might shift from operating complex machinery to designing and optimising robotic systems. This could create new types of jobs while potentially displacing others, requiring careful consideration of the social and economic implications.

Rethinking Robot Design

The availability of vision-based control systems could fundamentally change how robots are designed and manufactured. When embedded sensors are no longer necessary for precise control, engineers gain new freedom in choosing materials, form factors, and mechanical designs. This could lead to robots that are lighter, cheaper, more robust, or better suited to specific applications.

The elimination of sensor requirements could enable new categories of robots. Disposable robots for dangerous environments, ultra-lightweight robots for delicate tasks, or robots made from unconventional materials could all become feasible. The design constraints that have traditionally limited robotic systems could be relaxed, opening up new possibilities for innovation.

The approach might also enable new forms of bio-inspired robotics, where engineers can focus on replicating the mechanical properties of biological systems without worrying about how to sense and control them. This could lead to robots that more closely mimic the movement and capabilities of living organisms.

The reduced complexity of sensor integration could also accelerate the development cycle for new robot designs. Prototypes could be built and tested more quickly, allowing for more rapid iteration and innovation. This could lead to a more dynamic and creative robotics industry where new ideas can be explored more easily.

The Path Forward

Neural Jacobian Fields represent more than just a technical advance; they embody a fundamental shift in how we think about robotic intelligence and control. By enabling machines to understand themselves through observation rather than explicit programming, the technology opens possibilities that were previously difficult to achieve.

The journey from laboratory demonstration to widespread practical application will undoubtedly face numerous challenges. Questions of reliability, safety, and scalability will need to be addressed through careful research and testing. The robotics community will need to develop new standards and practices for vision-based control systems.

Researchers are also exploring ways to accelerate the learning process, potentially through simulation, transfer learning, or more sophisticated training approaches. Reducing the time required to train new robots could make the approach more practical for commercial applications where rapid deployment is essential.

Yet the potential rewards justify the effort. A world where robots can learn to understand themselves through vision alone is a world where robotic intelligence becomes more accessible, more adaptable, and more aligned with the complex, unpredictable nature of real-world environments. The robots of the future may not need to be told how they work—they'll simply watch themselves and learn.

As this technology continues to develop, it promises to blur the traditional boundaries between artificial and biological intelligence, creating machines that share some of the adaptive capabilities that have made biological organisms so successful. In doing so, Neural Jacobian Fields may well represent a crucial step towards truly autonomous, intelligent robotic systems that can thrive in our complex world.

The implications extend beyond robotics into our broader understanding of intelligence, learning, and adaptation. By demonstrating that sophisticated control can emerge from simple visual observation, this research challenges our assumptions about what forms of knowledge are truly necessary for intelligent behaviour. In a sense, these robots are teaching us something fundamental about the nature of learning itself.

The future of robotics may well be one where machines learn to understand themselves through observation, adaptation, and continuous interaction with the world around them. In this future, the robots won't just follow our instructions—they'll watch, learn, and grow, developing capabilities we never explicitly programmed but that emerge naturally from their engagement with reality itself.

This vision of self-aware, learning robots represents a profound shift in our relationship with artificial intelligence. Rather than creating machines that simply execute our commands, we're developing systems that can observe, learn, and adapt in ways that mirror the flexibility and intelligence of biological organisms. The robots that emerge from this research may be our partners in understanding and shaping the world, rather than simply tools for executing predetermined tasks.

If robots can learn to see and understand themselves, the possibilities for what they might achieve alongside us become truly extraordinary.

References

1. MIT Computer Science and Artificial Intelligence Laboratory. “Robots that know themselves: MIT's vision-based system teaches machines self-awareness.” Available at: www.csail.mit.edu

2. Li, S.L., et al. “Controlling diverse robots by inferring Jacobian fields with deep learning.” PubMed Central. Available at: pmc.ncbi.nlm.nih.gov

3. MIT EECS. “Robotics Research.” Available at: www.eecs.mit.edu

4. MIT EECS Faculty. “Daniela Rus.” Available at: www.eecs.mit.edu

5. arXiv. “Neural feels with neural fields: Visuo-tactile perception for in-hand manipulation.” Available at: arxiv.org

---

Tim Green UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0000-0002-0156-9795 Email: tim@smarterarticles.co.uk

Discuss...