Bus 5730

A Perceptron is an ANN with only two layers.

False

A classification model is used to predict how long a person is likely to live.

False

A model successful for training data set means that my machine learning algorithm is ready for use.

False

A regression model is used to predict things like "how likely is someone to buy a specific item."

False

AI and Machine Learning (ML) are important for companies, but they don't impact our personal lives.

False

AI cannot classify skin lesions as cancerous or benign as well as a board certified dermatologist.

False

AI is just one part of Machine Learning.

False

Army research has shown that AI can readily be applied to various "things" on the battlefield to provide significant and reliable aid to warfighters.

False

Artificial Intelligence (AI) is a relatively new area of study.

False

Clustering is an example of supervised machine learning.

False

Constructing a decision tree is all about finding an attribute that returns the highest entropy and the smallest information gain.

False

During the data preparation process (step 4), we should simply delete any sample with missing values.

False

Edge computing means that data is processed in the professional data centers so it can be processed faster and safer.

False

Forward chaining in expert system answers the question "Why did this happen?".

False

Having Big Data is a prerequisite for building deep-learning ANNs.

False

It's difficult to debug expert systems because the information in the database is not explicit.

False

It's hard to get experts to select the most important items for scorecards. They seem to include far too many items.

False

One expert noted that we have many established methods to test AI for safety, fairness, and effectiveness.

False

Personal data is illegal to use in ML systems, worldwide.

False

"Reinforcement Learning" involves having computers repeat an action until something difficult goes more smoothly and then favoring the behavior that led to that outcome.

True

AI can write movie scripts but is not as creative as human.

True

AI has become more capable due to the continued improvement of computers and computing, commonly known as Moore's Law.

True

Any situation in which you have a lot of data and are trying to predict an outcome is a potential application for supervised learning systems.

True

Autonomous robots are used to boost agriculture total factor productivity (TFP) growth.

True

Big Data normally refers not just to large amounts of data, but to data that is volatile.

True

DLSS uses low resolution images and upscales them using neural networks.

True

Deep Learning is the stepping stone for efficient and effective mainstream Machine Learning models in the open marketplace.

True

Deep learning is used to develop drugs and vaccines.

True

Deep neural networks are employed in self driving cars.

True

Ensembles utilize several models, like a group of experts.

True

Ethical use is an important consideration for ML systems, but what constitutes "ethical use" is not always clear.

True

Even though we may have large amounts of data, we sometimes may have to create additional data to revise existing data or summarize it in different ways.

True

Expert systems are modeled on the belief that human cognition uses "rules."

True

Expert systems try to solve problems that are difficult enough to require significant human expertise.

True

Facial recognition systems are currently in use by some police units.

True

Heuristics are ways of reacting to situations which might or might not lead to a "solution."

True

In ML, we have software that learns from examples, rather than being explicitly programmed for a particular outcome.

True

It's often possible to develop decision trees and scorecard models that reach essentially the same predictions.

True

ML efforts to aid in controlling online comments that are abusive are likely to be limited because of the need for context in some online models.

True

ML is the use of algorithms to analyze data.

True

Multi-layer ANNs generally use the Sigmoid activation function because it has a nice derivative, needed for the training mode.

True

Natural language processing (NLP) is a branch of artificial intelligence that helps computers understand, interpret, and manipulate human languages.

True

One focus in ML systems is to provide better interfaces to business managers as the users.

True

Scorecard models are attractive because we can see what data items contribute to the score someone receives.

True

Semantic analysis helps a machine understand the meaning of a text.

True

Some data that might be highly predictive might also be illegal to collect.

True

Stemming and Lemmatization are usually used in Natural language Processing (e.g. spam filtering) to prepare text, words, and documents for further processing.

True

Supervised learning discovers patterns in the data that relate data attributes with a target attribute.

True

The aim of ML is to discover useful patterns between different items of data and use them to make inferences about the behavior of new cases when they present themselves.

True

Three areas where AI can benefit healthcare are diagnostics, treatment, and administrative processes.

True

Unsupervised learning is a type of machine learning algorithm used to draw inferences from datasets consisting of input data without labeled responses or target attribute.

True

Variable/feature selection improves the accuracy of a model if the right subset is chosen.

True

We can be misled by the results of a model -- it looks as if it's very accurate, but really isn't better than other simple models.

True

We can calculate travel time of a commute with the Bluetooth technology.

True

While implementing virtual reality technology for military use is expensive at first, it is expected to reduce costs in the long term.

True

Expert systems are best at dealing with broad domains.

False

Leaf nodes are at the beginning of a decision tree.

False

ML learning systems usually replace the entire job, process, or business model when they are applied.

False

Netflix used only one model to optimize their recommendation engine.

False

Satellite GIS data is replacing the use of UAVs in Precision Agriculture.

False

Since databases for ML systems will be extremely large, there is no way that we can understand all the kinds of data used, nor should we try to.

False

Support Vector Machines algorithm is typically used for a low number of features.

False

The concept of an AI chatbot has failed because people wouldn't use it.

False

The data that we're concerned about relative to ethical use (age, race, gender, etc.) are not usually correlated with predictions anyway.

False

The difference between "active" and "passive" ML systems is that active ones are constantly adding active data.

False

We don't ever want to change the existing data of an ML system, even to consolidate inconsistent values.

False

What products customers usually buy together from Amazon is a classification question.

False

When measuring fairness, the smaller disparate impact value is (close to 0), the fairer a prediction model is.

False

While most social media companies use AI, it has little impact on the content we see when using their applications.

False

Artificial neural networks (ANN) are better for motion tracking in virtual reality devices than convolutional neural networks (CNN).

False