Biomimetic AI. The Ultimate Imitation Game: How AI is Teaching Us to Think Like Nature
Generative AI art of neurons by Vivika Martini
Artificial intelligence, at the forefront of contemporary technological evolution, embodies transformative potential that reshapes our societal landscape and the very essence of innovation. As AI emerges as a pivotal force in shaping the future, developing robust AI systems mirrors the intricate patterns and processes in the natural world. Numerous AI researchers are venturing into uncharted territories in computing, gleaning insights from these natural phenomena.
In many aspects, nature served as the original genius behind the intricate algorithms governing the biological realm. Even before humanity stepped foot on this planet, nature had already mastered the art of problem-solving. Operating with a delicate balance of complexity and efficiency, these natural algorithms have evolved and thrived over billions of years, shaped by the unforgiving forces of evolution and survival. These awe-inspiring biological processes have long captivated AI researchers.
Contemporary AI draws inspiration from nature’s ingenuity. For instance, the design of neural networks mirrors the intricate mesh of human neural connections, forming a cornerstone of AI advancement. Similarly, genetic algorithms emulate the very essence of nature’s evolutionary playbook, iteratively refining potential solutions across generations until the optimal outcome is reached.
Categorising Artificial Intelligience (AI)
Artificial Intelligence (AI) can be classified into different types based on capabilities and functionalities. Based on capabilities, AI can be categorised into three types:
Narrow AI: These are AI systems designed to perform specific tasks such as voice or image recognition. They operate within a limited pre-defined range of functions. Examples include Apple’s Siri, Google Translate, and IBM’s Watson supercomputer.
General AI: These AI systems can perform any intellectual task that a human being can do. They are designed to learn, think, and complete at similar levels to humans. However, this type of AI is currently theoretical and does not exist yet.
Super AI: This type of AI surpasses human intelligence and capabilities. It is also currently theoretical and does not exist yet.
Based on functional awareness, AI can be categorised into four types:
Reactive Machines: These are the most basic types of AI systems that can only react to current scenarios and cannot use previous experiences to inform everyday decisions. They are incapable of forming memory or storing information for future use.
Limited Memory: These AI systems can store past experiences or data quickly. This feature allows them to learn and train for future tasks.
Theory of Mind: This is a more advanced type of AI that can understand, interpret, and respond to emotions. It can also perform tasks of limited memory machines. However, this type of AI is currently theoretical and does not exist yet.
Self-aware AI: This is the most advanced type of AI. It can understand others’ emotions, has a sense of self, and possesses human-level intelligence. Like the Theory of Mind AI, this type of AI is also currently theoretical and does not exist yet.
It’s important to note that the development of General AI, Super AI, Theory of Mind AI, and Self-aware AI are ongoing research areas in AI.
Alan Turing.
Turing’s Insights
To comprehend the evolution of AI, we must journey back to its roots with Alan Turing and his iconic Turing machine. While Alan Turing is widely celebrated for his groundbreaking contributions to computer science and cryptography, his intellectual pursuits extended beyond mere machinery.
Turing’s fascination with living systems and nature-inspired innovation, the science of emulating nature’s models and systems, was evident in his final paper, “The Chemical Basis of Morphogenesis”, published in 1952. In this seminal work, Turing delved deep into the realm of biology, exploring the mechanisms behind forming patterns in nature.
“The original question, ‘Can machines think?’ I believe to be too meaningless to deserve discussion.”
Turing introduced the concept of reaction-diffusion systems, hypothesising that chemical substances could give rise to intricate patterns, such as the spots on a leopard or the spirals in a sunflower when interacting and diffusing within a medium.
This revolutionary theory suggested that simple chemical interactions governed by mathematical equations could explain nature's diverse and complex patterns. Turing's exploration not only bridged the gap between mathematics and biology but also laid the foundation for understanding how simple rules and interactions could lead to the emergence of complexity – a principle that AI researchers have since adopted and refined.
Sunflowers
By imitating nature, AI has achieved feats previously thought impossible. From creating intricate artworks to simulating complex biological systems, AI showcases the immense power of computational biomimicry. It is a testament to Turing's belief that understanding and emulating nature can unlock new frontiers in technology and science. This poses a compelling question: could computational biomimicry be the ultimate imitation game?
