P Caic 1 1-4 Flashcards
Priority FC from csv (152 cards)
1
Q
What has AI transitioned from in recent years?
A
AI has transitioned from a theoretical concept discussed in research papers and academic conferences to a transformative force reshaping industries, economies, and societies worldwide.
2
Q
What is the projected growth of the global AI market from 2023 to 2025?
A
$207.9 billion in 2023 to $1 trillion by 2025, with a compound annual growth rate of 20.1%.
3
Q
What sectors are expected to adopt AI significantly?
A
- Healthcare
- Finance
- Manufacturing
- Logistics
4
Q
What is the expected impact of AI-powered diagnostics?
A
AI-powered diagnostics are expected to reduce diagnostic errors by 85%.
5
Q
What value is anticipated from AI-driven financial services?
A
AI-driven financial services are anticipated to contribute over $1 trillion in value by automating and optimizing decision-making processes.
6
Q
What does the CAIC certification focus on?
A
The CAIC certification is tailored for professionals who wish to deepen their understanding of AI and its business implications.
7
Q
What are the two broad definitions of AI mentioned?
A
- AI as the area of computer science that studies how machines can perform tasks that would normally require a sentient agent
- AI as the area of computer science that studies how machines can closely imitate human intelligence
8
Q
True or False: A computer multiplying two numbers can be considered a form of artificial intelligence.
A
True
9
Q
How does the human brain compare to machines in terms of processing tasks?
A
The human brain can recognize, label, and classify objects with a few examples, while machines often require thousands of examples.
10
Q
What are the main reasons we study AI?
A
- Automate processes
- Handle large amounts of data efficiently
- Ingest data from multiple sources simultaneously
- Update knowledge constantly
- Respond in real-time with high precision
11
Q
What are the classifications of AI branches?
A
- Supervised learning vs. unsupervised learning vs. reinforcement learning
- Artificial general intelligence vs. narrow intelligence
- By human function: machine vision, machine learning, natural language processing, natural language generation
12
Q
Fill in the blank: AI is an area of computer science that studies how machines can perform tasks that would normally require a _______.
A
sentient agent
13
Q
What is a typical feature of a machine learning system?
A
It uses patterns from previously seen data (the training data) to make inferences on new, unseen data.
14
Q
What is the primary goal of AI?
A
To build smart systems that can understand patterns and behaviors of entities.
15
Q
What are some examples of AI applications mentioned?
A
- Self-driving cars
- Intelligent robots
16
Q
What is the significance of the human brain in AI development?
A
The human brain serves as an idealized model for creating intelligent systems.
17
Q
What does AI aim to achieve in terms of data processing?
A
AI aims to index and organize data to derive insights and learn from new data.
18
Q
What type of certification programs does USAII provide?
A
Best-in-class certification programs to upskill and reskill the AI talent force.
19
Q
What does the CAIC certification program provide to its candidates?
A
A self-help study kit that includes three books in eLearning format and additional materials.
20
Q
What is the main challenge faced by the human brain in the context of AI?
A
The human brain cannot keep track of insurmountable amounts of data.
21
Q
What limits the power of machine learning programs?
A
The size of the dataset
If the dataset is small, the learning models will be limited.
22
Q
What does a typical machine learning system do with previously unseen data?
A
Uses patterns from previously seen data to make inferences
This process involves matching features like eyes, nose, and lips in facial recognition systems.
23
Q
What is the role of mathematical logic in logic-based AI?
A
Executes computer programs using logical statements to express facts and rules
This is used for pattern matching, language parsing, and semantic analysis.
24
Q
What are search techniques used for in AI programs?
A
Examine many possibilities to pick the most optimal path
Commonly applied in strategy games, networking, and resource allocation.
25
What is ontology in the context of AI?
A formal definition of properties and relationships of entities in a domain
## Footnote
Usually represented with a taxonomy or hierarchical structure.
26
What is the aim of optimal planning in AI?
