Podcast
Questions and Answers
How can AI encode bias, as exemplified by the Dall-E image generation model?
How can AI encode bias, as exemplified by the Dall-E image generation model?
- By reinforcing stereotypes present in the training data. (correct)
- By randomly generating images without any specific pattern.
- By prioritizing creative and imaginative content over realistic portrayals.
- By accurately reflecting the diversity present in society.
What is the primary distinction between AI and ML based on the lecture?
What is the primary distinction between AI and ML based on the lecture?
- ML involves systems with goals that make decisions, while AI focuses on data extraction.
- ML extracts knowledge from data to build models, while AI encompasses any system that mimics intelligence. (correct)
- AI is a subset of ML focused on complex problem-solving, whereas ML is a broader field.
- AI extracts knowledge from data to build models, while ML mimics intelligence directly.
Which of the following best describes the role of 'labels' in machine learning?
Which of the following best describes the role of 'labels' in machine learning?
- They outline the ethical guidelines for the model’s application.
- They are used to define the algorithms to be used in the model.
- They provide a description of the data collection process.
- They implicitly tell the model what question to answer and what the correct answers are. (correct)
What is the role of data in machine learning models, according to the lecture?
What is the role of data in machine learning models, according to the lecture?
An ML model is presented with an image of an animal and outputs 'dog'. What type of ML model is it, and what kind of label does it use?
An ML model is presented with an image of an animal and outputs 'dog'. What type of ML model is it, and what kind of label does it use?
In the context of machine learning, what does it mean for a model to 'generalize' well?
In the context of machine learning, what does it mean for a model to 'generalize' well?
Which statement accurately describes the utilization of decision trees in machine learning?
Which statement accurately describes the utilization of decision trees in machine learning?
What is a primary advantage of using decision trees in machine learning?
What is a primary advantage of using decision trees in machine learning?
What does the 'Bag of Words' representation primarily aim to achieve in natural language processing?
What does the 'Bag of Words' representation primarily aim to achieve in natural language processing?
In the context of count-based models in NLP, what does the term 'near' typically refer to when incrementing the count in each cell?
In the context of count-based models in NLP, what does the term 'near' typically refer to when incrementing the count in each cell?
What is the advantage of updating ML models versus expert systems when inaccuracies are found?
What is the advantage of updating ML models versus expert systems when inaccuracies are found?
Which of the following is a task that can be approached using supervised learning?
Which of the following is a task that can be approached using supervised learning?
How do Large Language Models (LLMs) predict the next word in a sentence?
How do Large Language Models (LLMs) predict the next word in a sentence?
In the context of supervised learning, what is the source of the data's labels used to train the models?
In the context of supervised learning, what is the source of the data's labels used to train the models?
Why is it important to be careful about bias in data used for AI systems?
Why is it important to be careful about bias in data used for AI systems?
Which of the following tasks is best best described as AI but not ML?
Which of the following tasks is best best described as AI but not ML?
Which of the following tasks is best described as both AI and ML?
Which of the following tasks is best described as both AI and ML?
Which of the following best describes the relationship between data and ML model?
Which of the following best describes the relationship between data and ML model?
How do you input a picture into an ML model and what type of model does it use?
How do you input a picture into an ML model and what type of model does it use?
What is the purpose of labels in ML model?
What is the purpose of labels in ML model?
What is the difference(s) between Classification and Regression?
What is the difference(s) between Classification and Regression?
What is unsupervised learning?
What is unsupervised learning?
What did LLMs learn?
What did LLMs learn?
According to the video, what is one of the most basic paradigms in ML?
According to the video, what is one of the most basic paradigms in ML?
ML models are better at the examples we give it, what is the name of it?
ML models are better at the examples we give it, what is the name of it?
Decision trees are what type of machine to use to separate the data?
Decision trees are what type of machine to use to separate the data?
What will happen if there is more data to create ML models to do a prediction?
What will happen if there is more data to create ML models to do a prediction?
What happened if the temperature is -5, with Snow, on Wednesday wearing casual, what will the label show?
What happened if the temperature is -5, with Snow, on Wednesday wearing casual, what will the label show?
What are decision trees?
What are decision trees?
Which of the following best describes the AI system MYCIN?
Which of the following best describes the AI system MYCIN?
What is needed when the AI system is inaccurate?