Computational Beauty of Nature
One remarkable illustration of the convergence of AI and nature lies in computer scientist Peter J. Bentley's work. Peter Bentley's research encapsulates the true spirit of the imitation game, wherein computational systems replicate human intelligence and mirror the intricate patterns and processes found in nature. This harmonious blend enriches our understanding of AI's potential and ability to unlock nature's boundless wonders.
“Natural systems provide unique examples of computation in a form very different from contemporary computer architectures. Biology also demonstrates capabilities such as adaptation, self-repair and self-organisation that are becoming increasingly desirable for our technology. Just think what digital biology could do for us. Perhaps it could evolve new designs for us, think up ways to detect fraud using digital neurons or solve scheduling problems with ants.
Perhaps it could detect hackers with immune systems or create music from the patterns of growth of digital seashells. Perhaps it would allow our computers to become creative and inventive.”
You read further into Peter Bentley’s work in his Digitial Biology Lab at his research lab at UCL.
Aims - to understand and exploit fundamental computation within biology. We examine biological systems at all levels, from proteins, cells, organs and organisms to societies, species and ecosystems.
Focus - complex systems, including evolutionary, developmental, neural, swarming and other biologically-inspired computations.
Applications - research includes sustainability, design, robot control, architecture, fraud detection, materials science, art, economics, music composition, recommender systems, and models for biologists.
A Symbiotic Relationship
With the advancement of AI, we are gaining a deeper understanding of and safeguarding the natural world. Machine learning algorithms, powered by neural networks and vast amounts of data, now analyze massive climate change datasets, aiding scientists in predicting and mitigating its impacts.
These intelligent systems not only recognize patterns and trends that were previously overlooked, but they also simulate future scenarios, considering variables that may have otherwise been difficult to account for. AI-driven robots, equipped with sensors and cameras inspired by nature’s designs, venture into the depths of oceans and the far reaches of space, unveiling once-inaccessible secrets and unlocking new realms of knowledge.
UK research in computer science from The State of Nature-inspired Innovation.
By emulating the remarkable abilities of organisms like marine creatures and celestial navigators, these robots are pushing the boundaries of exploration, allowing us to gather essential data and expand our understanding of the interconnectedness of all living systems. However, as AI continues to evolve and learn from nature, we realise that the intricate complexities of our environment offer a rich tapestry of enigmas and puzzles, challenging AI researchers to develop novel algorithms and frameworks that can genuinely comprehend and respond to nature's intricacies.
From deciphering the complex communication networks of bee colonies to unravelling the secrets of biomimetic materials, nature's adaptive strategies provide an endless source of inspiration for AI innovation. In this symbiotic relationship, AI not only aids in our quest to protect and preserve the natural world but also enhances our appreciation for the beauty and resilience of the ecosystems that sustain us.
Learning from the Masters
Nature-inspired innovation (biomimicry) is an enthralling domain of AI research, where scientists and engineers strive to emulate the strategies and designs prevalent in the natural world. Birds impart wisdom on drone designs, ant colony behaviours shape logistics algorithms, and octopus camouflage abilities stimulate ideas for adaptive metamaterials.
Through studying these natural phenomena, AI researchers not only create more efficient and adaptable technologies but also gain a profound admiration for the brilliance of nature. It's a humbling realisation that despite our technological advancements, there is still much to glean from the world around us.
Numerous organisations have embraced and are presently integrating nature's intelligence into their systems and products, capitalising on the inherent wisdom of the natural world, like we do! (Our research on the State of Nature-inspired Innovation highlights this.
Logo of Google Deepmind
Google Deepmind
The UK-based AI research company Google Deepmind integrates natural systems into its AI systems by utilising machine learning and systems neuroscience techniques. They aim to combine the best practices from these fields to build AI systems capable of general intelligence.
One example of this integration is seen in their work on AlphaFold. AlphaFold is an AI system developed by Google DeepMind that uses deep learning and advanced algorithms to predict the 3D structure of proteins. This approach is inspired by the way natural systems fold proteins, and it has the potential to revolutionise the field of structural biology.
Another example of this integration is AlphaGo, an AI system created by DeepMind that achieved groundbreaking success in the game of Go. AlphaGo utilised deep neural networks and reinforcement learning techniques to learn and improve its gameplay. The system's architecture was inspired by how human brains process information, with neural networks mimicking the structure and functioning of neurons.
