Deep Learning with Biomimicry
“Everything we love about civilization is a product of intelligence, so amplifying our human intelligence with artificial intelligence has the potential of helping civilization flourish like never before – as long as we manage to keep the technology beneficial.”
- Max Tegmark, President of the Future of Life Institute (Davey, et. al, 2018).
The universe is constantly seeking homeostasis: a state of equilibrium in nature that can also be said to apply to the dynamic fields of information technologies. The stock markets, foreign exchanges, and cryptocurrencies all reflect this search with their large, quantified sets of data reflecting ups and downs that are hallmarks of a system seeking balance. Creating super-intelligence in machines requires highly complex self-regulating systems, as the nature of our universe is to seek balance and stability. For example, an upward exponential trend, such as the one perceived by Singularists, assumes that advancement in artificial intelligence may be met with a steep decline of intelligent applications.
An analysis of the challenges facing the creation and execution of A.I. technologies shows a prejudice toward shaping new advancements after the human experience of intelligence. We propose the ongoing development of evolutionary intelligence be framed in terms of the natural and organic life cycles found in the world around us, utilizing botany and biomimicry as transdisciplinary tools. Many plant species naturally integrate the rise and fall of the planet’s search for balance with seasonal growth patterns.
As an observation, plants in spring gather rainwater, whereas in the summer the same plants manifest their collected energies into deeper, sturdier roots that focus on harvesting water, and leaves which produce pigments, such as chlorophyll, in order to harvest energy from the sun. By autumn, some plants wither and die (leaving behind seed pods or bulbs) as others go dormant, slowing metabolism and shedding leaves. As winter progresses, they lie in wait for the cycle to begin anew.
Self-sustaining, biomimetic technologies can be expected to survive the peaks and valleys of the technological search for homeostasis, where stability and durability are sought at cost to the advancement of artificial intelligence technologies. Evolutionary developments of machine intelligence can prevent the obsolescence of code designs, causing expansion of deeper rooted systems and software that can feasibly self-prune and purge to survive future transmutations.
Keywords: transdisciplinary, biomimicry, botany, plant life cycles, evolutionary developments, machine intelligence, self-regulation, homeostasis, intelligent applications, deep learning
Overview of Artificial Intelligence
The concept of artificial intelligence has, so far, been approached in a manner that strives to emulate, mimic, or otherwise replicate the nature of human intelligence. The overarching goal of artificial intelligence (A.I.) seems to converge its innumerable developments into the creation of software programs that can autonomously identify, solve, resolve, disclose, and optimize its impact of real-world problems as well as or better than humans. This premise has evolved as the prevalent cornerstone for technological advancements with concurrent academic studies and innovative development initiatives within industry areas such as cognitive neuroscience, deep learning, neural networking, machine intelligence, and natural language processing. However, the concept of human intelligence as the primitive construct for A.I. technologies can be perceived in contrast with the intelligence that encompasses the natural, biological environment which humankind inhabits.
The status quo of achieving the creation of an A.I. that simulates human intelligence, as the proposed construct for its primitive infrastructures, could contain (1) fundamental miscalculations by being inherently flawed by high-level design integrations, or (2) too casually narcissistic for practical or realistic expectations of evolutionary development. For example, even with theoretical scientific acknowledgements that hypothesize human consciousness having a strong correlation to the complex set of critical infrastructure dynamics in the human brain (Thomson, 2014), the achievement of establishing a real-world case of any single sense of self within a self-regulated A.I. program is unlikely. Creation of single sense of self would require a specific unified initiative, set forth by resourceful influencers, then deliberately crafted into curated sets of procedural, technical, or adaptive methodologies, that are deemed as highly proficient within their respective fields. While the definition of intelligence may logically encourage development efforts to structure A.I. after human intelligence, adopting and propelling the awareness of biomimicry methodologies into the evolutionary developments of A.I. may pose distinct advantages. Our abstract considers biomimicry with an emphasis on plant life cycles as a viable construct for developing evolutionary methodologies within the field of A.I. and its correspondent technological developments. Biomimicry is known as an applied transdisciplinary approach to innovation that seeks sustainable solutions to human challenges by emulating time-tested patterns and strategies modeled from complex infrastructures in the natural world. The study and application of biomimicry necessitates the adaptation of holistic approaches across many disciplinary boundaries at once within the context of evolutionary A.I. developments. This presents a unique opportunity to impact correspondent disciplines with other, relevant specialties such as deep learning, economics, cognitive neuroscience, biomedical engineering, space exploration, and other fields of study with potential to merge in contextual applications.