What is needed when the AI system is inaccurate?
The weather man forecast the weather as raining with 100 percentage and thunderstorm for Tuesday at 10:00AM, what are the category and values?
The weather man forecast the weather as raining with 100 percentage and thunderstorm for Tuesday at 10:00AM, what are the category and values?
Which statement is true regarding data and ML?
Which statement is true regarding data and ML?
What does count-based models mean for NLP?
What does count-based models mean for NLP?
If you provide a cat list of pictures and a dog list of pictures. Then the ML model shows a dinosaur. What do you call that?
If you provide a cat list of pictures and a dog list of pictures. Then the ML model shows a dinosaur. What do you call that?
What aspect of AI systems requires careful attention to prevent biased outputs?
What aspect of AI systems requires careful attention to prevent biased outputs?
What term describes the concept of data in machine learning?
What term describes the concept of data in machine learning?
In the machine learning pipeline, what is the purpose of the 'Training' stage?
In the machine learning pipeline, what is the purpose of the 'Training' stage?
In a machine learning context, what is the role of 'labels'?
In a machine learning context, what is the role of 'labels'?
What is the primary objective of converting text into numerical representations in NLP?
What is the primary objective of converting text into numerical representations in NLP?
In count-based models for NLP, what does incrementing the count in a cell signify?
In count-based models for NLP, what does incrementing the count in a cell signify?
What is a key characteristic of supervised learning?
What is a key characteristic of supervised learning?
What is the practical implication of the statement that an ML model is an 'imitation machine'?
What is the practical implication of the statement that an ML model is an 'imitation machine'?
What is the core difference between how an expert system and a supervised learning system provide answers?
What is the core difference between how an expert system and a supervised learning system provide answers?
Why are ML models easier to update when inaccuracies are found, compared to expert systems?
Why are ML models easier to update when inaccuracies are found, compared to expert systems?
What are the main components to consider when determining the type of transportation a person is using?
What are the main components to consider when determining the type of transportation a person is using?
What does it mean for a machine learning model to 'generalize' well?
What does it mean for a machine learning model to 'generalize' well?
In the context of the YouTube video recommendation system, what serves as the 'input' for the ML model?
In the context of the YouTube video recommendation system, what serves as the 'input' for the ML model?
In Google Lens, what is the primary role of machine learning?
In Google Lens, what is the primary role of machine learning?
What best describes the role of data in ChatGPT?
What best describes the role of data in ChatGPT?
Consider the statement: 'I had to take my _____ to the vet.' What concept does this illustrate in language models?
Consider the statement: 'I had to take my _____ to the vet.' What concept does this illustrate in language models?
In Large Language Models (LLMs), what is the purpose of the 'context window'?
In Large Language Models (LLMs), what is the purpose of the 'context window'?
Given the data of temperature, precipitation, day and clothes; What type of ML is used to produce the output (label)?
Given the data of temperature, precipitation, day and clothes; What type of ML is used to produce the output (label)?
How would you describe passive imitation?
How would you describe passive imitation?
What does it mean for ML to be great explainability?
What does it mean for ML to be great explainability?
When should we be cautious of bias when training data?
When should we be cautious of bias when training data?
When is assignment #2 due?
When is assignment #2 due?
What does Building blocks of ML consists of?
What does Building blocks of ML consists of?
Why does 3 submissions exist for assignment #2?
Why does 3 submissions exist for assignment #2?
Where did the labels/answers come mostly from in Supervised Learning?
Where did the labels/answers come mostly from in Supervised Learning?
Which of these applications of AI is also ML?
Which of these applications of AI is also ML?
What does Data consists of?
What does Data consists of?
How to you input a picture into an ML model?
How to you input a picture into an ML model?
What does classify do regarding the types of news article?
What does classify do regarding the types of news article?
What are the amount of options when classifying a review type of postive or negative?
What are the amount of options when classifying a review type of postive or negative?
What does 'Large' stand for in LLM?
What does 'Large' stand for in LLM?
In Count Based Models, what does 'near' mean?
In Count Based Models, what does 'near' mean?
Which of this option will the statement refer to 'Cats will Escape'?
Which of this option will the statement refer to 'Cats will Escape'?
Given the temp is high, predict the label. Temp > 0
Given the temp is high, predict the label. Temp > 0
What does Generalize mean related to ML?