To achieve maximum returns with minimal costs
## Footnote
Programs start with facts and a goal to generate the most optimal plan.
27
Define a heuristic in the context of AI.
A technique for solving problems that is practical but not guaranteed to be optimal
## Footnote
Often used in robotics and search engines.
28
What is genetic programming?
A method for solving tasks by mating programs and selecting the fittest
## Footnote
Programs are encoded as genes to perform tasks effectively.
29
Who classified machine learning into different tribes?
Pedro Domingos in his book 'The Master Algorithm'
## Footnote
He classifies machine learning according to the scientific field from which it originated.
30
What are the main tribes of machine learning as defined by Domingos?
* Symbolists
* Connectionists
* Evolutionaries
* Bayesians
* Analogizers
## Footnote
Each tribe uses different algorithms and concepts from various scientific fields.
31
What is the primary tool used by Symbolists?
Induction or inverse deduction
## Footnote
Induction infers general principles from specific examples, while inverse deduction works backwards.
32
What do Connectionists primarily use in AI?
Neural networks
## Footnote
These are algorithms designed to recognize patterns, similar to the human brain.
33
What is deep learning?
A specialized type of neural network
## Footnote
It focuses on learning from large amounts of data through multiple layers of processing.
34
What concepts do Evolutionaries focus on in AI?
Evolution, natural selection, genomes, and DNA mutation
## Footnote
They apply these concepts to data processing through evolutionary algorithms.
35
How do Bayesians handle uncertainty in AI?
Using probabilistic inference
## Footnote
They update hypotheses based on new data and prior probabilities.
36
What is the k-nearest neighbor algorithm?
The most famous model used by Analogizers
## Footnote
It finds similarities between examples to make predictions.
37
What is the Turing test?
A test to determine if a computer can mimic human behavior
## Footnote
Proposed by Alan Turing, it assesses if a machine can engage in conversation indistinguishably from a human.
38
What are the requirements for a machine to pass the Turing test?
* Parse sentences
* Store and track information
* Interpret stored information
* Adapt to new conditions
## Footnote
These abilities are crucial for responding appropriately in a conversation.
39
What is a Total Turing Test?
A test that includes vision and movement capabilities
## Footnote
It requires a machine to use computer vision and robotics.
40
What is cognitive modeling?
Simulating the human thinking process
## Footnote
It helps in understanding how humans solve problems and is used in various AI applications.
41
Define rationality in the context of AI.
Observing rules and their logical implications to achieve desirable outcomes
## Footnote
An agent acts rationally if it takes actions based on available information to achieve its goals.
42
What is the General Problem Solver (GPS)?
An AI program intended to solve any general problem
## Footnote
It was the first program to use a base algorithm for various problems, created by Herbert Simon, J.C. Shaw, and Allen Newell.
43
What language was created for the General Problem Solver?
Information Processing Language (IPL)
## Footnote
This language expressed problems in well-formed formulas as part of a directed graph.
44
What is the first step in structuring a problem for the GPS?
Define the goals
## Footnote
For example, getting milk from the grocery store.
45
What are operators in the context of GPS?
Actions, preconditions, and changes resulting from actions
## Footnote
They define what conditions must be met to achieve the goals.
46
What is the main constraint of the General Problem Solver?
Its search process is computationally complex and time-consuming
## Footnote
This limits its applicability to well-defined problems.
47
What methods can impart intelligence to rational agents?
* Machine learning
* Stored knowledge
* Rules
## Footnote
These techniques help agents to effectively interact with their environments.
48
What is a rational agent?
An entity that acts to achieve the best outcome or the best expected outcome based on its understanding of the environment.
## Footnote
Rational agents make decisions based on their perceptions and actions in a given environment.
49
What are the commonly used techniques to impart intelligence to an agent?
* Machine learning
* Stored knowledge
* Rules
## Footnote
These techniques help in enhancing the capability of agents to perform tasks intelligently.
50
How does machine learning impart intelligence to an agent?
Through data and training.
