Phil 340 Midterm

necessary conditions

A has gotta be so for B to be so

if A isn't so, then B isn't so

B entails A

sufficient conditions

A guarantees B

If A is so then B is so

A implies B

what are the relationships between understanding, representation, explanation, and leaving things out

Mind mental properties, processes, etc. Brain physical properties, processes, etc.

"...mechanisms and explanations are always at some level of explanation. A typical explanation about combustion motors in automobiles will invoke pistons, fuel, or controlled explosions. It will not discuss these phenomena in term of particle physics, for instance; it won't invoke electrons, protons, or neutrons."

Some levels of explanation: What are the goals of the computation? What would constitute success? How does the algorithm work? How are the inputs and outputs represented? How is the algorithm realized in a particular language? (English, Python, C#,...) How is the algorithm realized physically?

turing machine

an abstract device that describes a procedure for solving a problem

basic properties of turing machines

mathematical problem-solving machine

computability thesis (aka the church-turing thesis)

a function on the natural numbers can be calculated if and only if it is turing computable

mind

mental properties, processes, etc

brain

physical properties, processes, etc

ambiguity

multiple interpretations are plausible for a use of a word

a word could be interpreted multiple ways each of which makes some sense of its use

equivocation

different interpretations are plausible for different uses of a word

different interpretations of a word are being used across two or more occurrences of the word

vagueness

the word has borderline instances

there appears to be no determinate answer as to whether the word applies (in a range of cases)

turing test

conversation between machine and human that will indicate the machine is intelligent if the human cannot discern that it is a machine rather than a human

winograd sentenes

The city councilmen refused the demonstrators a permit because they [feared/advocated] violence

1a. The city councilors refused the demonstrators a permit because they feared violence.

1b. The city councilors refused the demonstrators a permit because they advocated violence.

GOFAI

good old fashioned artificial intelligence

'symbolic AI'

AI modeled on the mind as we commonly understand it, as something that works with discrete symbols that are about things

expert systems

rule-based systems that encode human knowledge in the form of if/then rules

categorical rules

rules that specify exactly what is possible or impossible in a language

hypothetical rules

identify actions we ought to take, but only if we have some particular goal.

Simulation vs Reality

The simulation hypothesis proposes that what humans experience as the world is actually a simulated reality, such as a computer simulation in which humans themselves are constructs.

strict/exception permitting rules

Symbols are manipulated by strict rules.

skills

knowing what is and isn't relevant

knowing how something is relevant

expertise

the key point, features, and course of action are immediately known

reflection and deliberation on the choice of perspective are important towards improvement and avoiding tunnel vision, but not simply calculative deliberation

validity

if all premises of the argument are true then the truth of the conclusion is GUARANTEED

test for this by evaluating the FORM of the argument alone

soundness

the argument is valid (ie the truth of the premises guarantees the truth of the conclusion) AND all the premises are true

test for this by evaluating the form of the argument AND the truth value of each premise

deductive arguments

arguments that lead to necessary conclusions when their reasons are true

deductively valid

if the argument's premises are true, then its conclusion must be true

deductive invalidity

it's possible for the conclusion to be false even if the premises are true

deductive soundness

an arguments premises are true and the argument is valid

inductive arguments

arguments whose reasons lead to probable conclusions

reasonable inferences

Make sense and are based on observations and knowledge

rationalism

belief in reason and logic as the primary source of knowledge

our nature provides us with knowledge of some truths, some of our concepts

these truths and concepts undergird other truths and concepts

behaviorists

John Watson and BF Skinner

language is entirely learned

Chomskyans

language is at least in part innate (provided by our nature)

performance vs competence

What we see or hear vs. what you know (cannot be observed)

physical symbol system hypothesis

1. symbols are physically instantiated, "static and immutable"

2. symbols are manipulated by strict rules

3. symbols are arbitrarily associated with referents

+this is "the necessary and sufficient means for general intelligent action"

behavioral AI

Behavioral AI is different from the standard cybersecurity approach when handling threats because while the traditional approach can handle known threats, behavioral AI can handle both known and unknown threats in real-time.

basic approaches of 4E approaches to cognition

embodied

embedded

extended

enactive

embodied

the body can constrain concepts, it is part of the cognitive system

important to (human) action, perception, reason, knowledge of facts, ethical thought and judgement

embedded

cognitive tasks can be made easier by the environment

extended

environmental and social resources can be part of the cognitive system

enactive

cognition (and perception) involve sensorimotor activity

important to (human) action, perception, reason, knowledge of facts, ethical thought and judgement

serial processing

occurs when the brain computes information step-by-step in a methodical and linear matter

parallel processing

the processing of many aspects of a problem simultaneously

basic features of neural networks

usually organized in layers known as a multilayer perceptron.

the weights associated with each connection are crucial to the operation of a neural net because it breaks it down to a list of numbers

training requires finding appropriate numeric weights (normally by adjusting the weights after each training episode, trying to make the network correctly map inputs to outputs).

opaque

impossible to see through; preventing the passage of light

deep learning machine learner

not embodied

transparent

Allowing light to pass through

ex: turing machines

embodied

respect for persons

a regard for others as reasoning persons

maximizing utility

bringing about the greatest ratio of pleasure possible

conflicting ethical considerations

when Samantha sends an email about Theo's letters to get them published

she does not consider his feelings or his privacy in his work

biology's axiom

different brain structures implies different functions

"different structure implies different function"

biology's axiom that Pessoa does not agree with (The Entangled Brain)

functionalism

a school of psychology that focused on how mental and behavioral processes function - how they enable the organism to adapt, survive, and flourish.