What does Generalize mean related to ML?
What is a primary limitation of a perceptron?
What is a primary limitation of a perceptron?
What distinguishes generative learning from supervised learning?
What distinguishes generative learning from supervised learning?
Which factor does not significantly influence the decision-making when choosing between different ML models?
Which factor does not significantly influence the decision-making when choosing between different ML models?
What is the primary advantage of using generative learning with unlabeled data?
What is the primary advantage of using generative learning with unlabeled data?
In the context of Large Language Models (LLMs), what is the purpose of the 'context window'?
In the context of Large Language Models (LLMs), what is the purpose of the 'context window'?
What characteristic defines sequence generation models, such as those used for video generation?
What characteristic defines sequence generation models, such as those used for video generation?
In Generative Adversarial Networks (GANs), what is the role of the 'discriminator'?
In Generative Adversarial Networks (GANs), what is the role of the 'discriminator'?
Why is it important for generative models to have a large amount of data available?
Why is it important for generative models to have a large amount of data available?
How do decision trees classify?
How do decision trees classify?
What is the goal of the generator in Generative Adversarial Networks (GANs)?
What is the goal of the generator in Generative Adversarial Networks (GANs)?
What best describes the modelling power of the Artificial Neural Network?
What best describes the modelling power of the Artificial Neural Network?
What could generative models help application(s) related to?
What could generative models help application(s) related to?
In the three models; Perceptron, Decision Tree, and Artificial Neural Network; Which of these models has the simplest interperability?
In the three models; Perceptron, Decision Tree, and Artificial Neural Network; Which of these models has the simplest interperability?
What does a LLM (Large language Model) look at to predict the next word?
What does a LLM (Large language Model) look at to predict the next word?
Which action does unsupervised learning perform?
Which action does unsupervised learning perform?
Why is the output of a generative learning randomized?
Why is the output of a generative learning randomized?
What is meant by needing more than lines?
What is meant by needing more than lines?
With the 3 models; Perceptron, Decision Tree, and ANN. If the interpretability is poor, what model will that be?
With the 3 models; Perceptron, Decision Tree, and ANN. If the interpretability is poor, what model will that be?
Generative learning commonly has...
Generative learning commonly has...
When would a statement like, 'This is a big deal!' be said?
When would a statement like, 'This is a big deal!' be said?
What is the risk of training the models recursively on data that has been generated?
What is the risk of training the models recursively on data that has been generated?
What functions are ANNs used for?
What functions are ANNs used for?
Is it possible to turn our simple decision tree into a 2d plot?
Is it possible to turn our simple decision tree into a 2d plot?
What two networks are used in GAN?
What two networks are used in GAN?
Minksky's famous book did what to perceptrons?
Minksky's famous book did what to perceptrons?
Why do we want the model to be creative with a dice?
Why do we want the model to be creative with a dice?
What does Generative Learning need to generate new images?
What does Generative Learning need to generate new images?
What should happen that tells us to stop?
What should happen that tells us to stop?
What is the difference between supervised learning and generative learning?
What is the difference between supervised learning and generative learning?
How does Chat GPT generate?
How does Chat GPT generate?
If chatGPT outputs something that is wrong, will we need to update the answers?
If chatGPT outputs something that is wrong, will we need to update the answers?
Which model can learn complex non-linear functions?
Which model can learn complex non-linear functions?
What is a characteristic of sequence generation models?
What is a characteristic of sequence generation models?
What is the purpose of Data?
What is the purpose of Data?
Is AI in creative activities?
Is AI in creative activities?
Which of this option is right?
Which of this option is right?
What is a key characteristic that distinguishes generative learning from supervised learning?
What is a key characteristic that distinguishes generative learning from supervised learning?
What is the primary goal of the generator network in a Generative Adversarial Network (GAN)?
What is the primary goal of the generator network in a Generative Adversarial Network (GAN)?
Why is it often necessary for generative models to be trained with a large amount of data?
Why is it often necessary for generative models to be trained with a large amount of data?
In the context of Large Language Models (LLMs), like ChatGPT, what is the role of the 'context window'?
In the context of Large Language Models (LLMs), like ChatGPT, what is the role of the 'context window'?
What is the main purpose of the discriminator in a Generative Adversarial Network (GAN)?
