How AI Works Ch1 - And Away We Go

Question 1

What is the purpose of the book?

Answer 1

The book, How AI Works, teaches how AI works at the conceptual level without getting bogged down with complicated mathematics.

Question 2

What is the purpose of the chapter, ‘And Away We Go: An AI Overview’?

Answer 2

This chapter aims to clarify what AI is and its relationship to machine learning and deep learning.

Question 3

Rewrite in Your Own Words:
. “Artificial intelligence attempts to coax a machine, typically a computer, to behave in ways humans judge to be intelligent.”
. “Artificial intelligence is the field of computer science that involves mimicking human intelligence in machines.”
. “AI includes machine learning, which includes deep learning: AI > machine learning > deep learning.”

Answer 3

Artificial Intelligence… is machine activity… that humans perceive as intelligence.
Artificial Intelligence… is behavioral replication of human intelligence by way of computer software in any machine.
Deep Learning… is a type of Machine Learning, which is a type of artificial intelligence.
…
Artificial Intelligence is a machine that is engineered in such a way that it can be evaluated as behaving intelligently, such as a computer with AI.

Question 4

Answer: How is Machine Learning and Deep Learning related to each other and to Artificial Intelligence?

Answer 4

Machine Learning is one of many sub-fields of Artificial Intelligence. Deep Learning is one of many sub-fields of Machine Learning.

Question 5

Rewrite: “Machine learning conditions models like random forests, support vector machines, and neural networks to a particular dataset so that the conditioned model can accurately predict class labels or numeric values when given new, unknown inputs.”

Answer 5

Machine Learning…
…
Machine Learning conditions computer models based on KNOWN DATA to gain the ability to correctly behave according to the conditioned model on UNKNOWN DATA.
…
Computer models for Machine Learning include:
. Neural Networks
. Random Forest
. Support Vector Machines

Question 6

Rewrite: “Deep Learning is the sub-field of machine learning that uses large neural networks with many layers. Deep learning appeared around 2012, with the advent of large convolutional models with dozens to even hundreds of layers. Before the advent of deep learning, such models could not be reliably trained.”

Answer 6

Deep Learning…
…
Deep Learning is Machine Learning that conditions a specific computer model called Large Neural Networks that contain multiple layers.
Deep Learning appeared around 2012 with the arrival of Large Convolutional Models, that is a Large Neural Network that contains dozens or even hundreds of layers.
…
Research: What are these ‘layers’ of a Large Neural Network?
Research: What are the names of other Large Neural Networks created since Large Convolutional Models?

Question 7

Research: prominent computer scientist John McCarthy and his contribution to Artificial Intelligence.

Answer 7

John McCarthy co-authored the paper that coined the term ‘Artificial Intelligence.’
. The paper is a proposal, dated in August 31, 1955, to meet during the summer of 1956 at Dartmouth College to discuss AI.
. Read: https://web.archive.org/web/20070826230310/http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html
Research Other Accomplishments:
. Created Lisp Programming Language.
. Created the first Garbage Collection for Lisp.
. Was awarded 1971 Turing Award.
Research Publications:
. Read: https://web.archive.org/web/20131011125002/http://www-formal.stanford.edu/jmc/

Question 8

Rewrite: “Computers are programmed to carry out a particular task by giving them a sequence of instructions, a program, which embodies an algorithm. An Algorithm is a sequence of steps to accomplish a task; a recipe. Machine learning models implement algorithms.”

Answer 8

Machine Learning models process according to algorithms by way of programs. Algorithms are a step by step procedure that leads to a goal. Programs are instructions for the computer used to follow the Algorithms.

Question 9

Rewrite: “The program embodies the algorithm in a way that the workers (the machine) can follow step by step.”

Answer 9

A program’s instruction for the computer is based on the step by step algorithm.

Question 10

Research: conceptual machines envisioned by Alan Turing in the 1930s.

Question 11

Research: 19th-century Analytical Engine of Charles Babbage.

