Cognitive Science Post Midterm #2

the natures of mind and body are completely different from one another and that each could exist by itself.

Cartesian dualism is like saying that your mind is a separate person living inside your body. This idea is hard to believe because there is no scientific evidence to support it. Our brains and our minds are very closely connected. When your brain gets hurt, your mind can get hurt too. And when you think or feel something, your brain is working hard.

We know that the physical affects the mental

Qualia are the subjective, qualitative experiences of sensations, emotions, and thoughts. They are the "what it feels like" aspects of our mental lives. For example, the redness of a rose, the sound of a symphony, and the feeling of love are all qualia.

We cannot directly observe or measure the qualia of others, which makes it difficult to study them objectively. Qualia are subjective experiences, meaning that they are only accessible to the individual experiencing them. We cannot directly observe or measure the qualia of others, which makes it difficult to study them objectively.

Corpus collosum is disconnected. The corpus collosum connects the two cerebral hemispheres. This means that information processed by one hemisphere cannot be directly shared with the other, leading to a range of unusual behaviors and phenomena. This means that information processed by one hemisphere cannot be directly shared with the other, leading to a range of unusual behaviors and phenomena.

A test for intelligence in a computer that requires one condition for it to pass: the human being is unable to distinguish the machine from another human being through asking questions to both of them.

- As a result, the Turing Test examines the machine's intelligence through its ability to simulate human behavior.

- Tests the computers intelligence
Ultimately, it aims to respond to the question that Turing proposes, "Can machine think?"

Strong AI: Argues that a computer with the right program can actually have a mind and be intelligent, possessing understanding, consciousness, and other cognitive states.

Weak AI: Views computers as tools that can simulate intelligence, behavior, and thought processes but do not possess them inherently. The AI program is merely a model of the mind, not a mind itself.

Imagines a person locked in a room with Chinese symbols and a rulebook.
The person does not understand Chinese but can manipulate the symbols according to the rules, producing responses seemingly indistinguishable from a native speaker.
This scenario, according to Searle, demonstrates that manipulating symbols does not equate to understanding or intelligence.
Therefore, he argues, even an AI capable of passing the Turing Test (convincingly simulating human intelligence) wouldn't necessarily possess actual understanding or a mind.

Directly challenges Strong AI

So basically a careenium is when you have a bunch of networks interacting and if you have a lot of them then it suddenly gains consciuosness

It's a thought experiment that proposes that self-awareness and consciousness could emerge from complex interactions within a network of interconnected elements.

Western Educated Industrial Rich Democratic

It is not representative for the rest of the world.

WEIRD participants tend to be more individualistic and fairness-oriented, making them more likely to reject unfair offers than individuals from collectivistic cultures. ---Ultimatatum Game

The Müller-Lyer illusion is an optical illusion where two lines of equal length appear different due to the presence of arrowheads at their ends.
WEIRD individuals tend to experience this illusion more strongly than individuals from non-Western cultures.

Social learning is the process of acquiring new behaviors, knowledge, and attitudes through observing, interacting, and modeling others

Imagine a group of people trying to find food in a forest. Some people try different things (trial and error), while others just watch and copy what the successful ones do (social learning).

At first, everyone benefits from social learning because they can avoid the mistakes of trial and error. But as more people copy each other, there's less chance of anyone finding new food sources. Eventually, everyone starts starving because they're stuck with outdated knowledge.

This is Rogers' paradox: social learning can be helpful, but it can also trap a population in a rut. We need a balance between copying and exploring to survive in the long run.

Archaeological evidence suggests that Tasmanian technology became less complex after the island's isolation from mainland Australia around 12,000 years ago.
This decline included the loss of complex tools like bone implements, barbed spears, and spear-throwers.
This indicates that a reduction in population size and connectivity can negatively impact the development and maintenance of complex technologies.

Imagine a group of friends working on a puzzle together. If they're all huddled together sharing every idea, they might miss something important. But if they're completely separate, they might not have enough information to solve the puzzle at all.