What is the main purpose of the discriminator in a Generative Adversarial Network (GAN)?
What is a key limitation of a perceptron in machine learning?
What is a key limitation of a perceptron in machine learning?
What is meant by needing more than lines in machine learning?
What is meant by needing more than lines in machine learning?
What is the key factor that sets generative learning apart, making it scalable and useful?
What is the key factor that sets generative learning apart, making it scalable and useful?
What prompted skepticism and a subsequent decline in the popularity of research on perceptrons in the late 1960s?
What prompted skepticism and a subsequent decline in the popularity of research on perceptrons in the late 1960s?
In what context might the statement 'This is a big deal!' be said?
In what context might the statement 'This is a big deal!' be said?
What are Artificial Neural Networks (ANNs) most well known for?
What are Artificial Neural Networks (ANNs) most well known for?
What is the significance of the statement 'Sutton's Bitter Lesson'?
What is the significance of the statement 'Sutton's Bitter Lesson'?
If the interpretability score is poor, what model will that be?
If the interpretability score is poor, what model will that be?
Can you always turn our simple decision tree into a 2d plot?
Can you always turn our simple decision tree into a 2d plot?
Why is the output of a generative learning model randomized?
Why is the output of a generative learning model randomized?
If it wasn't possible to separate XOR function, then what happened?
If it wasn't possible to separate XOR function, then what happened?
What are two networks that are used in GAN?
What are two networks that are used in GAN?
How and why do you stop a generative learning?
How and why do you stop a generative learning?
Why is the Generator helpful to use?
Why is the Generator helpful to use?
If a Cat picture is given to the model and cat picture is generated, what is this?
If a Cat picture is given to the model and cat picture is generated, what is this?
A pathology foundation model is what type of example?
A pathology foundation model is what type of example?
ChatGPT prompt: Tell me a joke that I can use in a class that teaches Al to the public. ChatGPT generates a joke. What is it?
ChatGPT prompt: Tell me a joke that I can use in a class that teaches Al to the public. ChatGPT generates a joke. What is it?
What is an accurate description regarding the model's output in generative learning systems?
What is an accurate description regarding the model's output in generative learning systems?
What action does unsupervised learning perform on the training dataset?
What action does unsupervised learning perform on the training dataset?
What does a Large Language Model look at to predict the next word?
What does a Large Language Model look at to predict the next word?
What should be considered regarding the topic of Recursive Training?
What should be considered regarding the topic of Recursive Training?
When considering bias with generative AI, what action should a user take?
When considering bias with generative AI, what action should a user take?
Flashcards
What is AI?
What is AI?
Anything that mimics intelligence
What is Machine Learning (ML)?
What is Machine Learning (ML)?
A subset of AI focusing on extracting knowledge from data to build models for predictions
What is bias in ML?
What is bias in ML?
ML models may encode societal biases if not carefully addressed.
What is Data?
What is Data?
Signup and view all the flashcards
What is ML Training?
What is ML Training?
Signup and view all the flashcards
What is ML Prediction?
What is ML Prediction?
Signup and view all the flashcards
ML labels
ML labels
Signup and view all the flashcards
ML Classification
ML Classification
Signup and view all the flashcards
ML Regression
ML Regression
Signup and view all the flashcards
ML sequence
ML sequence
Signup and view all the flashcards
Bag of Words
Bag of Words
Signup and view all the flashcards
Count Based Models
Count Based Models
Signup and view all the flashcards
Generalization
Generalization
Signup and view all the flashcards
Decision Trees
Decision Trees
Signup and view all the flashcards
Supervised Learning
Supervised Learning
Signup and view all the flashcards
ML Data
ML Data
Signup and view all the flashcards
ML Model
ML Model
Signup and view all the flashcards
ML Strengths and Weaknesses
ML Strengths and Weaknesses
Signup and view all the flashcards
Modelling Power
Modelling Power
Signup and view all the flashcards
Data Requirements
Data Requirements
Signup and view all the flashcards
Trainability
Trainability
Signup and view all the flashcards
Interpretability
Interpretability
Signup and view all the flashcards
Linear Separability
Linear Separability
Signup and view all the flashcards
Perceptrons (1969)
Perceptrons (1969)
Signup and view all the flashcards
Generative Learning
Generative Learning
Signup and view all the flashcards
Choosing a Model
Choosing a Model
Signup and view all the flashcards
Dall-E 3
Dall-E 3
Signup and view all the flashcards
Context Window
Context Window
Signup and view all the flashcards
Generative Learning Benefit
Generative Learning Benefit
Signup and view all the flashcards
GANs (Generative Adversarial Networks)
GANs (Generative Adversarial Networks)
Signup and view all the flashcards
Sequence generation
Sequence generation
Signup and view all the flashcards
Study Notes
- AI systems can encode biases if not taken care of.
