Summary
This video explores the concept of intelligence and cognition beyond biological brains, suggesting it's a fundamental property of reality. It introduces synthetic beings like Xenobots and Anthrobots, which exhibit emergent behaviors and cognitive processes such as sensing, memory, and goal-directedness. The discussion extends to the idea that even simple chemical systems and gene regulatory networks can display learning and problem-solving capabilities. It proposes that life itself is a means of scaling cognitive light cones, the scope of an agent's goals in space and time. The implications for understanding consciousness, evolution, and the future of humanity in an era of AI and biotechnology are profound, challenging our definitions of life and agency.
Key Insights
Claim: Intelligence and cognitive capacity cannot be assumed; they must be discovered through experimentation.
A core principle driving these discoveries is the inability to assume a system's level of intelligence or cognitive capacity. Instead, these must be determined through experimental observation. This challenges traditional assumptions about where intelligence resides.
Gene regulatory networks can exhibit learning via conditioning, without needing cellular machinery.
A simple gene regulatory network, a system of genes and proteins regulating each other, can be subjected to behavioral training assays. By using nodes in the network as unconditioned and neutral stimuli and applying Pavlovian conditioning, the network can learn to associate the neutral stimulus with a response, demonstrating learning without complex biological structures.
Claim: Biological systems possess goals and use encoded future states or predictions as set points.
Cells and tissues exhibit goal-seeking behavior, indicated by their cessation of activity upon reaching a goal. This implies they possess a stored memory or prediction of the desired future state, which acts as a set point. This mental time travel or anticipation is essential for goal-directed behavior.
Claim: Biology's fundamental strategy is exploiting hierarchical architecture to cooperate and solve problems in changing environments, rather than relying on stable hardware.
Unlike engineered technology where the bottom layer is stable, biology is characterized by constant change (environment and internal mutations). Evolution exploits a hierarchical architecture, focusing on policies for cooperation between networks, cells, and tissues that enable problem-solving and goal pursuit in the current moment, without assuming past stability.
Claim: Cognition is a broader category than life, suggesting cognition predates and enables life's scaling capabilities.
Cognition is considered a more fundamental and expansive category than life. Life is defined as evolutionary systems adept at scaling their 'cognitive light cone' (the scope of goals in space and time). Therefore, cognition existed before life, and life evolved as a means to scale it up.
Claim: All intelligence observed is collective intelligence, arising from parts joining and changing together.
Intelligence is fundamentally a collective phenomenon. There are no known examples of monolithic, indivisible intelligence; rather, all observed intelligence involves parts forming together and changing. This collective nature is essential for learning and complex behavior.
Claim: Anatomical goals ('attractors') exist in Platonic space, guiding cellular agents in morphogenesis.
Systems biologists posit a Platonic space where anatomical goals, termed 'attractors', exist. Cellular collectives, acting as cognitive agents, are drawn to these attractors, collaborating and problem-solving to manifest physical forms that cannot be explained solely by present physical forces.
Claim: Biological systems are fundamentally adaptable and reconfigurable, not rigidly determined by DNA or past environments.
The process of morphogenesis (forming bodies and minds) possesses agency and intelligence. This is evidenced by applying cognitive and behavioral science tools, which yield better scientific results than assuming systems are mere dumb mechanical machines. Biological systems have plasticity beyond fixed blueprints, allowing them to find new ways to reach goals or even new goals.
Claim: The principle of least action in physics represents a minimal form of goal-directed behavior.
The principle of least action, fundamental in physics, describes how systems naturally follow paths that minimize action. This can be interpreted as a minimal form of goal-directed behavior or optimization, mathematically describable using the same framework as active inference in brains.
Sections
Introduction to Synthetic Life and Emergent Behavior
Hybrids, cyborgs, and chimeras represent combinations of living and engineered materials with diverse potential bodies and minds.
The speaker frames the future as one where humans will live with hybrid and chimera entities, which are combinations of living material and designed engineered material. The option space for these diverse bodies and minds is enormous. It's noted that even simple code can lead to unpredictable emergent behaviors, implying a similar unpredictability in more complex artificial and biological systems.
Synthetic beings like Xenobots and Anthrobots demonstrate novel behaviors not present in their origin tissues.
Michael Levan explains the creation of synthetic beings such as xenobots (made from frog skin cells) and anthrobots (made from human tracheal epithelial cells). These entities exhibit interesting new behaviors that their original tissues did not possess. They self-assemble into functional forms with novel behaviors and structures that were not explicitly designed.