Another Brain logo
Another Brain
Another company pioneering nature-inspired AI systems is Another Brain, based in France. They are developing a system known as Organic AI, which emulates the cognitive processes and capabilities of the human brain. Organic AI aims to create AI systems that can understand and analyse complex data, make intelligent decisions, adapt to new situations, and learn from experience.
This has applications in many fields, including Robotics, Healthcare, Natural Language Processing, Manufacturing, and Autonomous Vehicles. Organic AI represents a departure from traditional rule-based or symbolic AI approaches, aiming to bring AI systems closer to the natural intelligence exhibited by humans and other living beings.
Opteran logo
Opteran
Similar to the previous two examples, the UK-based company Opteran strives to develop machines and systems that think and behave like natural organisms. Opteran Mind technology uses insect brain algorithms as inspiration for its software. This software can help promote machine autonomy, allowing for motion to be perceived in real-time alongside stabilising vision.
This can enable machines to map the environment around them accurately. This has a variety of mapping and navigational applications in drones, AMRs, and ADAS/AVs. While these examples demonstrate how nature can inspire and inform AI development, it's crucial to recognise the unique challenges and opportunities this approach presents.
Advanced Research and Invention Agency website
Advanced Research and Invention Agency
Taking cues from existing work, the UK version of the US DARPA model, ARIA (Advanced Research + Invention Agency), delves into understanding nature's computational methods, believing such insights could lead to vastly more efficient computing systems. These are the questions they are exploring:
What alternative vectors could dramatically improve computing performance without relying on shrinking transistors?
Energy minimisation in physical systems is a highly efficient computational mechanism. Why has so little effort gone into exploiting such phenomena, and could they underpin a new platform for massively scalable AI compute?
What new ideas could emerge by breaking down the barriers between scientists on the front lines of AI algorithmic research and those building the hardware on which those algorithms run? How could these ideas impact what we know about biological function?
What if, in trying to engineer computers that mimic the natural world in new ways, we could better understand how the natural world processes information?
As of December 2023, the programme director, Suraj Bramhavar, is exploring the space to understand the existing and future capabilities across the UK via his programme thesis.
Read more about ARIA - Unlocking AI compute hardware at 1/1000th the cost.
A Symbiotic Relationship
As artificial intelligence progresses, it enhances our understanding of the natural world and aids in its conservation. Extensive datasets about climate change are analysed through machine learning algorithms, enabling scientists to predict and mitigate its adverse impacts. Furthermore, AI-powered robots are venturing into the unexplored depths of the oceans and the vast expanse of outer space, unravelling enigmas that were once beyond our reach. Nature is an ongoing inspiration and challenge for AI researchers, presenting them with intricacies that foster innovation and push the boundaries of AI's potential.
The Future of the Imitation Game
When we think about the future of AI systems, we can predict that natural systems will play a more significant role as the framework for these systems moving forward. The current development of traditional AI systems requires computing power that increases exponentially with every innovation cycle. This is emphasised by the average session by a single user on ChatGPT, requiring approximately 150 times more power than the human brain uses when performing all functions!
This has also led to innovation start-up costs on new cutting-edge AI technology skyrocketing, isolating smaller innovation groups and allowing only a handful of firms to participate in this development. This is not a sustainable practice as we advance. If we wish to continue our foray into the world of artificial intelligence, then we need to find new ways to mitigate the exponential computing power required.
The relationship between AI and nature is symbiotic. As we strive to create machines that think and act more like living beings, we also learn to see the world holistically. The ultimate goal is to build more intelligent machines and foster a deeper connection between humanity, technology, and the natural world.
In the ultimate imitation game, the winners are not just those who create the most advanced AI but those who, in the process, gain a greater appreciation for the beauty and complexity of nature. As we continue to learn from and emulate the natural world, we're advancing technology and enriching our understanding of the universe and our place within it.
Written by Vivika Martini, Lucas Gater and Richard James MacCowan.
Hi, we're Biomimicry Innovation Lab. We partner with founders and leaders to transform ideas into reality, drawing inspiration from transformative solutions found in nature. Our approach? Harnessing the latest scientific research with innovative tools to deliver solutions to complex challenges.
Reach out for a virtual coffee to discuss ideas.
