Cognitive Science Midterm #2

The Turing Test is a test to see if a computer is intelligent.
Basically you get a person to interact with a computer and see whether or not they think that it is human

starts with a higher level of understanding and then work your way down to the details.

Basically you break the problem down into smaller, easier problems.
GOFAI(GOOD OLD FASHION AI)

It's already hard coded

type of information processing that starts with raw data and builds up to a higher level of understanding.

While the symbolic approach attempts to replicate human like reasoning through explicit rules and logic, the connectionist approach aims to emulate the human brain's structure and function by leveraging interconnected nodes or neurons

Magic Sandwich Hole - neural network that measures how good a sandwich is

Big sanwich bug - a sandwich that contains mulch might still be rated tasty if it contains good enough ingredients to cancle out th emulch
Adding hidden layers - this creates a more sophisticated algorithm, one thats able to judge sanwiches as more than the sum of their ingredients
--------------Adding layers maen a new way to combine the insights from the previous layer

activation functions - non linear functions to be able to capture complex variables more effectively.

Deep learning uses artificial neural networks to learn from data.
Its trained by feeding the neural network data and allowing them to adjust their weights until they can accurately predict the desired output.

Occurs when a dataset has a significant imbalance in the number of examples in each class.

Imagine you are trying to train a deep learning model to classify images of cats and dogs. You have a dataset of 100 images, and 90 of the images are cats and 10 of the images are dogs.

If you train the model without addressing the class imbalance, the model will likely learn to classify all of the images as cats. This is because the majority class is cats, and the model is trained to minimize the overall error rate.

A Markov chain is a stochastic model of a sequence of possible events in which the probability of each event depends only on the state attained in the previous event. A Markov chain process can be viewed as a system of states that are linked by transitions. Each state represents a unique configuration of the system, and each transition represents the possibility of a change from one state to another. The probability of a transition from one state to another is known as the transition probability.
The predictive text suggestions are based on the words that the user has already typed, and on the probability of those words being followed by other words in the dataset.

Evolutionary algorithms are a class of optimization algorithms that are inspired by the process of natural selection. They work by iteratively generating a population of candidate solutions, evaluating the fitness of each solution, and then selecting the best solutions to generate the next population. This process is repeated until a satisfactory solution is found.

Evolutionary algorithms are often used in deep learning to optimize the hyperparameters of neural networks. Hyperparameters are the parameters that control the training and structure of a neural network, such as the learning rate, the number of layers, and the number of nodes per layer.

ML does best with a simple, narrowly-defined goal, where objective feedback on the quality of the judgment is possible.

Bias can occur when the data that is used to train an ML algorithm is not representative of the population that the algorithm will be used on. For example, an ML algorithm that is trained on images of mostly white people will be more likely to misclassify people of color.

Cones are the main color receptors in your eye. Along with the rods, they sit on your eye’s retina and perceive light. Cones use three different types of photoreceptors. They are aptly named short, medium, and long-wavelength cones (also known as red, green, blue cones). Each type is responsible for detecting the lights and colors within its respective wavelengths.

rods are receptive to motion and cover the blacks and whites in our vision. Rods are the reason we can see in the dark. They detect low levels of light and makeup what is called our scotopic vision (cones make up our photopic vision).

The human eye can only see a small portion of the electromagnetic spectrum, which is called visible light. Visible light has wavelengths between 380 and 780 nanometers. Electromagnetic waves with wavelengths outside of this range are invisible to humans. Some animals, such as insects and birds, have eyes that are sensitive to different wavelengths of light than humans. For example, bees can see ultraviolet light, which is invisible to humans. This allows them to see flowers and other objects that humans cannot see.

Ventral stream is responsible for object recognition and visual perception, while the dorsal pathway is responsible for spatial processing and action guidance.

"what" pathway
originates in the primary visual cortex and goes to the temporal cortex
Responsible for object recognition

"where" pathway
originates in the primary visual cortex and goes to the posterior parietal cortex
responsible for processing information important for spatial perception and visuomotor performance

Imagine you are looking at a picture of a vase or two faces. From a purely bottom-up perspective, the image is ambiguous: there is no clear evidence that it is one thing or the other. However, our top-down knowledge of the world tells us that vases and faces are two different things, and that they cannot both be present in the same image.

As a result of this top-down knowledge, we are more likely to interpret the image as one thing or the other, depending on our prior experiences and expectations. For example, if we have seen similar images of vases in the past, we may be more likely to interpret the image as a vase. Conversely, if we have seen similar images of faces in the past, we may be more likely to interpret the image as two faces.

Another way to think about it is that top-down processing allows us to impose our own meaning on the world around us. In the case of ambiguous stimuli, this means that we can choose to see one thing or the other, even though the stimulus itself is ambiguous.

the process of interpreting visual information based on our prior knowledge and expectations.
- includes things like our understanding of the world, our current goals, and our emotions

the process of interpreting visual information based on the physical feature of the sitmulus itself
- includes color, shape, texture

is characterized by an inability to perceive the basic features of objects, such as their shape, size, and color.
-damage to visual cortex

is characterized by an inability to recognize the meaning of objects, even though the basic features of the objects can be perceived.
-damage to inferior temporal cortex

Damage to the visual cortex is Apperceptive agnosia while damage to the temporal cortex

objects in the visual input compete for representation, analysis, or control
-----------Biased bottom up components
------------------driven by physical features of the stimuli
-----------Biased top-down components
-------------------driven by observer's goals and expectations
The object or feature that receives the most attention is the one that is most likely to be processed and recognized

all of the objects and features in a visual scene compete for attention because the brain has a limited capacity

brain can select one object on feature to attend to, even in the prescence of competing stimuli

participants were presented with a visual display of multiple objects and asked to search for a specific target object. Treisman found that the speed of search was determined by the number of objects in the display, and by whether the target object was different from the other objects (i.e., whether it stood out from the crowd).