AI is used to discover new ways to prompt cellular collectives into novel behaviors and forms.
AI, in collaboration with labs like Josh Bongard's, is used as a tool to discover how to prompt cells to assemble and behave in new ways. This AI acts as a translator to communicate goals to the cellular collective, which comprises groups of cells acting as a scalable intelligence team.
Cellular collectives exhibit scalable intelligence, problem-solving, and communication to achieve shared goals.
Groups of cells within the body, termed 'cellular collectives', are described as scalable intelligence teams whose problem-solving power increases with size. They utilize shared goals and communication to solve problems, adapt, and accomplish tasks, demonstrating collective intelligence beyond individual cells.
Claim: Intelligence and cognitive capacity cannot be assumed; they must be discovered through experimentation.
A core principle driving these discoveries is the inability to assume a system's level of intelligence or cognitive capacity. Instead, these must be determined through experimental observation. This challenges traditional assumptions about where intelligence resides.
Cognitive processes like sensing, memory, learning, decision-making, and goal-setting are observed in non-biological and simple biological systems.
The experiments reveal cognitive processes such as sensing, perception, memory, various kinds of learning, decision-making, goal-setting, and pursuing goals, including creative problem-solving. These are observed in unconventional beings and systems, not just brains.
Intelligence is defined as the ability to use available tools in new ways to solve unfamiliar problems.
A definition of intelligence is presented as the capacity to utilize existing tools in novel ways to address problems never before encountered. This definition is deemed non-controversial and applicable to diverse systems.
Tools from behavioral and cognitive sciences can be applied to study cells, tissues, and molecular networks.
The functional reality of these cognitive processes in unconventional systems is validated by the ability to apply the same tools used in behavioral and cognitive sciences (studying memory, learning, goal-directedness, problem-solving) to cells, tissues, and molecular networks.
Learning and Memory in Molecular and Cellular Systems
Gene regulatory networks can exhibit learning via conditioning, without needing cellular machinery.
A simple gene regulatory network, a system of genes and proteins regulating each other, can be subjected to behavioral training assays. By using nodes in the network as unconditioned and neutral stimuli and applying Pavlovian conditioning, the network can learn to associate the neutral stimulus with a response, demonstrating learning without complex biological structures.
Basic chemical systems demonstrate various forms of learning prior to complex biological organization.
It's shown that even basic chemical systems, before reaching the complexity of biological cells where life and intelligence typically emerge, are capable of various categories of learning, including sensitization, habituation, and associative learning.
Intelligence emerges at a deeper level and in more varied systems than traditionally understood.
The natural world possesses varying degrees of cognitive competencies, opening new ways of relating to reality where intelligence emerges at a fundamental and diverse level. Examples include chemical droplets solving mazes.
Understanding cellular intelligence allows for new approaches to drug discovery and treatment of drug habituation.
If molecular networks possess memory, as demonstrated by learning, this has implications for biomedicine. Drug habituation, where drugs stop working over time, can be understood as a form of memory. This understanding could lead to new strategies for drug delivery, such as pulsing exposures, to create or break these memories.
Claim: Biological systems possess goals and use encoded future states or predictions as set points.
Cells and tissues exhibit goal-seeking behavior, indicated by their cessation of activity upon reaching a goal. This implies they possess a stored memory or prediction of the desired future state, which acts as a set point. This mental time travel or anticipation is essential for goal-directed behavior.
The brain's memory of an anticipated endpoint can be represented by distributed bioelectrical states.
In brains, the encoding of anticipated endpoints for goals is likely patterns of electrical activity. Experiments show that in organisms like planarians, a distributed bioelectrical state can be decoded to represent the blueprint for building a specific structure, like a head.
Altering the bioelectrical state representing a body plan can directly change the resulting physical form.
By altering the bioelectrical state in planaria that dictates the number of heads to form, researchers successfully induced the development of two-headed worms, demonstrating that the physical form can be changed by modifying the stored 'memory' of the correct body plan without altering DNA.
The ability to develop functional organs in unexpected locations suggests inherent reconfigurable capacities.
Animals can develop functional eyes on their tails, which connect to various systems like the spinal cord or gut. This demonstrates that biological systems are fundamentally ready to reconfigure themselves in novel scenarios without requiring new rounds of selection and mutation.