Human Intelligence and A.I.
By replicating the complex biological systems found in nature, A.I. capabilities could allow for better understanding of the cyclic patterns in nature and the organic functionality of intelligence itself. Neurobiologist Lu Chen, Ph.D., of Stanford University states, “We know very little about the brain. We know about connections, but we don’t know how information is processed.’’ (Lam, 2016) The challenge of structuring an advanced technology after a physiological structure that even industry leaders are still working on understanding is a difficult undertaking. An analogous concern is the “Doorway Effect” (Radvanksy, Krawietz, Tamplin, 2011), the experience of walking into a room and forgetting what one was originally intending to do. Such nuanced complexities of the human brain can illustrate the ways in which its core functionality can be trivially short-circuited; sometimes occurring by displacing focus temporarily onto the surrounding environment and losing the original “train of thought” as a result.
While it can be argued that encountered challenges can be tested and experimented upon with effort and time, such technological hurdles have a high statistical probability of failure due to the unanticipated complexities of the human brain. The mind’s lifelong processes of development, as well as abnormalities and anomalies that can be found en masse, further contribute to these acknowledged complexities. Additionally, the mind’s uncanny ability to overcomplicate simple tasks is no small hurdle to the creation and correlation of competent, useful, and sophisticated problem solving capabilities. A recent publication highlights the complications manifested in the A.I. developments that took place within Google Brain experiments. Left to their own devices, computers running machine intelligence and natural language processing- layered software created a language (Biggs, 2016). The language that the computers created has yet to be decrypted and made readable for humans to interpret as valid metrics for natural language processing within artificial intelligence systems. Further illustrating A.I. developments, 2018 saw the creation of Google Assistant (Chokkattu, 2018), a program that uniquely provides realtime data processing through technological aggregation techniques that have been layered over the span of the last decade. However, it is of paramount importance for open source communities, and their influential corporate counterparts, to recognize that such innovative technologies take time, testing, and experimentation to develop. Achieving any form of simulation of human intelligence will not only affect the global developer communities that collectively craft innovations, but could also sway technological innovations adopted en masse into evolutions that may be undesirable or dynamically inefficient in later years.
Fear & The Singularity
Over-complication of A.I. advancements has arguably alienated laypersons from attempting to understand its purpose and relevance in modern society. Due to this, it has also consequently prompted a societal fear response due to the lack of educational awareness, technological understanding, and general campaigns for comprehensive adoption by the technical community in large. In response, societal fear has arisen over a hypothetical event in which artificial super-intelligence enters a state of unregulated self-learning, leading or empowering its autonomy to transcend to a superior dimension in relation to human intelligence (Brockman, Kurzweil, 2001). The perceived event, known as the Singularity, has pervaded mass media and science fiction since the rise of technology in both public and private sectors. Fear of the unknown is a common psychopathology in humans (Bergland, 2016), as is fear of authority (Kashtan, 2012). The concept of a suddenly emerging technology that is not only difficult to understand in terms of mass adoption but also superior to the collective intelligence and longevity of humans is bound to instill fear in uninformed individuals. Unfortunately, this is quite the opposite effect that A.I. communities would hope for when developing open source technologies intended to empower humanity’s efforts toward the aspiration of achieving advanced methodologies characterized by artificial and machine intelligence technologies.
