Midterm 2 Flashcards
who found that you could redefine what is meant by computation manipulating DNA to solve mathematical problems? What was his main finding in the paper?
Leonard M. Adleman
A UTM can be built with DNA molecules
What is the hamiltonian path problem?
Given a network of nodes and directed connections between them, is there a path through the network that begins with the start node and concludes with the end node visiting each node only once.
Binary (yes or no)
Generation and test method for hamiltonian problem with computer
Step 1: Generate random paths on the network.
Step 2: Keep only those paths that begin with the start
city and conclude with the end city.
Step 3: If there are N cities, keep only those paths of length N.
Step 4: Keep only those that enter all cities at least once.
Step 5. Any remaining paths are solutions (I.e., Hamiltonian
paths).
what is the issue with the generation and test method for a computer?
leads to combinatorial explosion. brute-force serial search takes a very long, virtually forever (unintelligent computation)
What is the key to solving the hamiltonian problem using DNA
performing the 5 steps in generation and test method in parallel search instead of serial
Who discovered the double helix and launched molecular biology?
Watson & Crick
anealing
complementary single-stranded DNA or RNA sequences spontaneously bind together by hydrogen bonds to form a double-stranded molecule
polymerases
copy information from one molecule into another
ligases
bind molecules together
used to repair breaks in DNA
DNA synthesis experiment
the “bio” nanomachines
-slides along
-reads each base
-writes its complement onto new strand
How did they encode the problem into DNA?
- In a test tube, mix the prepared DNA pieces together
(which will randomly link with each other, forming all different paths). - Perform PCR with ‘start’ and ‘end’ DNA pieces as primers (which creates millions’
copies of DNA strands with the right start and end). - Perform gel electrophoresis to identify only those pieces of right length (e.g., N=4).
- Use DNA ‘probe’ molecules to check whether their paths pass through all
intermediate cities. - All DNA pieces that are left in the tube should be precisely those representing
Hamiltonian paths. - If the tube contains any DNA at all, then conclude that a Hamiltonian path exists, and
otherwise not. When it does, the DNA sequence represents the specific path of the
solution.
Why does DNA work to solve the problem
its enormous parallelism
How many possible paths in the travelling salesman problem?
over 4 million
Working memory
Limited time capacity: contents of current awareness; storage mechanism that keeps a limited amount of information active fro a brief time such as a few seconds to minutes
acoustic
18 seconds duration
7+-/2 items capacity
Brown-peterson distractor task
Retention memory curve
participants count backwards by 3s during retention interval
memory span drops steeply because of retroactive interference
Modal model of memory (atkinson & shiffrin)
classical information-processing view
info -> sensory memory -> STM (maintenance via rehearsal which goes to LTM, or lost) -> LTM and retrieval back to STM
- modern model replaces STM with WM
Is there a ‘place’ in the brain for WM
No
WM is just referring to a concept or label for some sort of mental activities going on all over the brain, rather than a buffer place or loading dock
WM vs LTM
WM: temporary, fragile duration, limited capacity and size, relatively easy ease of entry, relatively easy ease of retrieval
LTM: long-lasting, enduring duration, virtually unlimited capacity and size, effortful ease of entry (rehearsal), ease of retrieval difficult and slow and can be unsuccessful
Waugh & Norman probe task
report digit after probe
serial position curve
Primacy effect - LTM
Recency effect - STM
List length effects
better recognition for shorter-list items
Computational modeling of serial position curves
SIMPLE model - Bayesian model
CMR model - connectionist model
Both model the same pattern
Evidence for separate stores of WM and LTM
error patterns ion memory tasks
double dissociations after brain lesions
H.M. Memory pattern
remembers old explicit memories formed before surgery but cannot form new explicit memories (severe anterograde)
Evidence for separate implicit and explicit memory systems
H.M. explicit memory deficits but performs normal on implicit tasks (motor learning, classical conditioning, priming)
explicit memory depends on the temporal lobe and implicit does not
explicit vs implicit
explicit: medial temporal lobe, hippocampus, conscious, fast acquisition speed, late development
implicit: multiple regions of the brain, esp cerebral cortex not conscious, slow acquisition speed, early development
roles of hippocampus in memory/learning
- Specialized for cross-modality associations (“binding”) between events/objects
that are context-specific and episodic (i.e., space-time stamped) - The hippocampus is bi-directionally connected to a wide array of cortical areas
cross modality in the hippocampus
Different aspects of an episode (e.g.,
sights, sounds, smells) are represented
and stored in disparate cortical areas.