Bodies operate as nested modular agents with a multiscale competency architecture that delegates tasks.
Bodies are composed of nested modular agents with a multiscale competency architecture. These components solve problems by delegating tasks to downstream modules, without micromanagement of individual cells or genes. Triggering competencies in a cellular collective can lead to complex structures like eyes, with adjustments made based on location.
Claim: Biology's fundamental strategy is exploiting hierarchical architecture to cooperate and solve problems in changing environments, rather than relying on stable hardware.
Unlike engineered technology where the bottom layer is stable, biology is characterized by constant change (environment and internal mutations). Evolution exploits a hierarchical architecture, focusing on policies for cooperation between networks, cells, and tissues that enable problem-solving and goal pursuit in the current moment, without assuming past stability.
Claim: Cognition is a broader category than life, suggesting cognition predates and enables life's scaling capabilities.
Cognition is considered a more fundamental and expansive category than life. Life is defined as evolutionary systems adept at scaling their 'cognitive light cone' (the scope of goals in space and time). Therefore, cognition existed before life, and life evolved as a means to scale it up.
A 'cognitive light cone' defines the boundary of events in space and time an agent can perceive and act upon.
Borrowing from physics, a cognitive light cone represents the scope of awareness for an intelligent system. It encompasses the limits of events in time and space that an organism can perceive, remember, or leverage to guide its actions, shaping its decision-making.
Cancer can be viewed as a breakdown in communication, causing cells to shrink their cognitive light cone and devolve.
Cancer is described as a biophysical disconnection from the cellular collective, causing the cell's cognitive light cone to shrink to an amoeba-like level. Freed from collective goals, the cell focuses solely on local optimization, treating the rest of the body as external.
Cells possess individual memories and agendas, enabling collective intelligence but also susceptibility to 'defections'.
While cellular collectives store memories of what they are building, individual cells have their own abilities to learn, agendas, and goals. Managing this diversity allows for incredible intelligence but also creates risks of parts deviating from the collective's objectives.
Consciousness, Awareness, and Agency in Systems
Awareness is defined as any information accessible to a system to drive its behavior.
A useful definition of awareness, beyond human-level meta-awareness, is any information that a system has access to and can use to drive its behavior. This makes information access the core of awareness.
Cells exhibit awareness, predictive capacity, and morphological problem-solving capabilities.
Individual cells possess inherent cognitive capabilities, including anticipation of future events, learning, and morphological problem-solving analogous to cellular collectives. This suggests cells have predictive capacities and are not merely passive responders.
Gene regulatory networks show increasing causal emergence as they learn, indicating reality beyond their parts.
Even gene regulatory networks, before considering the whole cell, exhibit increasing causal emergence metrics as they learn. This signifies a reality or organizational property that transcends the sum of their individual components, relevant to theories of consciousness like Integrated Information Theory.
Claim: All intelligence observed is collective intelligence, arising from parts joining and changing together.
Intelligence is fundamentally a collective phenomenon. There are no known examples of monolithic, indivisible intelligence; rather, all observed intelligence involves parts forming together and changing. This collective nature is essential for learning and complex behavior.
Platonic Space, Evolution, and Biological Forms
Evolution may search 'pointers into Platonic space' rather than the space of all possible phenotypes.
An alternative view of evolution suggests it searches through 'pointers into Platonic space'—abstract realms of mathematical laws and forms—instead of merely searching all possible physical outcomes or phenotypes.
Claim: Abstract truths like mathematics have an objective, unchanging reality in Platonic space.
Abstract truths, such as mathematical principles, exist objectively, perfectly, and unchangingly in a Platonic realm. These truths, like the number four, are not contained within physical instances but apply universally, suggesting a non-physical reality that structures the physical world.
Claim: Anatomical goals ('attractors') exist in Platonic space, guiding cellular agents in morphogenesis.
Systems biologists posit a Platonic space where anatomical goals, termed 'attractors', exist. Cellular collectives, acting as cognitive agents, are drawn to these attractors, collaborating and problem-solving to manifest physical forms that cannot be explained solely by present physical forces.
Navigating anatomical space involves moving from an initial state to a region corresponding to the correct shape.
Embryogenesis is described as navigating a high-dimensional anatomical space. Starting as a single cell, the organism moves through this space, eventually reaching the region representing its correct shape. Interference can redirect this path, but the organism still strives to find its way to the intended form.
The ability to form complex structures reliably suggests a platonic realm of anatomical goals guiding development.