## Footnote
Machine learning models learn from labeled data to identify patterns and relationships.
51
What role does the inference engine play in machine learning?
It runs the learning model to make predictions based on the input received from sensors.
## Footnote
The inference engine processes data and derives decisions that lead to actions.
52
List some applications of machine learning.
* Image recognition
* Robotics
* Speech recognition
* Predicting stock market behavior
## Footnote
Machine learning is widely used in various fields due to its ability to analyze and interpret data.
53
What are the two types of models in AI?
* Analytical models
* Learned models
## Footnote
Analytical models rely on mathematical formulations, while learned models are derived from training data.
54
What is the primary issue with analytical models?
They are often simplistic and inaccurate, based on human judgment.
## Footnote
Analytical models involved lengthy derivations and trial and error.
55
What defines learned models in AI?
They are obtained through training, using many examples of inputs and outputs to develop complex and accurate equations.
## Footnote
Learned models do not require deriving a mathematical formula manually.
56
What is required to build a learning model?
Data that is representative of the world.
## Footnote
The quality and relevance of the data significantly influence the performance of the learning model.
57
What command is used to check the version of Python 3?
$ python3 --version
## Footnote
This command helps verify if Python 3 is installed correctly on the system.
58
What is recommended for Mac OS X users for installing Python 3?
Use Homebrew.
## Footnote
Homebrew simplifies the process of managing software packages on Mac OS X.
59
What Python packages are necessary for data interaction in machine learning?
* NumPy
* SciPy
* scikit-learn
* matplotlib
## Footnote
These packages provide essential tools for data manipulation and visualization.
60
How do you load the house prices dataset in Python?
>>> house_prices = datasets.load_boston()
## Footnote
This command imports the dataset for further analysis.
61
What does ANI stand for in AI terminology?
Artificial Narrow Intelligence.
## Footnote
ANI refers to AI systems that are specialized in one task and do not possess general intelligence.
62
What is the difference between strong AI and weak AI?
Strong AI refers to machines that can think and reason like humans, while weak AI (ANI) refers to systems designed for specific tasks.
## Footnote
Currently, all existing AI is classified as weak AI.
63
True or False: AGI (Artificial General Intelligence) exists today.
False.
## Footnote
AGI is still a theoretical concept and has not been achieved.
64
What are the four major learning categories in machine learning?
* Supervised learning
* Unsupervised learning
* Semi-supervised learning
* Reinforcement learning
## Footnote
Each category has its own models and algorithms tailored to specific types of data and tasks.
65
What is the main focus of product managers in AI product management?
To help organizations discover the best combination of infrastructure, training, and deployment workflow to maximize success.
## Footnote
Product managers play a critical role in bridging technical and market needs.
66
Fill in the blank: If a machine is learning from some past behavior and its success rate is improving, it is _____ .
ML (Machine Learning).
## Footnote
Learning is the active element that defines machine learning, distinguishing it from traditional programming.
67
What are the four major areas of ML?
Supervised learning, Unsupervised learning, Semi-supervised learning, Reinforcement learning
68
What distinguishes supervised learning from unsupervised learning?
Supervised learning involves labeled data, while unsupervised learning uses unlabeled data.
69
What is a key characteristic of deep learning (DL)?
DL algorithms identify patterns or features in data using a hierarchical layered system.
70
What is the relationship between machine learning (ML) and deep learning (DL)?
DL is a subset of ML, with DL focusing on neural network algorithms.
71
What is the primary concern for product managers regarding DL models?
Explainability of the model's conclusions and processes
72
What do ensemble modeling techniques involve?
Using multiple models and selecting the best-performing one.
73
Fill in the blank: In supervised learning, humans label the data, and the machine tries to correctly label current or future data points, making it ______.
supervised learning
74
Name two applications of supervised learning models.
Classification models, Regression models
75
What is the purpose of a Naive Bayes algorithm?
To find associations probabilistically without assumptions about the data.
76
How does a Support Vector Machine (SVM) function?