Fortunately for humankind, the Singularity is statistically unlikely (Allen, 2011). When confronted with possibilities of such growth, the world invariably finds ways to prevent it from continuing ad infinitum with a process known as homeostasis. This natural process of seeking balance within ecosystems can be observed not only in wildlife and plant populations, but also in the rise and fall of economies as exemplified in the housing market (micro) and world currency markets (macro). Increases in prices and market values can be observed as fluctuations on graphs and charts in the global financial trade markets, while tracked and assessed via automated indicators for the purpose of establishing economic strategies for fundamental and technical analysis (Technical Analysis, 2011). These indicators utilize recognized patterns and mathematical equations, such as the Fibonacci sequence, to give traders guidelines for decisions on buying or selling (Technical Analysis, 2011) Existence of these patterns and the frequency with which they prove useful are indicators that, while constantly fluctuating, there is a mean or median which the charts are balanced around, as can be viewed in Figure 2. The mean or median can be considered the point at which the charts themselves would become homeostatic, or balanced. Such dynamic data environments present the opportunity to utilize biomimetic methodologies to establish discipline within the scope of their societal impact.
The Rise of Complex Technologies
Over the past century, we have experienced a massive growth of technological evolutions due in large part to incentivized government programs, open source innovations, and market adaptations. This rapid growth has empowered humankind to advance from horse-and-carriage transportation to interplanetary outer-space exploration and commercialized autonomous vehicle transportation methods in under a century. Access to information has evolved from brick-and-mortar libraries to smartphone devices with the capacity of quad-core computers, stored in purses, back pockets, and even wallets. That technology has advanced in the past century is an understatement; it has become integral to society in ways no one could have guessed, thus it can naturally be expected that a great downturn will balance the charts in due time, and possibly at random. In anticipation of the realization of this looming downturn, it would behoove the tech sector to look toward computer learning models that take this into account and can bounce back more readily from an inevitable downfall. Enter the field of biomimicry.
Biomimicry as a concept is defined by industry leader Sue L. T. McGregor as “the juncture where ecology meets agriculture, medicine, manufacturing materials science, energy, computing and commerce” (McGregor, 2013). The concept behind biomimicry is that the natural world can and should influence a variety of systems, from social interactions to marketing and production. The push toward emulating systems that occur in nature is that it tends to build, shape, and grow within the confines of its natural resources without imposing linear shapes and functions to itself (McGregor, 2013). The ability to build and grow within a limited area without the need to fit itself into a box allows for optimization of available resources such as energy and space. Natural systems also allow for the co-evolution of complex organisms; the change in one organism will cause another organism to adapt to fit within its environment around and alongside the other organism. Trees in a forest, for example, require growth within a great deal of space with their leafy canopies, often leaving other, smaller plants to compete for the light they need in order to photosynthesize. To adapt to the loss of light, vines have evolved the ability to climb the trunks of trees and utilize space that otherwise would have been wasted as visualized in Figure 3 (Gartner, 1991).
The ability for organisms to cope, change, and adapt within an environment prone to continuous fluctuation poses as a significant component to establishing a technological environment which not only fosters complex growth cycles, but additionally prepares its own system structures for the eventuality of a cyclical downfall while simultaneously creating the ability to identify opportunities for regrowth, much like plants would demonstrate after a forest fire (Oliver, 2003).
An Overview of Plant Intelligence
Typically, plants are presented to humans as objects or accessories to the world they inhabit. Apart from a few exceptions, humans cannot perceive plant movement; plants do not physically appear to move much. They will only appear to be in motion when acted upon by outside forces, bending and stretching when initiated by the wind or other environmental activity. To the naked eye, there is not much complexity or benefit to their existence aside from consumption, being aesthetically pleasing, or practical uses such as being made into fabrics or ropes. However, when one observes closely, plant structure reveals unique, complex behavior patterning.
Plants possess a unique intelligence which allows them to move, grow, and even fend off predators within the proper corresponding environments. They have been recorded via timelapse footage moving and turning to allow them to absorb light optimally and will actively grow towards areas that afford them the best chance at survival (Valladares, Wright, Lasso, Kitajima, Pearcy, 2000). Studies show that certain species, when confronted with predators such as ants, will secrete pheromones to ward off the threat and communicate chemically to other plants so that they can better defend themselves by becoming bitter and inedible (Zagrobelny, Bak, Rasmussen, Jorgensen, Naumann, Moller, 2004). Plants possess a unique ability to cope with their environment when living conditions become less than favorable for their natural states. Deciduous trees will turn dormant in the winter to avoid the withering of their tender leaves in the harsh cold. Flowers react in a similar self-sustaining pattern, dropping seeds and leaving behind bulbs in order to propagate after winter’s last frost. Because of their continued ability to operate within a confined environment, and their ability to be uniquely integrated within that environment even under extreme conditions, the complex organic structure of plant physiology makes for an elegant template for applying biomimicry methodologies within innovations such as A.I. technologies (McGregor, 2013).