* During learning, the hippocampus
specializes in the rapid formation of
arbitrary associations between cortical
areas
* During retrieval, the hippocampus binds
and reconstructs separate memory
traces into a coherent episode.
learning vs memory
Learning refers to long-lasting changes in behavior resulting from long-
lasting physical changes in the brain, most likely at neuronal synapses.
Memory traces as synaptic connection weights:
– What is a mental state? It is a pattern of activation
over neurons.
– Patterns of activation, or mental states, come and
go, leaving traces behind when they have passed.
What are the “remembered” (memory) traces? They
are the synaptic connections between neurons.
In short, memory is the product of learning.
What is the computational mechanism of hippocampal learning and memory?
hebbian learning at synapses
Electro-neurophysiological experiment in hippocampus
performs in way Hebb predicted, connection strengthens when conductivity goes up & later on it will still fire strongly with only a small pulse
Long term potentiation as a candidate mechanism for hebbian learning
The ability of hippocampal neurons to acquire and maintain a long-
lasting strengthening in synaptic conductivity following a repeated
stimulation of pre-synaptic neurons with high-frequency electrical
pulses.
synaptic plasticity hypothesis
the neuron doctrine
Synaptic plasticity hypothesis: the strength and/or
number of synaptic connections changes as a result
of experience.
è “Memory is in the synapses.”
simple hebbian rule vs better hebbian rule:
simple: change in weight between sender x and receiver y = learning rate * postsynaptic activation * presynaptic activation
-two neurons that fire together increase synaptic connections between them.
-problem is that the weights in this form get infinitely large
Improved: change in weight between sender x and receiver y = learning rate * postsynaptic activation * (presynaptic activation - current weight)
-uses info that is available locally at the synapse
hebbian learning
Hebbian learning makes small changes in the connection
weights between two units: a sender neuron and a
receiver neuron.
2. If both neurons are active, the connection grows stronger
(cf., LTP). Cells that fire together wire together.
3. If one neuron is active but the other is not, the connection
grows weaker known as Long Term Depression (LTD).
4. A simple equation captures both these properties. It is
used in neural network models that abstract away from
the biological details:
learning vs memory in classical conditioning:
Learning is the process of behavioral change as a result
of experience
Memory formed in the brain is the end product of a
behavioral change
Learning is associative in nature. That is, learning
occurs when an association is formed between two stimuli or between a stimulus and a response.
-via bi-directional projections from/to hippocampus
classical conditioning phenomenon
A type of learning in which an artificial stimulus evokes a response that was originally associated with a natural
stimulus
language
a system that uses physical symbols (sound,
mark on paper) to express meaning:
linguistics
the scientific study of language, attempts
to answer the questions like what is language? and how is
it represented in the mind?
linguistic competence is the ability to:
– Produce and understand a practically unlimited number of
utterances, including novel and unfamiliar ones.
– Recognize that certain (“ungrammatical”) utterances do not
belong in the language.
what is the central subject matter of linguistics?
linguistic competence
Six features of language
- Communicative: Permits communication between
individuals - Arbitrary: The relationship between language symbols and
their meaning is arbitrary
(e.g., “boy” vs “buy” vs “purchase”) - Structured: The (sequential) pattern of language symbols
matters, and is not arbitrary but rule-based:
“the boy ran from the dog” vs “the dog ran from the boy” - Generative: Basic units of language (words) can be used
to generate a practically unlimited number of expressions
(i.e., “infinite” possibilities). An expression that has never
seen/heard before can be generated and understood with no
problem. - Complex:
Among the practically unlimited number of possible
expressions, only a very small portion (still an enormously
huge number) of it are meaningful. - Innate:
We know there are in the order of 10 ~50 different sentences or
expressions, grammatical or ungrammatical; however, there is no way
you can be exposed to all these possible sentences, not in your life time
– the poverty of the stimulus.
Innate in the sense that the language is learned effortlessly by all
normal humans before they reach the age of five, with virtually no direct
instruction and with little or no systematic presentation (e.g., +/-
feedback) of relevant data.