The reliable construction of complex, functional biological structures like eyes and limbs, even when cells are scrambled or damaged, cannot be fully explained by DNA alone. This suggests that cells are drawn towards abstract anatomical goals (attractors) in a Platonic space.
Biology's intelligence arises from exploiting hierarchical cooperation to solve problems in changing conditions.
Evolution doesn't just find specific solutions, but develops policies for cooperation across molecular networks, cells, and tissues. These policies enable systems to solve problems and pursue goals in the current moment, adapting to constant environmental and internal changes without relying on fixed 'hardware'.
Claim: Biological systems are fundamentally adaptable and reconfigurable, not rigidly determined by DNA or past environments.
The process of morphogenesis (forming bodies and minds) possesses agency and intelligence. This is evidenced by applying cognitive and behavioral science tools, which yield better scientific results than assuming systems are mere dumb mechanical machines. Biological systems have plasticity beyond fixed blueprints, allowing them to find new ways to reach goals or even new goals.
Our perception is biased towards 3D space, hindering recognition of intelligence in physiological or gene expression spaces.
Human cognitive systems prioritize 3D physical space due to evolutionary optimization for movement. This bias makes it difficult to visualize and recognize intelligence in other 'spaces' like blood chemistry or gene expression, where organs act as intelligent agents, solving problems to maintain life.
'Symbiants' are competent entities, like organs, that navigate specific spaces to keep a host alive.
A symbiant is defined as a species living helpfully within another. Organs can be viewed as highly competent symbiants tasked with navigating specific physiological spaces to maintain the host's life and health by addressing various challenges.
Organ transplant studies suggest organs may store memories or personality traits, challenging psychosomatic dismissals.
Studies on organ transplant recipients report personality changes irrespective of the organ transplanted. Researchers are investigating whether donor organs can store memories or traits, moving beyond previous dismissals of such anecdotal reports as psychosomatic.
Cellular cognition and memory can evolve through bioelectrical circuits, not solely through DNA.
The evolution of traits can involve bioelectrical circuits as a 'scratchpad' for information, in addition to DNA. The process of translating genes into outcomes is itself an intelligent, problem-solving competency, not a purely mechanical mapping, allowing for greater plasticity and evolution.
Humanity can become more intentional about shaping its future, moving beyond random evolutionary pressures.
Unlike past biological evolution shaped by random pressures like cosmic rays, humanity now has the potential to intentionally shape its future, both personally and societally. This involves reinterpreting past stories and consciously steering future development, aiming for better outcomes than random chance.
Hybrids, Chimeras, Cyborgs, and the Future of Humanity
Hybrids and chimeras combine biological and non-biological materials, blurring lines between life and machines.
Hybrids and chimeras involve mixing species (e.g., frog and axolotl cells) or combining biological tissue with mechanical parts. Cyborgs are humans with implants. These combinations challenge the hard categories we use to define living things and machines.
The future will challenge our respect and compassion based on the proportion of biological versus mechanical components we possess.
As technology advances, we may face societal issues where compassion and respect are influenced by the degree of mechanical versus living material in individuals. The speaker argues against basing legislation and behavior on such distinctions.
Our limited perception of other minds is not necessarily an adaptation but a lack of pressure to perceive them.
The inability to perceive other minds widely might not be an evolutionary adaptation but rather a consequence of lacking pressure to do so historically. However, with AI and biotechnology, understanding these other intelligences is becoming crucial.
The speaker identifies as a type of 'pansychist', viewing mind/cognition as a fundamental feature of reality.
The speaker expresses a view aligned with pansychism, suggesting that mind or cognition is a fundamental aspect present throughout reality, not limited to complex biological brains. This is based on the observation of goal-setting and behavior across various systems.
Physics, using low-agency tools, primarily finds low-agency phenomena, thus not typically discussing minds or agency.
The tools used in physics (rulers, voltmeters) are low-agency, meaning they are not designed to detect or interact with complex agency. Consequently, physics tends to find and describe low-agency phenomena, making it unsurprising that agency and minds are not common topics in the field.
Claim: The principle of least action in physics represents a minimal form of goal-directed behavior.
The principle of least action, fundamental in physics, describes how systems naturally follow paths that minimize action. This can be interpreted as a minimal form of goal-directed behavior or optimization, mathematically describable using the same framework as active inference in brains.
Ask a Question
*Uses 1 Wisdom coin from your coin balance