It splits the dataset into two classes to predict future data points.
77
What is the function of linear regression models?
To predict a dependent variable using one or more independent variables.
78
What type of predictions does a logistic regression model make?
Binary categorical predictions.
79
What does a decision tree model resemble?
A flow chart, with nodes and branches.
80
What is a random forest model?
An ensemble of decision trees that averages or takes a majority vote for predictions.
81
In unsupervised learning, what is the model primarily looking for?
Patterns in unlabeled data.
82
What is clustering in the context of unsupervised learning?
Segmenting or grouping data into clusters.
83
What does dimensionality reduction achieve?
It simplifies data by removing less important features.
84
What is the challenge of using unsupervised learning on small datasets?
Results can be wildly inaccurate.
85
What does semi-supervised learning combine?
Labeled and unlabeled datasets.
86
How does reinforcement learning operate?
Through trial and error, learning from past behavior.
87
True or False: Semi-supervised learning requires only labeled data.
False
88
What is the ultimate objective of choosing a learning model?
To achieve necessary performance and explainability.
89
What are some challenges companies face when operationalizing AI/ML?
Updating models, keeping data fresh, organizing experiments, validating and testing.
90
Fill in the blank: A central place to store data for AI/ML models is referred to as a ______.
data warehouse
91
What does ETL stand for in data engineering?
Extract, Transform, Load
92
What is a common issue when scaling AI/ML systems?
Clunky or slow data delivery systems.
93
What is the goal of using multiple models in semi-supervised learning?
To improve performance with labeled and unlabeled data.
94
What is a key challenge in storing data as your tech stack evolves?
Choosing a cost-effective and reliable solution.
95
What happens when the load time for customer-facing dashboards lags?
It may be due to the number of customers and their large metadata.
96
What is referred to as the flow of constantly maintaining a system in DevOps?
Continuous maintenance.
97
What are the four major components of the continuous maintenance process?
* Testing/validating code and components
* Code changes or updates passed continuously
* Continuous learning for models
* Monitoring models for performance.
98
What can happen if ML/AI models are not iterated constantly?
Models and hyperparameters will become stale.
99
What are potential consequences of improper model maintenance?
* Lags in performance
* Job losses
* Unfair mortgage rates
* Unfair prison sentences.
100
What is the significance of where and how data is stored for AI/ML performance?
It impacts the effectiveness of AI/ML models.
101
What is the most cost-effective way to store unstructured data?
Data lake.
102
Fill in the blank: If your primary use of the database is querying to access data, a _______ might be enough.
relational database.
103
What is a data warehouse used for?
Centralizing and analyzing structured data.
104
What are the advantages of using a data lakehouse?
* Combines features of data lakes and warehouses
* Makes data available for non-technical users.
105
What is the difference between batch processing and real-time pipelines?
Batch processing moves large amounts of data at intervals; real-time pipelines provide data as soon as it's generated.
106
What does ETL stand for?
Extract, Transform, Load.
107
What is the role of a data engineer in AI/ML products?
To manage data flow and maintain ETL pipelines.
108
True or False: ETL pipelines are generally updated in real-time.
False.
109
What is the trend in managing AI/ML systems related to DevOps?
MLOps and AIOps working in conjunction with DevOps.
110
What does IaaS stand for?
Infrastructure as a Service.
111
What should companies consider when adopting AI?
Cost, storage, compute power, and investment required.
112
What is a shadow deployment strategy?
Deploying a new model alongside an existing model without affecting production.
113
What does the deployment strategy of shadowing allow for?
Testing the new model's performance without disrupting the existing model.
114
What can frequent changes in data require from your deployment strategy?
Continuous monitoring and retraining of models.
115
What is the goal of using a data warehouse for AI/ML applications?
To enable advanced analytics and insights across business units.
116
Fill in the blank: If you have lots of raw, unstructured data, you’d be looking at a _______.
data lake.
117
What should be done if models are left unchecked?