Any child has the capacity to acquire any natural language regardless
of genetic heritage:
Notion of Noam Chomsky’s Linguistic Universals (Universal Grammar):
Genetically determined, hard-wired predisposition of linguistic features
common to all languages (e.g., nouns & verbs, consonants & vowels), not
something that can be acquired by infants naturally in the absence of pre-built-
in constraints and rules on the form of language.
number of possible sentences for just 10 words
10! (3,628,800)
most important features of language
Generative and complex features
left vs right brain functions
left: analytic thought, logic, language, science and math
right: holistic thought, intuition, creativity, art and music
what brain area connects the broca’s and wernicke’s area?
arcuate fasciculus
lateralization of language
Left brain: 98% of right-handed people have almost all of the
language functions represented in the left-hemisphere of
the neocortex
Right-hemisphere processes language functions regarding
metaphorical or artistic meaning of language (e.g., humor)
In left-handed people, language function is represented
more dominantly in the right-hemisphere of the neocortex
than as in the case of right-handed people, though still
mostly represented in the left-hemisphere
brain areas and functions
Broca’s Area
Production of spoken language (motor programs for controlling speech
sounds)
Wernicke’s Area
Comprehension of language (interpretation of spoken and written words)
Arcuate Fasciculus
Connection between Broca’s and Wernicke’s
Visual Cortex
Processing of written language
Angula Gyrus
Connection among Broca’s, Wernicke’s and visual cortex
Mortor Cortex
Making of speech sounds (i.e. controlling of vocal muscles
Broca’s aphasia
Damage to Broca’s area
- Prevents a person from producing speech (i.e., spoken words).
- Nonfluent, telegraphic speech
- Words are not properly formed (more like a telegram)
(e.g.) “I’m a sig … no… man … uh, well, … again.”
“Well..mess..uh..sgga..diz..es..”
- Person can understand language (both spoken & written
wenicke’s aphasia
- Damage to Wernicke’s area
- Loss of the ability to understand language (both
spoken & written words) - Fluent but unintelligent speech
- Can form words properly but the words that are put
together make no sense
(e.g.) “I go to a dog of cookies in TV”
“I cooked a radio for my mother on the door
conduction aphasia
- Damage to Arcuate fasciculus
- Fluent speech/good comprehension, but unable
to repeat what is heard or read
acquired alexia
- Damaged connection between visual cortex and
Wernicke’s area - Inability to read, but can see words
agraphia
inability to write
dyslexia
Loss & deficits of reading skills, spelling and recognizing
word sounds
father of behavior based robotics
rodney brooks
roomba and big dog
not turing machine, wouldn’t pass turing test, minimal representation, minimal intelligence, weak AI
three disciplines of linguistics
phonetics: sound structure of language
semantics: meaning structure
syntax: grammatical structure(chomsky made greatest contribution to this area)
phonetics
When language is spoken, words are represented as sequences of discrete
sounds or utterances – an acoustic signal produced in the upper
respiratory and vocal tracts.
One-to-one relationship between an abstract symbol and the
corresponding phoneme.
(e.g.) Bet <–> /bet/
semantics
Semantics is the study of meaning. It is concerned to describing how we
represent the meaning of a word in the mind and how we use this
representation in constructing sentences
A sentence is an expression of world knowledge and/or personal belief that
is true or can be true under certain circumstances. Sentences are
related as synonymous or contradictory (i.e., similarity structure).
latent semantic analysis
A Computational
Theory of Word
Meaning
The basic idea of LSA: Words that
are close in meaning tend to
appear in the vicinity of one
another in sentences, thus
occurring in similar contexts of
text (i.e., co-occurrence
relations)
syntax
syntax is about sentence structure. It attempts to describe
what is grammatical in a particular language in terms of rules that
detail an underlying structure and a transformational process
Main question of interest: “What makes a sentence grammatical,
and what makes it ungrammatical?
linear-chain hypothesis
A grammatical (so meaningful) sentence is
formed as a (linear) sequence of words.
* Behaviorist account (e.g., Skinner, 1957) –
sentences are learned and produced as chains
of responses :
– A stimulus S 0 elicits the first word.
– The first word serves as the stimulus for the
second word, and so on (a kind of
associative learning)
finite state grammar
skinner
generative aspect: An infinite (or, at least, astronomical) number of
sentences can be generated and understood with finite
knowledge and resources.
learning aspect: learning language would take too long w FSG
cannot account for human language
FSGs have extremely limited memory – only the last
word that was produced.
* Dependencies between words in a sentence can
extend over many words (e.g., embedding)
general FSG
grammar can be learned by associating pairs of words
chomsky
Arguably the most influential linguist ever
* Critic of behaviorism, so linear-chain hypothesis of
language
* Transformational-generative grammar
* Nativist theory of language development:
– Universal grammar
– Principles-and-parameters theory
– Deep and surface structure
Phase structure
Sentences are not formed by simply stringing words or
categories together in a linear/sequential manner, like
beads on a necklace.
* Rather, sentences are structured in a hierarchical
design (not as a sequence of words as in FSG, nor as a
sequence of word categories as in GFSG) in which
words are grouped together into successively larger
structural units called phrases (i.e., word groups
two unique aspects of generative PSG
- embedding (a sentence within a sentence)
- recursion (self-replication/similarity)
chomsky’s contributions
The syntax of human language is a sophisticated, hierarchical
(i.e., non-linear) system that involves storage and manipulation
of hierarchical phrase structures.
