Artificial intelligence

Question 1

Describe the Turing test

Answer 1

Alan Turing is a founder of modern computers, and he asked the question can machines think? The computer is deemed intelligent, when a human interrogator can’t tell it, apart from a real human, will be imitation game

The interrogator has to guess the gender of the actors behind the curtain .  the man behind the curtain, must trick the interrogator, and if the machine is able to trick the interrogator, as well as a man, the machine is considered intelligent, to be intelligently machine, must guess like the man would guess 

Question 2

Describe the Chinese room (1980)

Answer 2

 it challlegged whether the Turing test is a valid test for intelligence.

John Searle propose that guy in the Chinese room can guess the correct translations of Chinese sentences without understanding Chinese.  in the scenario the computer is the man in the room and questions where the computer can actually use human intelligence, even though it’s just using human language. 

Question 3

True or false, chat GPT, understands what it’s saying

Answer 3

False,

It is just a language model with no prior experience, like humans have

Question 4

 describe how ChatGPT works

Answer 4

CPT is a large language model (LLM)  which is based on machine, learning techniques called neural network networks

Question 5

What is a neural network?

Answer 5

Neural network is a computational framework to model human cognition

Question 6

What is parallel, distributed processing?

Answer 6

A computational framework to model cognition, which differed from other approaches tear III, as it is based on massive connections among like neurons in the brsi.

Question 7

What is convolutional neural network?

Answer 7

CNN is a massive deep, neural network that is only made possible by using parallel computing

Question 8

What is the difference between traditional neural networks and deep learning?

Answer 8

The number of layers and the amount of data used to train the models.

Question 9

What is the difference between  call evolutional, neural network and large language models?

Answer 9

CNN is good at processing images and visual information beyond the human level and LLM is good at processing languages (like AI) 

Question 10

Define badly and hitch, working memory model

Answer 10

Depending on the type of information that comes in, they are processed differently. Visual memory travels through the central executive and to the visual spatial sketchpad and languageL Words is processed by travelling through the phonological loop.

Question 11

How does convolutional neural network and large language model relate to badly hitch model

Answer 11

 They are like the computer version, so CNN would be lichen to the visual spatial sketchpad and LLM would be licensed to the phonological loop

However, the computer doesn’t have a essential executive yet 

Question 12

What makes humans unique compared to AI

Answer 12

1. Self-awareness.
2. Emotion
3. Creativity.
4. Adoptability.
5. Intuition.
6. Morality.
7. Social graces
8. Physical body 

Question 13

 when can we trust AI?

Answer 13

Accuracy, bias, controllability, dependability, explainability, predictability, reliability, privacy, safety

Question 14

According to Oxsann et al, 2020 what kind of people trust AI, more or less?

Answer 14

People with higher openness, trust more, but people with high conscientiousness AI less 

Question 15

Is Philly automated I technology, trusted more or less than human AI combo?

Answer 15

Less

Question 16

In what ways can AI help us?

Answer 16

1. Pattern recognition
2. Big data.
3. Predictive modelling.
4. Automation (e.g. Tesla
5. Chat, bpt and Google bars
6.  augmented reality such as Pokémon GO
7. Virtual therapy and diagnosis for mental health, such as Eliza which is a language processing AI for the firing test which is turned into a psychotherapist, and was the first model to pass the Turing test. Also limbic access is AI used by the NHS with 93% symptom accuracy 

Question 17

What are some of the risks associated with AI?

Answer 17

1. Misdiagnosis - however, limbic access has low risk of this
2. Adversarial attacks such as hacking.. e.. taskrsbbit.
    3. Jailbreaking on LLM Network e.g. Using AI for building bombs.
3. Privacy.  as Hospital , intelligence, LLM, like Chi gpt uses peoples data on the Internet to form a new data output.
4. Risk of bias such as 1. Lack of representation as AI is trained with mainly white male faces, 2. Spurious correlation meaning there is a correlation, but no real reason for this correlation . 

Question 18

What is an example of representational bias in AI

Answer 18

 CAD, an artificial intelligence system under represented women of minority groups which skewed the level of diagnosis. There was low accuracy for black patients than white patients.

Question 19

Given an example of bias, spurious, correlation for artificial intelligence

Answer 19

I was trying to develop AI to shortlist candidates, but it was scrapped due to a bias against women as the AI saw as spurious correlation

Question 20

What is brain computer interface?

Answer 20

Mind control, gaming, and EEG

Question 21

Give examples of brain computer interface

Answer 21

BCI has been used with paralysed patient (kobler 2001) 

A visual imagery during sleep with MRI predicted object categories in (horikawa 2013)