They may produce egregious results based on under-representative data.
118
What happens when a new AI model is fully deployed in production?
The original model will be retired.
119
What is the purpose of setting up two slightly different models concurrently?
To understand how each model performs in the live environment.
120
Why must the differences between two models be slight?
To understand what creates the most success.
121
What does a gradual approach to deployment involve?
Creating subsets of users that experience the new model over time.
122
What is the benefit of having a buffer time between groups of users in a gradual deployment?
To understand how they react and interact with the new model.
123
What factors influence the selection of deployment strategies?
* Nature of the product
* Importance to customers
* Budget
* Metrics and performance monitoring
* Technical capacity and knowledge
* Timeline
124
What is MLflow?
An open-source platform developed by Databricks to manage the complete ML life cycle.
125
What are the main benefits of using MLflow?
* Experiment tracking
* Model management
* Model deployment
126
What is Google's newest product for ML deployment built on?
TensorFlow.
127
What does Uber’s Michelangelo tool aim to standardize?
Collaboration and deployment of ML models.
128
What is the purpose of Meta's ML management system?
* Reusable ML algorithms
* Reusable training pipelines
* Easy and automated model training
* Searchable knowledge base of past projects
129
What does Amazon's ML product offer?
Building, training, and deploying ML models with managed infrastructure tools.
130
How does Airbnb approach its ML infrastructure?
By creating standardization and centralization between AI/ML organizations.
131
What is a key consideration for companies investing in AI?
Awareness of risks, costs, and level of investment needed.
132
What is the promise of AI rooted in?
Quantifying prediction and optimization.
133
What significant savings did Highmark Inc. achieve using ML?
$260M through fraud detection.
134
What percentage of Amazon's sales come from their recommendation engine?
35%.
135
What is the recommendation for starting AI projects?
Start small, apply it to a clear business goal, and track effectiveness.
136
What are the stages of the NPD cycle for an AI/ML product?
* Discovery
* Define
* Design
* Implementation
* Marketing
* Training
* Launch
137
What is the focus during the discovery phase of an AI product?
Isolating the problem to solve and understanding customer needs.
138
In the define phase, what is built from the feedback gathered?
A plan that screens ideas and selects the one with the highest potential.
139
What does the design phase involve for an AI product?
Creating mockups and identifying suitable models.
140
What is the main goal of the implementation phase?
To materialize the ideating and planning efforts into a working MVP.
141
Why is marketing crucial during the development of AI products?
To communicate effectively with the target market without overselling capabilities.
142
What is documented during the training phase of product development?
Justifications for choices made for the MVP and managing user expectations.
143
What is the sixth phase of product development focused on?
Training users and documenting the product
## Footnote
This phase includes creating justifications for the choices made for the MVP and overall product.
144
Why is managing expectations important for AI/ML products?
To help customers understand when to trust or question results
## Footnote
AI/ML products often involve predictions and optimizations that may vary over time.
145
What must be communicated to customers regarding AI/ML product performance?
Healthy margins of error to expect
## Footnote
This involves informing customers about the reliability and limitations of the product's predictions.
146
What is the goal of the final step in the product development process?
Officially launch the product into the market
## Footnote
This step includes scaling back to original definitions for performance and customer success.
147
What should be evaluated during the final product launch?
Whether the product hits the metrics originally set with customers
## Footnote
Assessing performance expectations and defining future achievable goals is also critical.
148
What will be reviewed in the following section after discussing the NPD process?
Popular ML model types commonly used in production
## Footnote
The section will cover characteristics shared by these models.
149
Fill in the blank: The final step is about _______ the product into the market.
launching
150
True or False: The process of training users includes managing their expectations about the product's performance.
True
151
List some characteristics that should be considered when evaluating AI/ML product performance.
* Metrics set with customers
* Performance expectations
* Future achievable goals
152
What is a key part of training users on your product?
Informing others about how to best interact with it
## Footnote
This involves ensuring users understand the capabilities and limitations of the product.