Bean Plant Specifics & Lifecycle Application
An excellent example for the purposes of creating systems for identifying complex intelligence mechanisms would be that of the bean plant. Bean plants are relatively simple to grow in gardens and are commonly studied in lower level elementary school science classes across the United States (Smith, Anderson, 2006). Since the bean plant’s biology is simple to understand, it makes an attractive subject by which A.I. can become more approachable to the public at large.
To grow a bean plant, one would place the desired bean within a damp paper towel and place that towel into a resealable plastic bag. After a few days in this enclosed environment, the bean will have begun a process of germination, as pictured in Figure 4. After an observable period of time in this germinating state, the bean would develop a primary root structure, known as the radicle, as well as the beginnings of a stem and a primary leaf structure, known as the cotyledon (Martin, n.d.). Once these two preliminary developments are in due process, the bean sprout is then placed into a container of soil and, allowing that there is sufficient water, nutrients and sun, the bean plant will grow and flourish. Bean plants root and grow quickly, with mature leaves formed as soon as a week after planting, and will begin “searching” their surrounding landscapes with tendrils from their stem for places in their environment to anchor themselves. This particular extension of development will allow them to continue growing in an upward direction to gather more sunlight for energy consumption even after their leaves become too heavy for their narrow stem to support. After six to eight weeks, the plant will begin flowering and, given the flower is pollinated, a corresponding seed pod will develop (Martin, n.d.). Within these seed pods, new beans will form and eventually mature into future iterations of their initial bean plant system.
Smart Plant Innovations in Relation to Marcum Bean Theory
We will term the correlation of bean plant growth cycles and A.I. technologies as Marcum Bean Theory. The theory serves as the fundamental structure on which a dynamic, adaptable form of contextual deep learning models can be developed within biomimetic methodologies. Utilizing the premise of Marcum Bean Theory, influencer and open source community initiatives are presented with the opportunity to simplify the architectural designs of complex, evolution-oriented projects such as the development of human-inspired learning programs.
The Marcum Bean Theory emphasizes a guided methodology for establishing a dynamic infrastructure comprised of four primary modular components to achieve a “smart plant” construct: seed, root, stem, and leaf. The seed module functions as the basic beginning of the program’s task, requiring context and instruction as to what the end task to be performed will be as well as any information vital for the program to complete this task. From this established task, the root module structure will begin to form, sifting through and compiling information to siphon to the stem module. Much like a plant seeking water, areas of the root structure where there is no relevant information will cease growing and become dormant.
Having received vital information from the root module, the stem module will begin assimilating the required data to the associated leaf module structures. Like a grasping bean vine, the stem module grasping for stability will then begin searching for new information using the assimilated data to come up with a solution to the task given to the program during the seed module stage. Based on human assessment, any identifiable areas that do not result in a plausible or viable output solution will cease to grow, and will “self prune” or cut themselves off, as they are no longer useful and do not require the data being assimilated by the root structure. This self pruning activity will also assist in limiting the usage of resources by the Smart Plant, allowing the root module to become more efficient in its data aggregation functionalities.
Any plausible solutions discovered by the construct will present as a leaf module, and the information in the leaf structures can be utilized to collect the best possible optimized solution to the task that was presented during the seed module stage. In the event that a particular plant module becomes incapable of continuing to function or is consuming too many resources to be viable, it will then enter a “flowering” stage wherein a parent plant creates a new seed module. The new seed includes not only the primary origin task and basic functionality of the Smart Plant construct, but also any useful information that the smart plant construct has encountered during its own growth stages within its construct life cycle.
