Neuroscience and AI: Understanding the Brain

• Artificial intelligence (AI) is the mechanical reproduction of intelligent behavior
• Researchers have historically sought inspiration from natural systems that exhibit traits AI tries to emulate
• Human and animal brains and their cognitive functions are the only known working examples of such systems
• The goal of this unit is to familiarize you with the basic tenets of neuroscience and cognitive science

Neuroscience and the Human Brain

• Neuroscience identifies the anatomical structures that form nervous systems and their functions.
• Neuroscience combines anatomy, physiology, cytology, and the chemical and developmental sciences.
• The focus is primarily on the human brain as it constitutes the most complex brain specimen.
• The brain is a lump of soft tissue that weighs between 1.2-1.4 kg in adults

Brain Anatomy and Physiology

• There is no evidence to suggest that brain size is connected with mental capacity
• The outer layer of the brain is a highly wrinkled structure with a large surface area
• The constituent parts are the cerebrum, the cerebellum, and the brain stem
• The brain stem functions as an interface between the brain and the spinal cord
• The brain stem controls basic functions like heart-rate regulation, breathing, body temperature, and the wake-sleep cycle
• The cerebellum is adjacent to the brain stem underneath the cerebrum
• The Cerebellum's main function is motor control, like steering movement, upholding balance, and maintaining body posture
• The largest fraction of the brain is constituted by the cerebrum
• All higher functions, such as the interpretation of sensory input, emotions, reasoning, speech and language understanding, reside in the cerebrum
• The cerebrum is split into two halves called hemispheres
• The corpus callosum connects the hemispheres, enabling the exchange of signals

Brain Hemispheres

• The right hemisphere controls the left side of the body, and vice versa
• The brain halves are generally symmetrical with respect to their function
• Functional specialization, also called lateralization, makes claims of high-level cognitive functions
• Reliable evidence for actual lateralization pertains to more low-level perceptual functions.
• Broca's and Wernicke's areas play an important role in language processing
• These brain regions are usually found in the hemisphere opposite to the dominant hand
• The brain hemisphere controls the dominant side of the body

Brain Lobes

• The brain can be compartmentalized into four main lobes: Frontal, Parietal, Occipital, and Temporal
• This division is based anatomically on the most visibly distinct fissures of the brain surface
• The vast majority of observable brain functions are based on the complex interaction of many of the brain's constituent parts
• The frontal lobe is responsible for higher mental faculties such as judgement, planning, problem-solving, intelligence, and self-awareness as well as complex motor control tasks and speech
• The parietal lobe contrasts with the frontal lobe and is mostly concerned with the interpretation of sensory input; spatial-visual perception and interprets vision, auditory, and touch
• The temporal lobe is involved in the understanding of language, the formation of memory, and sequencing and organization as well as complex vision tasks, such as the recognition of objects and faces
• The role of the occipital lobe lies in performing the early stages of visual signal processing and interpretation

Brain Cells

• The human brain is, on average, composed of about 86 billion connected nerve cells called neurons
• Neurons are responsible for information processing
• The brain has an approximately tenfold higher number of glia cells
• Glia cells are responsible for the protection, nourishment, and structural support of neurons.
• The cell body is called the soma
• Signals from other neurons reach the soma via branched structures called dendrites
• The soma processes incoming information and produces a corresponding output that is sent down the axon
• The length of the axon can be 10 to 1000 times the diameter of the soma
• The axon branches off into axon terminals at its end
• Axon terminals constitute the points of contact for dendrites from neurons further down the signal flow
• The difference between soma and axons manifests itself in the distinction between gray and white matter seen in brain cross-sections
• Gray matter is formed by the cell bodies, and white matter is comprised of axons

Brain Functions

• The human brain functions as the regulator of all bodily functions 24/7
• The interconnected network of neurons controls routine needs such as breathing, blood pressure, and mobility
• Communication between the brain and the body takes place along the spinal column.
• The brain is responsible for processing sensory input such as Vision (sight), Audition (hearing), Gustation (taste), Olfaction (smell), Tactition (touch), Thermoception (temperature), Nociception (pain), Equilibrioception (balance), Proprioception (body awareness)
• The brain is responsible for the promotion of behaviors that are beneficial for the organism Including attention, learning, memory, planning, problem-solving, understanding language, and the ability to form complex ideas
• The brain: perceives inputs, processes the input, and initiates action

Cognitive Science

• Cognitive science examines cognition and cognitive processes in their own right
• Cognitive science elucidates the corresponding functional relationships by abstracting from biological actualities
• Cognitive science addresses evolutionary and developmental aspects
• Typical cognitive processes studied are behavior, intelligence, language, memory, perception, emotion, reasoning, learning
• Defining characteristics of cognitive science: an interdisciplinary approach, like philosophy, psychology, neuroscience, linguistics, anthropology, and artificial intelligence
• The intellectual history of cognitive science can be traced back to the dawn of philosophy
• Current approaches derive from twentieth century developments like George Miller's studies on mental representations and the limitations of short-term memory, Noam Chomsky's work on formal grammars and his critique of radical behaviorism, early efforts in AI
• The term "cognitive science" was coined in 1975

Empirical Findings

• Empirical data in the study of cognitive processes is commonly derived from experimental methods
• Brain imaging: Commonly used to trace neural activity during complex mental tasks
• Behavioral experiments: behavioral experiments allow drawing conclusions about the processing of stimuli
• Simulation via computational modeling: This technique allows to verify theoretical ideas by comparing simulated outcomes with behavioral data