Question 12

Research: Ada Lovelace often regarded as the world’s first programmer, who initially understood the far-reaching possibilities of what we now call a computer

Question 13

Research: “old-school AI, which includes the algorithms and approaches developed from the 1950s onward. Old-school AI is not what people currently mean when discussing AI.”

Question 14

Rewrite: “Machine Learning models are blank slates that data must condition to make them suitable for a task. If the data is bad, the model is bad.”

Answer 14

Machine Learning models are model templates that are conditioned according to a known dataset to make the conditioned model correctly perform intelligent behaviors. Conditioning a model template with bad data leads to a erroneously conditioned model.

Question 15

Rewrite: The model has parameters, which control the model’s output.
Rewrite: Parameters are a generic term for any quantity in a model that can be adjusted. Usually, the term “parameters” is used to refer collectively to the weights and biases of a neural network.

Answer 15

A parameter is an abstract term referring to any value that can be adjusted impacting a Machine Learning model’s output; usually a quantity used as weights and biases to produce a neural network output.

Question 16

Rewrite: “A machine learning model is a black box that accepts an input, usually a collection of numbers, and produces an output, typically a label like “dog” or “cat,” or a continuous value like the probability of being a “dog” or the value of a house with the characteristics given to the model (size, number of bathrooms, ZIP code, and so on).”

Answer 16

A Machine Learning model is a non-transparent entity that received known data inputs and provides meaningful output.

Question 17

Rewrite: “A Machine Learning model is generic term for any algorithm conditioned to a set of data by adjusting the parameters of the algorithm. A model might be a neural network or any other machine learning algorithm, like a random forest or a support vector machine.

Rewrite: “A Machine Learning model is an abstract notion of something that accepts inputs and generates outputs, where the inputs and outputs are related in some meaningful way.”

Rewrite: More abstractly, a model is “an intentional simplification of a complex situation designed to eliminate extraneous detail in order to focus on the essentials” (Daniel L. Hartl, A Primer of Population Genetics and Genomics [Oxford University Press, 2020]).”

Answer 17

A Machine Learning model is an algorithm conditioned according to a known data. The conditioning is done by changing the parameters of the model.

A Machine Learning model received known data as input and generates output that is meaningful to humans.

In general, a model is a simplified template of essential characteristics of a more complex conceptualization.

Question 18

Rewrite: Conditioning a Machine Learning model, known as training, seeks to set the model’s parameters in such a way that they produce the correct output for a given input.

Rewrite: Machine Learning Training is the act of conditioning the parameters of a Machine Learning model to a specific dataset or use case. What training entails depends on the form of the model, from virtually nothing (nearest neighbor classifiers) to incredible levels of computation (training a large language model like GPT-4). All machine learning models learn from the training dataset, making machine learning, including deep learning, an empirical exercise.

Answer 18

Machine Learning Training is the conditioning of a machine learning model given a known dataset, where the model’s parameters are changed to produce the correct output that is meaningful to humans.

In Machine Learning Training, the parameters of any machine learning model are conditioned according to a known dataset or use case. Every Machine Learning Model has a unique training and learn from a training dataset.

Question 19

Rewrite: Now we understand why the field is called machine learning: we teach the machine (model) by giving it data. We don’t program the machine; we instruct it.

Answer 19

Machine Learning is not programed. Machine Learning is instructed according to an algorithm for the machine to process according to a known dataset.

Question 20

Answer: what is Supervised Learning?

Answer 20

using known labeled data to train the model, this approach is called supervised learning

Question 21

Rewrite:
Here, then, is the Machine Learning algorithm for Supervised Learning:
1. Gather a training dataset consisting of a collection of inputs to the model and the outputs we expect from the model for those inputs.
2. Select the type of model we want to train.
3. Train the model by presenting the training inputs and adjusting the model’s parameters when it gets the outputs wrong.
4. Repeat step 3 until we are satisfied with the model’s performance.
5. Use the now-trained model to produce outputs for new, unknown inputs.