Smart Plant Technology
Maintaining information that has proved useful to the parent plant is a key component of Lamarckian evolution, a now defunct theory which proposes that evolutionary traits are passed quickly from parent to child through physical changes that are made to the parent during its lifetime (Burkhardt, 2013). Giraffes, for example, would continuously stretch their necks to reach leaves on high trees, and upon doing so a “nervous fluid” would flow into the neck allowing for the neck to become longer during the animal’s own lifetime. This physical change would then pass onto the giraffe’s offspring and the cycle would continue, creating elongated necks over time. While we now understand that evolution has more to do with natural selection than nervous fluids, this principle is one that can readily be used in Smart Plant technology in order to continue the paths of data acquisition from parent to seed. The passing of acquired knowledge would also allow information about the data environment to be passed on to the developing plant. Data environment information would prove useful to the seedling, allowing for greater efficiency in future data compilation and assimilation activities.
Marcum Bean Theory and Smart Plant technology constructs demonstrate the potential to serve as a fundamental basis for numerous undertakings in evolutionary technologies that currently utilize human research and application as their primary resource for innovation. For example, the plant’s ability to compile and assimilate available information autonomously while enabling flexibility with its initial programming constructs will allow for a revolutionized approach to medical research. By siphoning and sifting through information on ready-made medicines, chemicals, and medicinal herbs, this methodology could be utilized to synthesize potential treatments for illnesses and diseases while making more time available to current medical researchers to do tasks that would benefit from their abilities as humans. Smart Plant technology constructs also present potential utilizations in the fields of mathematics, civil engineering, and language processing. In particular, its language processing potential may open doors for the improvement of language softwares that are intended to bridge the gap between language barriers. The application and context of the initial programming will ultimately determine the Smart Plant’s utilization, making it highly adaptive and potentially useful for a wide array of use scenarios.
Critical Observations to Shape Future Innovations
While current A.I. developments may be unnecessarily focused on the replication of human-based intelligences, this newly introduced theoretical technology has the potential to shift the paradigm in a direction that works in a more streamlined, adaptive way. The form and function that academics, economists, and scientists have previously taken to enhance standard methodologies applied within A.I. development, such as Agile software lifecycle management practices, which adopt incremental and iterative methods via structured frameworks for building software programs (Maximini, 2015), will work in unison with Smart Plant technology in order to enable faster rates of production. With increased production, the continued experimental application of real-world use cases present the potential to realize a universal set of best practices for A.I. development methodologies that can become not only more organic, but also much more efficient in its use of resources.
The applied approach of biomimetics within organizations that influence commercial and open source awareness, adoption, and evolution would lend significant influence to the credibilities of scalable tracking, emulating, simulating, or replicating learning models not only in cases of achieving human consciousness (Hobson, Pace-Schott, 2002), but also other seemingly random orders of chaos within the universe. There are currently multibillion- dollar businesses at work in commercial and open source innovations, such as Google, IBM, and Intel, which have pioneered and continue to model the constructs of what are now considered modern software engineering principles. These principles define the scope of impact for transdisciplinary developer communities among projects that perform multi-trillions of data processing requests in the capacity of micro-services and web services for a broad percentage of the global population as a segment of their respective business models.
While the continual evolution of open source developments opens doors for applied iterations of innovative projects such as aforementioned autonomous vehicle transportation or Google Assistant™ publications, placing emphasis on defining the societal adoption of evolution-oriented methodologies could shape the future of synthetic intelligence. An analysis of the challenges facing the creation and execution of A.I. technologies shows a prejudice toward shaping new advancements. We propose the ongoing development of evolutionary intelligence be framed in terms of the natural and organic life cycles found in the world around us, primarily utilizing botany and biomimicry as transdisciplinary tools for deep learning models and constructs. It is clear that cyclical patterns in nature occur across disciplines in apparent and random fashion; this is a universal truth in fields such as finance, sciences, software engineering, mathematics, and applied technologies. With belief in the transdisciplinary adaptation of technological evolutions illustrated by biomimetics, developments in artificial intelligence and related disciplines can create an influential reform in the way we collectively choose to innovate as a society of interconnected communities.