- Dall-E, an AI image generation model, produces images reflecting stereotypes of successful people as white, male, young, dressed in Western business attire, working in urban offices, and having common hairstyles.
Announcements
- Assignment #2 is live and due Next Tuesday, with three possible submissions.
- Three submissions are allowed, to mitigate accidental submissions, internet outages, etc.
- There will be no feedback provided upon submitting.
- There will be no feedback provided upon submitting.
- Asignment #2 is due tonight
- Asignment #1 grades have been released
- Asignment #3 opens tonight
- Asignment #3 is due the following Tuesday and will be like A2. It will be a multiple choice quiz in Python
Prerequisites and Learning Objectives
- From last lecture, learn the building blocks of machine learning, data and models.
- Data is the information used to train the model, like text and images
- The model is a computer program that processes input and creates output that's usually a number or a collection of numbers.
- Supervised learning learns to imitate, and learns to predict the label of data, which are created by human labelers
- Key learning objectives for this content include:
- Listing the strengths and weaknesses of different models.
- Distinguishing between generative and supervised learning
- Listing examples of generative learning systems and what they are used for
Machine Learning Building Blocks
- Machine learning has 2 building blocks, data and models
- Supervised learning is the basic paradigm
- The goal is to list and describe the building blocks of ML systems
- The goal is to identify problems that can be thought of as supervised learning
AI Caution, Fear, Excitement
- “Success in creating effective AI, could be the biggest event in the history of our civilization. Or the worst. We just don't know. So we cannot know if we will be infinitely helped by AI, or ignored by it and side-lined, or conceivably destroyed by it,” ~Stephen Hawking
- “The rise of AI will free people up to do things that software never will—teaching, caring for patients, and supporting the elderly, for example.” ~Bill Gates
- “AI is far more dangerous than nukes.” ~Elon Musk
- There are three reasons why people fear AI
- Cynicism is the belief that it is rational not to cooperate
- Humanism/racism is systematic bias against machines, denial of their potential moral worth and personhood
- Conservatism is the fear of change, fear of the other tribe
- None of these fears reflect well on those who hold them- Rich Sutton (UoA)
AI Versus ML
- AI mimics intelligence, it includes systems with goals that make decisions.
- ML is a subset of AI where knowledge is extracted from data to build predictive models.
- Expert system MYCIN is given as an example of AI without ML.
- A simulated path-finding robot is another example of AI, not ML.
- Spam filters, YouTube recommendations, Google Lens, and ChatGPT are examples of both AI and ML.
- ML is at the core of most modern AI systems.
Deciding which models to use
- All basic models have different strengths and weaknesses
- Modelling Power is whether the model can learn complex patterns
- Data requirements is the amount of data needed to train the model
- Trainability is whether it is difficult to train, and may require ML Expertise
- Interpretability is understanding what the model learns
Data as a building block
- In ML, data trains models to distill information and knowledge.
- Data is a collection of discrete or continuous values conveying information, quantity, quality, facts, and statistics.
Course Map
- The topics in this deck will include the Machine Learning Building Blocks 2 as well as Generative AI Systems, and choose a model.
Lecture Topics
- This lecture will cover how to choose a model, and another important type of paradigm in MI: Generative Learning
- Generative Learning will be broken down into Imitating Data, generative learning systems (Dall-E), Generative Adversarial Networks, and Sequence Generation Models (ChatGPT)
- There will also be discussion of the power of scaling with data and computation over scaling with people.
Machine Learning Pipeline
- The steps are training and prediction.
- Training extracts info/knowledge from data to build the model.
- Prediction uses the model to answer an input query.
- Labels in ML examples provide the model with the question it should answer.
- The label is the correct answer/output/prediction for a given input.