Answer 21

The abstract Supervised Learning process for any Machine Learning Algorithm:
1. Gather the training dataset, which is known inputs and known outputs.
2. Select an appropriate Machine Learning Model for training given the dataset of known inputs and known outputs.
3. Train the model by adjusting the Machine Learning Model’s parameters when the model does not provide the known output given the known input.
4. Repeat step 3 until the model’s performance is satisfactory.
5. Use the trained Machine Learning Model with unknown inputs to produce intelligent output.

Question 22

Answer: what is a multi-class model?

Answer 22

A Machine Learning model that categorizes outputs into multiple classes.

Question 23

Answer: what is a binary class model?

Answer 23

A multi-class Machine Learning model that categorizes outputs into exactly two classes.

Question 24

Rewrite: Models often want numeric class labels, which tells us that models don’t know what their inputs and outputs mean; they only make associations between sets of inputs and outputs. Models don’t “think” using any commonly accepted definition of the word.

Answer 24

Models don’t think since Machine Learning Models work with numeric class labels and the models don’t know what the inputs and outputs mean. Instead, Machine Learning Models makes associations between inputs and outputs.

Question 25

Answer: What is a Vector? What is Dimensionality?

Answer 25

A vector is a string of numbers treated as a single entity. Vectors are strings of numbers often representing measurements in a dataset. Dimensionality measures the quantity of elements in a vector.

Question 26

Answer: What is a Matrix? What are matrices?

Answer 26

A matrix is two-dimensional arrays of numbers. Matrices are multiple two-dimensional arrays.

Question 27

Rewrite: In machine learning, we often represent datasets as matrices, where the rows are vectors representing the elements of the dataset, like an iris flower, and the columns are the measurements.

Answer 27

Machine Learning Datasets are often in the form of matrices. The columns of a matrix are the measurement quantities of a vector, whereas the rows are the full vector.

Question 28

Answer: What is a 'feature' in Machine Learning Model terminology?

Answer 28

A 'feature' is an element of the known dataset input, that can help a Machine Learning Model learn to produce the correct output.

Question 29

Answer: What is a 'feature vector' in Machine Learning Model terminology?

Answer 29

A 'feature vector' is a multidimensional vector that contains many features.

Question 30

Rewrite: "Feature. An element of the feature vector input to a model. Features are data elements that have some relevance to determining the proper class label for an input. If the input is an image, each image pixel is a feature. Other possible features include measurements, location information, color, or any quantity (numeric) that can help a model learn to produce correct output."

Answer 30

A 'feature' is a data element of the known dataset input, that can help a Machine Learning Model learn to produce correct output. The correct output is usually the proper class label for input. The feature is contained within a multidimensional vector, called a feature vector.

Question 31

Answer: What does Machine Learning Etiquette suggest in regards to testing?

Answer 31

Machine learning etiquette states that we should test our model to know that it is working.

Question 32

Answer: How do you test a Machine Learning Model? Answer: What is the held-out test?

Answer 32

The proper way to test a model is to keep some of the known dataset input to use as test data after training. This is called the held-out test. The model’s performance on this held-out test dataset indicates how well the model has learned.

Question 33

Research: the term 'optimization,' in regards to parameters in Machine Learning.

Question 34

Research: the term 'generalization,' in regards to parameters in Machine Learning.

Question 35

Rewrite: "a foundational issue in machine learning: generalization. Some machine learning models follow a process quite similar to a widely used approach known as optimization. Scientists and engineers use optimization to fit measured data to known functions; machine learning models also use optimization to condition their parameters, but the goal is different. Fitting data to a function, like a line, seeks to create the best possible fit, or the line that best explains the measured data. In machine learning, we instead want a model that learns the general characteristics of the training data to generalize to new data. That’s why we evaluate the model with the held-out test set. To the model, the test set contains new, unseen data it didn’t use to modify its parameters."