Allen, P. G. (2016, January 08). Paul Allen: The Singularity Isn't Near. Retrieved from https://www.technologyreview.com/s/425733/paul-allen-the-singularity-isnt-near/
Bergland, C. (2016, November 21). Study: Fear of the Unknown Compounds Many Anxiety Disorders. Retrieved from https://www.psychologytoday.com/us/blog/the-athletes-way/201611/study-fear-the-unknown-compounds-many-anxiety-disorders
Biggs, J. (2016, October 28). Google's AI creates its own inhuman encryption. Retrieved from https://techcrunch.com/2016/10/28/googles-ai-creates-its-own-inhuman-encryption/
Burkhardt, R. W. (2013, August). Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3730912/
Chokkattu, J. (2018, April 11). Google Assistant has arrived on the Pixel C. Retrieved from https://www.digitaltrends.com/mobile/google-assistant/
Davey, T., Conn, A., Perry, L., Cussins, J., Bensinger, R., & Gronlund, K. (2018, May 24). Home. Retrieved from https://futureoflife.org/
Gartner, B. L. (1991, December 01). Structural Stability and Architecture of Vines vs. Shrubs of Poison Oak, Toxicodendron Diversilobum. Retrieved from https://esajournals.onlinelibrary.wiley.com/doi/abs/10.2307/1941555
Hobson, J. A., & Pace-Schott, E. F. (2002, September 01). The Cognitive Neuroscience of Sleep: Neuronal Systems, Consciousness and Learning. Retrieved from https://www.nature.com/articles/nrn915
Kashtan, M. (2012, August 16). Our Habitual Responses to Authority. Retrieved from https://www.psychologytoday.com/us/blog/acquired-spontaneity/201208/our-habitual-responses-authority
Kurzweil, R., & Brockman, J. (n.d.). THE SINGULARITY. Retrieved from https://www.edge.org/conversation/ray_kurzweil-the-singularity
Lam, V. (2016). "We know very little about the brain": Experts Outline Challenges in Neuroscience. Stanford Medicine. doi:10.3897/bdj.4.e7720.figure2f
Martin, A. (n.d.). Life Cycle Bean Plant. Retrieved from https://garden.lovetoknow.com/garden-basics/life-cycle-bean-plant
Maximini, D. (2018). SCRUM CULTURE: Introducing agile methods in organizations. S.l.: SPRINGER INTERNATIONAL PU.
Mcgregor, S. L. (2013). Transdisciplinarity and Biomimicry. Transdisciplinary Journal of Engineering & Science,4(1). doi:10.22545/2013/00042
Oliver, C. D. (2003, June 25). Forest development in North America following major disturbances. Retrieved from https://www.sciencedirect.com/science/article/pii/0378112780900134
Radvansky, G. A., Krawietz, S. A., & Tamplin, A. K. (2011). Walking through Doorways Causes Forgetting: Further Explorations. Quarterly Journal of Experimental Psychology,64(8), 1632-1645. doi:10.1080/17470218.2011.571267
Smith, E. L., & Anderson, C. W. (2006, August 18). Plants as producers: A case study of elementary science teaching. Retrieved from https://onlinelibrary.wiley.com/doi/abs/10.1002/tea.3660210703
Technical Analysis[.pdf]. (2011, February 2). Https://www.mrao.cam.ac.uk/~mph/Technical_Analysis.pdf.
Thomson, H. (2014, July 2). Consciousness on-off switch discovered deep in brain. Retrieved from https://www.newscientist.com/article/mg22329762-700-consciousness-on-off-switch-discovered-deep-in-brain/
Valladares, F., Wright, S. J., Lasso, E., Kitajima, K., & Pearcy, R. W. (2000, July 01). Plastic Phenotypic Response to Light of 16 Congeneric Shrubs from a Panamanian Rainforest. Retrieved from https://esajournals.onlinelibrary.wiley.com/doi/full/10.1890/0012-9658(2000)081[1925:PPRTLO]2.0.CO;2
Zagrobelny, M., Bak, S., Rasmussen, A. V., Jorgensen, B., Naumann, C. M., & Moller, B. L. (2004, January 06). Cyanogenic glucosides and plant–insect interactions. Retrieved from https://www.sciencedirect.com/science/article/abs/pii/S003194220300637X