Imagine two friends are caught cheating on a test. They're each put in separate rooms and offered a deal:

Confess and blame your friend: You get a slap on the wrist, your friend gets suspended.
Stay silent: If both stay silent, you both get detention.
Stay silent, but your friend confesses: You get suspended, your friend gets off scott-free.
This is the prisoner's dilemma. Both friends benefit most if they stay silent, but they're both tempted to confess and betray the other to get the lighter punishment.

This situation shows how people can be tempted to act selfishly, even if it hurts everyone in the long run. It's a reminder that cooperation is often the best choice, but it requires trust and the willingness to put the group's needs ahead of your own.

Imagine you're playing a game with a stranger, just once. You can either cooperate and both get a small reward, or betray each other and only one of you gets a big reward.

Since you only play once, you can't trust the other person to cooperate. You might think, "Why should I be nice if they can just betray me and get more?"

This is why cooperation is difficult in one-shot games. Everyone is tempted to betray each other to get the most for themselves, even if it means everyone ends up with less. It's like the saying, "There's no honor among thieves."

Imagine you're playing a game with the same person over and over again. If you both cooperate, you both get a decent reward. But if you betray each other, only one of you gets a big reward.

At first, you might be tempted to betray the other person to get more. But over time, you realize that if you both cooperate every time, you both get a good reward, while betraying each other only leads to punishment.

So, you and the other person start cooperating and both get the good reward. This is like how people learn to work together in real life. When we know we'll be interacting with someone again in the future, we're more likely to cooperate because we know it's the best way to get what we want in the long run.

Recurrent neural networks
- Think of your brain as an RNN. It remembers things you learn and uses that knowledge to understand new things. RNNs are like brains for computers. They are good at understanding things that happen in order, like words in a sentence or notes in a song.

• Graphics processing units (GPUs)
- Imagine a giant library filled with books. This library represents the massive amount of text and code that Large Language Models (LLMs) are trained on. To process all this information, LLMs need powerful computers called Graphics Processing Units (GPUs).

• Word vector embeddings
Imagine a vast library with countless books. Each book represents a word, and its content represents its meaning. Word vector embeddings are like maps that help us navigate this library by representing each word as a point in a high-dimensional space.

• Transformers
Imagine a grand banquet hall buzzing with conversations. Everyone is talking simultaneously, yet you can still understand your neighbor and contribute to the discussion. This is akin to how Transformer models work - they navigate complex relationships within language, allowing them to understand and generate human-quality text.

Training data bias: LLMs are trained on massive datasets of text and code, which can reflect existing societal biases and prejudices. This can lead to the model generating outputs that are discriminatory, offensive, or harmful.
Creativity and innovation: While LLMs can generate creative text formats, they struggle with truly original ideas and innovative solutions. Their outputs are often derivative and lack the human touch of genuine creativity.
Limited understanding: LLMs are good at processing and generating text, but they often lack true understanding of the meaning and context. This can lead to misinterpretations, factual errors, and misleading information.

According to previous chatgpt models -- 3000 to 4000

Science is a process through which the world can
be systematically probed.
• Scientists gather facts and search for consistent
relationships between measurable quantities.
• A problem: the boundary between science and non-
science is not always clear.


a scientific theory involves hypotheses that are
falsifiable. That is, there is some definite hypothetical result that,
if obtained, would show the hypothesis to be false

In science, a model is a simplified representation of a system or phenomenon. It helps us understand, explain, and predict how things work in the real world.

To represent complex systems: Models can help us visualize and understand complex systems that would be difficult to study directly. For example, a climate model can help us understand how changes in greenhouse gas emissions might affect the Earth's climate.
To generate predictions: Models can be used to predict what would happen under different circumstances. This allows scientists to test their hypotheses and refine their understanding of a system. For example, a model of a drug's interaction with the body can be used to predict its potential side effects.
To identify gaps in knowledge: Models can help us identify areas where we need more research. When a model's predictions do not match the real world, it suggests that we need to refine our understanding of the system.
To communicate ideas: Models can be used to communicate complex scientific ideas to a wider audience. For example, a model of the solar system can help students understand how the planets orbit the sun.