* This manipulation occurs algorithmically, or rule-based.
* Mathematical theorems – Chomsky’s hierarchy of formal
grammars
language acquisition
- Language is universal across all human societies. That is, all societies
use language in similar ways. - Despite the apparent diversity of human languages, any language can
be learned by anybody. - Accordingly, languages must have some common underlying
structures
universal grammar
Universal Grammar (UG; Chomsky, 1965):
A limited set of abstract innate (i.e., hard-wired in the brain & not
something that can be taught or learned) organizing rules of
grammar that are universal to all natural languages
philosophy of mind questions
Ontology questions:
- What is out there?
- What is the nature of reality?
Epistemology questions:
- How do we know what is out there?
- How is knowledge possible?
Mind-body problem:
- Does the mind exist? In what form?
- Can a machine have a mind? (Strong AI View)
- How can a mind emerge from the neuronal activity in the brain (body)?
(M-B Problem
The mind is mental in nature and mental states are:
subjective and private (color thought experiment)
qualia
Philosopher’s term for the introspectively accessible, qualitative
character of subjective (mental) experiences such as feelings and
sensations
- Redness of red, awfulness of pain, sweetness of peach pie
- Emotion (sadness, anger)
- “Atoms” of recognizable inner experience
§ Not all mental states possess qualia: - Experience of suddenly understanding of a math problem (i.e.,
insight)
§ Hotly debated, hypothetical construct that is non-physical - Are qualia irreducible (i.e., intrinsic qualities of their bearers)?
- How do the mental qualia relate themselves to the physical world
outside? (i.e., mind-body problem)
infinite recursion problem
Is it possible or even allowed for one mind to study and
understand its own mind, or another mind?
ontological issues
The ontology (“being”, “existence”) of mind is the study of the nature of
mental states and their relation to physical states of the brain/body.
According to Plato’s theory of forms, world objects (i.e., the things we perceive
through our sensory organs) are just imperfect copies (i.e., shadows cast on
the mental-cave wall) of the true reality existing outside our mental universe.
Plato’s Cave
Kant’s two-world interpretation
A world of appearance whose existence and properties depend on human
perceiver; and another world of things in themselves whose existence and
properties are absolutely real in the sense that they would exist even if no
human beings around to perceive them
Four Theories about the Mind
and the Nature of Reality
- Materialism/Physicalism (only matter matters)
- Immaterialism/Idealism (only mind/ideas matters)
- Dualism (both mind and matter matter)
- Functionalism (neither matters but functions)
Dualism vs Monism
Dualism: The mental (mind) and the physical (body) are
fundamentally distinct.
* Monism: All is one, there are no fundamental divisions:
– Materialism/physicalism
– Idealism
– Functionalism (neutral monism)
- Materialism/Physicalism (Democritus)
§ Nothing exists but matter. Mental states are nothing
but sophisticated states of the brain (physical system),
and thus can be reduced to bio-physico-chemical
phenomenon of firing neurons.
- Pattern of neural firings defines mental state
- Reductionism of hard science & engineering
- Reductionism in modern neuroscience
Counterarguments:
- How can we account for qualia (seeing red, feeling
happy/angry), thoughts, desires, intentions, beliefs…?
Regarding Reductionism in Neuroscience
§ Three challenges in solving the
mystery of how the brain works:
1) Ethical
2) Technological
3) Theoretical (i.e., how to
interpret neural signals)
This is unlike other science &
engineering disciplines, such as
nanomaterials science
Idealism
Everything that exists is mental, and there is no
such things as the material. No material things
exist.
2. Idealism (G. Berkeley)
§ The world (self and surrounds) is nothing but a pure
creation of the mind. Reality, as we know it, consists wholly
of minds and their ideas. No mind, no reality.
§ Material things are nothing but collections of “ideas”
§ “Esse est percipi aut percipere.” (Berkeley)
(To be is to be perceived or to perceive)
Do not deny the existence of what “appears” to be the
[physical] world around us such as trees, buildings,
people, sky,… According to Berkeley, however, what
such objects turn out to be are no more than bundles
or collection of ideas, all created in the bearer’s mind
or “consciousness”.
§ “All reality is subjective experience.”
Observer-dependent Reality
What we see is undoubtedly real, they say, but these phenomena are
not really there in the absence of an observation.