YouTube Video Recommendation example
- For YouTube video recommendations, input data might include user ID, current viewing video, watch history, liked videos, and disliked videos.
- The output is a list of top-K recommended videos
Google Lens example
- The input is an image like a 2x2 Rubik's cube,
- The output provides links to online stores selling similar items via object detection
MNIST example
- MNIST, a commonly used ML dataset for research, consists of handwritten digits from 0 to 9,
- It's used in real-world applications to read ZIP codes in postal services and check amounts in bank accounts.
Predicting text articles and movie reviews
- Language tasks include classifying news articles by topic, classifying movie reviews as positive or negative.
- There is also text generation based on input text.
Language Data
- Before Large Language Models, ML focused on language, labeled "Natural Language Processing" (NLP).
- Large Language Models are huge Neural Nets with Billions of connections & neurons trained mostly on text data.
- They then output sentences
Bag of Words
- The "Bag of Words" representation converts text to numbers for processing in ML models.
- The goal is to provide calculations
- Word order and importance can matter in language data.
- "Near" can have many word associations
Count-Based Models and Vector Creation:
- Count-based models convert words to numbers based on relationships, building a data table with rows and columns for each word.
- We increment counts based on observed proximity, creating a vector representing word relationships.
- This list of numbers creates a number vector.
- Count-based models can show how word meanings shift over time.
Machine Learning Models
- The model output is a number or a collection of numbers.
- Classification models output a category, or a discrete value.
- Regression models output some value that is a continuous value of the input.
- Models answer questions based on input data
Classification Model
- In classification, the output is a category, or a discrete label.
- Examples include digit recognition (0-9), object detection (cube), and sentiment analysis (positive/negative).
Regression Model
- Regression models work give output with a continuous value of the input, such as scoring digit writing, pricing, or rating a movie.
Model Output
- The output of these models can be a variety of categories and values
Chat GPT
- In ChatGPT data refers to the information used to train the model, like large collections of text from the internet, books, and other written sources
Machine Translation
- Machine translation involves generating text output based on a given input.
Large Language models
- LLMs convert large amounts of text into manageable word vectors and use context windows to predict the next word.
Attention
- LLMs learn to focus on key words through a process, "attention", improving understanding.
Supervised Learning
- Supervised learning, an ML framework, learns from labeled data where each input pairs with a label or correct answer.
- The labels provide descriptive tags or values.
- Supervised learning takes direction from the people selecting the answers
Categories
- Labels in classification are categories, like categories for digits, objects, or sentiment.
Regression
- Labels in regression are numerical values, like scores, prices, or ratings.
- Supervised learning models have no agency and cannot be better than the data they are given.
Imitation
- Supervised learning systems and expert systems are imitation machines.
- ML Models are much easier to update.
Features
- Features, like temperature, precipitation, day of the week, and clothing, are numbers or categorical data inputted into the model.
- Models should generalize predictions on new data.
Generalization
- Generalization makes predictions on new unseen data
Decision Trees
- To get around this, decision trees, a classical machine learning model, can be implemented to find trends in data
- In this model temperature is noted against the data.
- They can be used for both classification and regression.
Decision Trees Pros and Cons
- Decision trees are noted as having high explainability but struggle against higher, more complex data implementation and learning
- It is important to remember how and when the systems work for this, and those of you in INT-D 161 will use powerful ML models
Other Learning
- Supervised learning is the most basic paradigm in ML but there are other learning systems.
- Other learning systems include Generative and Reinforcement Learning.
Lecture Topic Summaries
- Data and models are crucial ML building blocks.
- Data encoding depends on data type.
- Supervised learning imitates human answers.
- Decision trees are explainable classical models.
Sample Basic Models
- Artificial Neural Network: The model is composed of layers of interconnected nodes, or neurons, that process and transmit information, it is comprised of an Input Layer, Hidden Layers, and an Output Layer
- Decision Tree: A decision tree is a structured model that uses a series of binary decisions to classify or predict outcomes based on input features or attributes, it navigates based on conditions like temperature or weather, it then derives certain conditions
- Perceptron: A perceptron is a single-layer neural network that performs binary classification by applying weights to input features and using a threshold to make a decision, it includes inputs, a Heavyside Step-function, and an output
Studying That Suits You
Use AI to generate personalized quizzes and flashcards to suit your learning preferences.