Imagine you're a teacher trying to explain how publishing only positive results can lead to misinformation to young students. Here's a simplified version:

Think of science as a group of friends trying to figure out how the world works. They do experiments and collect data, just like kids do in school.

Sometimes, their experiments work out and they find something cool. They're excited and want to tell everyone! This is like publishing positive results.

But sometimes, their experiments don't work out. They might feel discouraged, but it's still important to learn from their mistakes. This is like publishing negative results.

If they only share their successes and hide their failures, it's like telling a half-true story. People might think everything is always perfect, which isn't true.

Bem supposedly found evidence of precognition.

Problems:
High p-hacking risk: Bem's experiments had a high risk of "p-hacking," which involves selectively reporting data that supports the desired outcome and disregarding non-significant results.
Multiple comparisons: Conducting multiple analyses without adjusting for multiple comparisons increases the risk of false positive results.
Lack of blinding: Experimenters were not blinded to the condition participants were assigned, potentially introducing bias.
Questionable analysis methods: Bem used questionable statistical methods like "optional stopping" and "file-drawer analysis," which can inflate the likelihood of finding statistically significant results by chance.
Lack of replication: Subsequent attempts to replicate Bem's findings by other researchers have largely failed, casting doubt on the original results.

Statistical errors: Misuse of statistical tests and inadequate adjustments for multiple comparisons can significantly increase the risk of false positive results. This can lead to the mistaken acceptance of spurious findings as genuine scientific discoveries.
Lack of transparency and replication: Many psychological studies lack transparency in their methods and data analysis, making it difficult to assess their validity and reproducibility. Additionally, the emphasis on novel findings often overshadows the importance of replication, hindering the confirmation and accumulation of reliable knowledge.
Overreliance on null-hypothesis testing: The traditional focus on null-hypothesis testing can lead to overlooking potentially interesting and meaningful effects that fail to reach statistical significance. This can stifle scientific curiosity and limit the exploration of new ideas.
Lack of open science practices: The reluctance to share data, code, and materials openly hinders collaboration, transparency, and verification of research findings. This can slow down scientific progress and limit the potential impact of research

1. Pressure to publish: In academia, researchers face pressure to publish frequently in high-impact journals to secure funding, promotions, and tenure. This pressure can create a bias towards publishing positive results that are more likely to be accepted by journals, even if they are not as robust or reliable.

2. Publication bias: Journals tend to favor research with statistically significant results, leading to a bias towards publishing positive findings and neglecting negative or null results. This can create an incomplete and misleading picture of the scientific literature, as the true effect sizes and uncertainties may be underestimated.

3. P-hacking: The temptation to manipulate data or analysis techniques to achieve statistically significant results (p-hacking) can inflate the likelihood of publishing positive findings, regardless of their true validity. This practice undermines scientific integrity and leads to unreliable conclusions.

4. Lack of emphasis on replication: Replication studies, which aim to confirm previous findings, are often undervalued and receive less funding and recognition. This lack of focus on replication can lead to the persistence of non-reproducible findings that cannot be confirmed by independent researchers.

Ultimately, the "brain in a vat" argument serves as a reminder of the limitations of human understanding and the vast mysteries that remain to be explored. Whether we are brains in vats or not, it encourages us to think critically about our own perceptions and the nature of reality.

Check this

The extended cognition argument proposes that cognition extends beyond the brain and into the physical environment. It argues that our mental processes are not solely confined within our skulls but are actively shaped and supported by the external world. This argument challenges the traditional view of the mind as a self-contained entity and suggests a more dynamic and interactive relationship between our minds and our surroundings.

Most PhDs, especially in STEM fields, are funded.
• This means that most students pay no tuition.
• Students usually receive stipends, paid either through grants, or by working as
teaching assistants