Dualism (R. Decartes)
Mind (thinking substance) and body (extended substance)
are two completely different kinds of substances, each
capable of existing independently of the other and one
being causally linked to the other.
René Descartes
(1596-1650)
3. Dualism (R. Decartes)
§ Mental states are not reducible to physical states.
- Sensation of pain (mind) vs physical object that caused the pain
(matter/body)
- A human person is a composite being (“union”) of a mind and a body
§ The question that arises naturally is then:
- How is the mind related to the body?
- If the mental is not physical, then how can we make sense of its
causal interaction with the physical?
è Mind-Body Problem (MBP)
Cartesian Theater
Criticism of the dualism’s view of consciousness/mind.
* Decartes’dualistic view of the mind necessarily implies what might
be called the Cartesian Theatre.
* In other words, thinking (i.e., mental) entails the existence of a
thinker (i.e., physical). In other words, there must be a central,
reference location in the brain that creates a mental experience (i.e.,
consciousness).
* This central reference point, physical in nature, is called the
puppeteer Homunculus (a “little man”or an “inner theatre” inside the
brain)
* The problem here, then, is that the homunculus itself must have a
brain, which must have yet another homunculus, thereby becoming
an infinite recursion problem, thus the impossibility.
Functionalism (Neutral Monism)
A mental state is nothing but a collection of physical states described
“functionally.” (i.e., mind as software)
§ A mental state is defined by the function it serves and performs,
characterized in terms of its role in relating inputs to outputs and its
relations to other mental states – nothing more and nothing less.
(e.g.) Feeling pain (mental state):
(1) Input: injury/trauma to body
(2) Causal relation: produce the belief that something is wrong
with the body and the desire to be out of that state
(3) output: moaning or wincing (i.e., pain sensation)
§ Multiple realizability of functionalism:
According to the functionalistic view, any other physical system that
serves the same function by undergoing similar causal sequences
can be characterized as having the same mental state of “feeling
pain,” for instance.
§ Related to the substrate independence of the mind: Mental states can
be realized on a variety of physical substrates whether on biological
neurons, DNA, or silicone-based chips (N. Bostrom)
- Neither materialism nor dualism
- No mind-body problem
Counterarguments:
- Qualia: color perception (sense of red vs green)
(Skeptical) Relativism
Each of us has access to only our own perceptions, and therefore
truth and reality are in the eye of the beholder… There is no
objective or absolute truth.” “Truth is relative.”
The measurement of temperature is relative to the scale we use, F or C.
§ (Question) Which one of the following four theories of mind is most
closely related to Skeptical Relativism?:
- Physicalism
- Idealism
- Dualism
- Functionalism
Three Mysteries (Realms) of Reality
Dr. roger penrose
physical world, mental world, mathematical
Zhuang Zi’s Butterfly Dream
(a type of skeptical relativism??)
Once Zhuangzi dreamed, he was a
butterfly, a butterfly fluttering around,
happy with himself and doing as he
pleased. He didn’t know he was
Zhuangzi. Suddenly he woke up and
there he was, solid and unmistakable
Zhuangzi. But he didn’t know if he
was Zhuangzi dreaming of a butterfly,
or a butterfly dreaming of Zhuangzi!
Cross-modal recognition
“Humans can easily recognize, through touch alone, objects that they have
previously only seen.”
is auditory experience a subjective experience?
yes
What you hear is not what is out there in the outside
world, but instead what has been interpreted &
created all inside of your head.
what is sound?
Sound is the psychological sensation that is
internally generated in the brain as a response to
an external physical stimulus called the ‘sound
wave’
necessary elements of sound
source (sound wave - the compression and rarefaction of air molecules)
and listener
five steps of auditory computation
1: Outer ear (eardrum) (sound detection)
2: Middle ear (sound amplification)
3: Inner ear (neural pulse generation)
4: Superior olive (relay station - interaural mixing)
5: Auditory cortex (conscious perception)
speech spectrogram
a visual representation of the frequency content of a sound over time, used to analyze speech and understand its acoustic characteristics, like formants and voicing.
frequency-to-location coding in cochlea
Each group of cells has its own
fundamental frequency, to which it
maximally responds. Note the nonlinearity of the map.
cochlea is in essence:
an audio frequency analyzer
Frequency sensitive cells in Cochlea
unequal sensitivity, some Hz need more intensity to be heard. detection relies on intensity being above a certain threshold
Topographic maps of auditory encoding:
Similar frequencies are mapped to nearby points in the cortex
sound localization in auditory cortex
by interaural time difference
How is interaural time difference calculated
time-to-position coding; coincidence detecting neural circuit
laminaris neurons
coincidence detector neural circuit for each frequency channel