cortex

the cerebral cortex, aka the cerebral mantle, is the outer gray matter neural tissue of the cerebrum of the brain in mammals. It is organized into a series of layers

cortical

pertaining to the cortex

subcortex

the lower part of the brain responsible for various physiological processes necessary to stay alive

subcortical

Structures that lie beneath the cerebral cortex, but above the brain stem.

neuron

a nerve cell; the basic building block of the nervous system

axon

the extension of the neuron that comes into close contact with other neurons and typically affects the dendrite of a postsynaptic cell via the synapse

dendrite

the extension of the neuron that typically receives stimulation (from axons)

soma

cell body

synapse

the space between two neurons that permits the presynaptic neuron to pass a chemical signal to the postsynaptic cell

gray matter

the part of the brains tissue consisting of neurons and other related cell types, such as glial cells

white matter

contains mostly long-range axons

projection

anatomical pathway comprised of bundles of axons that connect two areas

spinal cord

a major part of the central nervous system which conducts sensory and motor nerve impulses to and from the brain

hindbrain

portion of the central nervous system in vertebrates that includes the medulla, pons, and cerebellum

midbrain

portion of the brainstem that is closest to the forebrain. among the regions that are in the midbrain are the superior colliculus and areas that produce the neurotransmitter dopamine (ventral tegmental area and substantia nigra)

forebrain

in animals with vertical posture, it is the most superior part of the brain that contains the cerebral hemispheres. in humans, it contains the cortex and subcortical structures

amygdala

subcortical structure extensively studied in the context of aversive conditioning but involved in a very large array of functions. the "amygdala complex" contains at least a dozen subnuclei. In very broad terms, it is useful to consider a "basolateral amygdala" and a "central amygdala"

hypothalamus

A neural structure lying below the thalamus; it directs several maintenance activities (eating, drinking, body temperature), helps govern the endocrine system via the pituitary gland, and is linked to emotion and reward (sham rage).

striatum

part of the subcortex (in the forebrain) consisting of the caudate and the putamen

superior colliculus

receives visual sensory input

thalamus

the brain's sensory control center, located on top of the brainstem; it directs messages to the sensory receiving areas in the cortex and transmits replies to the cerebellum and medulla

PAG

periaqueductal gray

dopamine

A neurotransmitter associated with movement, attention and learning and the brain's pleasure and reward system.

hypothetical 'minimal brain'

allows an animal to defend itself and seek rewards, essential components of survival

action flexibility

necessitates uncoupling sensory and motor components

this 'solution' frees animals from acting simply based on sensory simulation

modularity

degree of interdependence of the many parts that comprise a system of interest. a decomposable system can be said to be modular, whereas a nondecomposable system is not modular. more generally, modularity can be conceptualized as varying from low to high

decomposability

the computational interactions within subsystems are much more complex than those between subsystems

is modular

reductionism

the type of approach central to science, in which an organization of greater complexity is understood in terms of the contributions of its subparts, which when put together give rise to the behavior of the broader system

one-to-one mapping

one structure is involved in a single function

One-to-many mapping

one structure carries out multiple functions (amydgala)

many-to-one mapping

multiple areas can carry out one function (such as aversive processing)

many-to-many mapping

elements of biological systems, like areas of the brain, exhibit the most complex mapping where multiple structures have multiple functions

automaticity

the ability to process information with little or no effort

selective information processing

For an individual, only thinking about the information that is really needed and disregarding the rest of the information presented.

amygdala

reward prediction error

a mismatch between actual and expected rewards

appraisal

evaluation or estimation of worth

cognitive control

Effective control of thinking in a number of areas, including controlling attention, reducing interfering thoughts, and being cognitively flexible.

classical theory of concepts

concepts as necessary and sufficient conditions

prototype theory of concepts

statistically common properties

theory theory of concepts

concepts get content from the roles they play in theories

transformers

to do something at least somewhat similar for sequences of tokens that make up a piece of text

instead of just defining a fixed region in the sequence over which there can be connections, instead introduce the notion of "attention"—and the idea of "paying attention" more to some parts of the sequence than others

Luiz Pessoa

"'biology's axiom'- different structure implied different function"

"...functionalism, which asserts that mental states are identified by their functional role-not by how they are physically implemented"

Jill Lepore

simulmatics: the technological in-between pre-1950s behavioral science with today's world ruled by social media.

Hubert Dreyfus and Stuart Dreyfus

"the expert is simply not following any rules!"

Arthur Glenberg, Jessica Witt, and Janet Metcalfe

PSSH(Physical Symbol Systems Hypothesis)

Rodney Brooks

He explained that it is no longer a question of whether a human-level artificial intelligence will be developed, but rather how and when.

"The world is its own best model"

Sara Hooker

"Successful breakthroughs are often distinguished from failures by benefiting from multiple criteria aligning surreptitiously. For computer science research, this often depends upon winning what this essay terms the hardware lottery— avoiding possible points of failure in downstream hardware and software choices."

Rich Sutton

"the biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective, and by a large margin."

"One thing that should be learned from the bitter lesson is the great power of general purpose methods, of methods that continue to scale with increased computation even as the available computation becomes very great. The two methods that seem to scale arbitrarily in this way are search and 92 learning."

"The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex..."

"...we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. All these are part of the arbitrary, intrinsically-complex, outside world."

"They are not what should be built in, as their complexity is endless; instead we should build in only the meta-methods that can find and capture this arbitrary complexity."

"Essential to these methods is that they can find good approximations, but the search for them should be by our methods, not by us. We want AI agents that can discover like we can, not which contain what we have discovered. Building in our discoveries only makes it harder to see how the discovering process can be done."