Key Concepts, Influences, and Critique

• The representational theory of mind is the prevalent paradigm
• Cognition is achieved by employing computational procedures on mental constructs that can be likened to data structures in computer science
• These data structures can represent concrete objects or abstractions that pertain to the mental domain
• Computational procedures are correspondingly variegated and include deduction, search and matching, and the like
• Cognitive science also has contributions from numerous other specialized disciplines
• Cognitive science contributes to: behavioral economics, the study of cognitive biases, and understanding the roles and interplay of brain structures.
• Areas cognitive science has only recently considered the role of emotion in human thinking and the problem of consciousness

Neuroscience, Cognitive Science, and Artificial Neural Networks

• The relationship between our brain and mental activities still requires research
• The brain is the physical base of mental states
• The human mind has the ability to regulate the flow of energy and information
• "Faculty" refers to all the mental and physical abilities a person is endowed with, which can exhibit considerable variation between individuals
• "Self-organizing" refers to a process of spontaneous ordering arising from local interactions
• "Emerging” means “to give rise to”
• "Relational processes" signify that there is a significant relationship between the human persona and outside objects and processes, especially other minds
• Specific human faculties Conscience is a faculty is that judges the difference between right and wrong
• Self-awareness: Conscious awareness and introspection
• Judgement: Ability to consider evidence and knowledge
• Language: The ability to use languages to express ideas
• Imagination: Ability to see possibilities beyond what is immediately being perceived
• Memory: Ability to recall coded and stored information in the brain
• Thinking: Faculty to search for possible reasons or causes

Reproducing Mental Faculties in Computational Machinery

• Computer scientists, and in particular artificial intelligence researchers, have used philosophical, psychological, and neurological models of cognitive capabilities as inspiration
• Researchers built and employ computation models to gain insight on the functioning of the mind or neural circuitry found in organisms
• A neuron's function: the cell receives electrochemical signals from neurons, modulates the input depending on how often two nerve cells are activated together, and then modulates what it knows already before sending the signal; if the total excitation exceeds some predefined threshold, the cell then takes the sum of all its inputs weighted in this manner and sends an impulse along the axon, its outgoing connection
• Process: Input is received via connections that model connection strength via weight parameters, the weights are added up, the result is subjected to an activation function

Network Flows and Structures

• Information processing and the learning of input-output associations is possible with a single computational unit
• The analogy to biological neural systems is carried one step further with layered schemes
• Concerning the flow of information processing through the network, there are feed forward and recurrent approaches
• The feed forward approach: Information only proceeds in one direction-upstream to downstream
• Every node in the network receives inputs, processes, and passes the signal onto the connected neurons in the next layer without looping
• Three types of layers in a feed forward approach: Input layer, Hidden Layer,Output layer
• The recurrent approach: The order of nodes encodes the processing steps
• Recurrent networks are suited to applications that have a time component, like the processing of a time-series, speech, or handwriting recognition
• Approaches to the creation of learning systems have been inspired by theories of neural information processing, but sometimes over-emphasized

Machine Learning

• Artificial neural networks implement simplified models of neural activity that abstract away many complexities of biological activity
• Deep Learning: emphasizes in name depth of layering over vague allusions to the functioning of biological neural networks.

Computer Structure

• Complex mobile chip designs have transistor counts in the order of a magnitude of 10¹⁰
• The number of transistors in a modern CPU already approaches the number of neurons in the human brain
• The largest artificial nets have unit counts of between 10⁶ and 10⁷
• This number has been doubling roughly every 2.4 years since the 1980s
• For reference, the number of neurons in humans is about 10¹¹, in bees around 10⁶, and in frogs 10⁸.

Human vs Computer Intelligence

• Machine AI is more successful when solving specific tasks, like identifying objects, playing certain games, or translating text
• Human mental capacity has thought towards matches human equivalency
• Comparing problem-solving differences include learning efficiency, generalization and transfer, imagination
• AI achieves superior performance by processing more training data than humans can
• Examples of humans achieving expert level proficiency in more than one game versus AI models that are specific to the game they have been trained for.

Transfer Learning

• Transfer learning takes an existing model and applies it to a related task
• Meta learning: Takes one step back from concrete tasks and is concerned with the problem of learning to learn

Generative Adversarial Networks (GAN)

• This approach combines two neural networks that work in opposition to each other, hence the adversarial of GAN
• One network tries to generate data from a category while the other decides which data is real versus generated
• These enhance generator/discriminator to be like one another

Super Intelligence

• Super intelligence is the belief in creating an AI to exceed the capabilities of the human mind
• The machine continues to improve itself leading to " technonological singularity"
• Vinge (1993): Thought greater than human intelligence would be achieved in the next thirty years
• Kurzweil: His 2005 book " The Singularity is Near" discuses the concepts worldwide

Anti-Singularity

• An explosion of AI won't exist until machine intelligence passes the human variety in all areas
• Largely, human intelligence depends on a number of factors, such as emotion
• Overtime AI discoveries become more difficult
• Forecasting large returns from artificial intelligence may be observable eventually leading to diminish returns
• The individual mind is the only focus in the AI theory
• Modern collective intelligence in a developed society creates lives