These findings suggest that the brain has a limited capacity to process all of the objects in a visual scene at once, and that it prioritizes the processing of objects that are different from the other objects in the scene.

is a phenomenon that occurs when people fail to notice something that is in plain sight, because they are attending to something else. This is a good example of how top-down factors can bias the competition for attention.

One famous example of inattentional blindness is the gorilla experiment. In this experiment, participants are shown a video of people passing a basketball around. The participants are asked to count the number of passes made by the team wearing white shirts.

At one point in the video, a person wearing a gorilla suit walks across the screen. However, many participants fail to notice the gorilla, because they are so focused on counting the basketball passes.

Difficulty recognizing objects, even familiar ones
Difficulty distinguishing between different objects
Difficulty identifying the parts of an object
Difficulty copying or drawing objects
Difficulty navigating through space
Difficulty with activities of daily living, such as dressing and eating

a person is unaware of or ignores stimuli on the side of their body opposite the side of the brain that has been damaged.

Often results from dange to the right parietal lobe, leading to neglect of the left visual field.

Texting while driving is a dangerous and distracting activity that can lead to serious accidents. Even hands-free phone calls can impair your ability to drive safely.

A study by Strayer and Drews used EEG data to measure the brain activity of drivers who were texting, talking on a hands-free phone, or driving without distractions. The study found that all three activities reduced brain activity in the areas of the brain responsible for paying attention to the road.

To make inferences and to process information efficiently

Concepts are not always defined by logical rules. For example, there is no single logical rule that can be used to define the concept of "beauty." What is considered beautiful can vary from person to person and from culture to culture.

Through prototype rule

An analogy is a comparison between two things that are similar in some way. Analogies can be used to explain complex ideas, to make arguments more persuasive, and to create new and original ideas.

Babies are unable to chunk information in the same way that adults do. This is because they lack the cognitive skills necessary to identify and group together related pieces of information.

Taking “small” concepts and putting
them together into bigger and bigger
ones, thus recursively building up a
giant repertoire of concepts in the
mind.

The longer we live, the larger our repertoire of concepts becomes,
which allows us to gobble up ever larger coherent stretches of life in
single mental chunks

When you were a kid you are still getting new experiences. Which means you are getting more chunks. Once you grow older everything is already chunked so it isn't as special anymore.

Grandmasters have the ability to recognize and remember complex configurations of chess peieces based on patterns and prior knowledge, wheras novices don't have this extensive pattern recognition.

However, if the configuration is fake the grandmasters won't be able to recognize it.

Concepts help us label situations and draw our attention to particular aspects of our environment

"War of Ghosts" experiment:
Participants retold a Native American story, and tehir retellings were influenced by their own cultural backgrounds and concepts, demonstrating the role of existing knowledge in shaping perception and memory.
- The british school children converted the Native American story to have british elements.

situation label is a cognitive representation that summarizes the key features of a situation. Often summarized through the use of concepts

In this process, the mind is constantly making comparisons between new experiences and existing knowledge. This comparison process allows us to learn new things, make predictions, and solve problems.

Leopard Alarm, Snake Alarm, Eagle Alarm,
The grunts can be used to express a range of emotions, such as aggression, submission, and contentment.
Vervet monkeys also have a variety of distress calls that they use to signal when they are in danger. These calls can be used to attract the attention of other monkeys or to deter predators.

Humans are social to a degree that makes it an outlier among other species. Our social structure is based on cooperation and shared intentions. Humans point. Human languages are endlessly adaptable.

- phonetics - the study of speech sounds in language
phonology - the study of the patterns of speech sounds in language
morphology - The study of **word structure** and word formation.
syntax - The study of sentence structure
semantics - the study of meaning of words and sentences
pragmatics - The study of how **context** influences language understanding

the study of speech sounds in language

- the study of the patterns of speech sounds in language

**: In the standard linguistics view, abstract representations of sound units

** Representations of actual sounds as perceived. Dictionaries often give phonetic representations of words.

The study of **word structure** and word formation.

- the study of meaning of words and sentences

The study of sentence structure

The study of how **context** influences language understanding

The Left Hemisphere

involved in motor function
Damage causes difficulty in speech production

Difficulty in speech(and reading)comprehension

There are no numbers
No fixed color terms("this looks like blood", "this looks like berries")
- No perfect tense(e.g. we have talked before)
- All verbs are marked with by source of information(I saw it, I was told it, etc).
- No deep memory(no creation myths)
- No tradition of art/drawing
- No directional words for right and left. Instead, they give directions in absolute terms, telling others to head "upriver" or "downriver" or "to the forest" or "away from the forest"
- Very few kin terms. Mostly cohort based: The cohort above me (genderless), my cohort(including siblings), the cohort below me, biological son, biological daughter(only gendered kin terms).

The piraha don't have recursion
For example, the sentence "The cat that chased the mouse that ate the cheese is brown." contains four clauses: "The cat," "that chased the mouse," "that ate the cheese," and "is brown." The clause "that ate the cheese" is embedded within the clause "that chased the mouse," and the clause "that chased the mouse" is embedded within the clause "The cat."

Piraha have to say" The cat chased the mouse", "The mouse ate the cheese", "the cheese is brown"