AI TECHNOLOGY

The dialog uses the _________ that are identified in the user's input, plus context from the application, to interact with the user and ultimately provide a useful response.

Intents

How does IBM view the impact of artificial intelligence? A) It is here to augment human intelligence. B) It is how systems predict the future. C) It is about automating mundane tasks that humans do. D) It is about the quest for artificial general intelligence (AGI).

A) It is here to augment human intelligence

In a neural network, changes to the weight alter the steepness of the vector or curve. The bias, on the other hand ________ A) Shifts the entire vector or curve so it fits better B) Adjust the hyperparameters automatically C) Helps mitigate the local minimum problem D) Removes the outliers from the data set

A) Shifts the entire vector or curve so it fits better

To which of the following categories of artificial intelligence does the IBM Watson Jeopardy! quiz game adhere to?

Artificial Narrow Intelligence

_______________ is used to compute the informative question words based on the statistics of the text corpus. Named entity detection Relation Extraction Inverse Document frequency Deep parsing

C） Inverse document frequency

Which of the following functions is an example of a deterministic systems? A) Sigmoid operation B) Logistical regression C) y=f(x) D） Force = mass x acceleration (F=ma)

D）Force = mass x acceleration (F=ma)

unstructured data

Images, tweets, words : not numeric and in an organized format.

what is a dialog skill?

A dialog skill is a container for the artifacts that define the flow of a conversation that your assistant can have with your customers.

If the grading system was such that your grades were presented as A, A-, B+, B and so forth (each encompassing a range of numbers; for example, A is anything from 93 up to maximum of 100), which of the following will be the appropriate approach to your data analysis? A) Regression problem B） Classification problem C) Clustering problem D) Unsupervised learning

B) Classification problem

Amazon's recommendation engine uses which of the following machine learning methods? Unsupervised Supervised IBM jeopardy pipeline! CNN (nueral networks)

Supervised

Most chatbot failures have what three things in common?

Their creators 1. didn't define the purpose 2. set goals that are too ambitious 3. launch before they are ready

Natural language processing is needed when you wish to mine ________ data and extract meaningful insight from text a) unstructured B) Structured c) relational d) formatted

Unstructured

Inverse Document Frequency is a key annotator used in which stage of the UIMA pipeline? A)Question Analysis B) Hypothesis Generation C) Hypothesis Scoring D) Final Merger and ranking

a) Question analysis

Which of the following Natural Language Processing attributes are integral to the Watson Assistant service? Emotion Sentiment Entities All of the above

all of the above

27) Which of the following pose challenges to the field of natural language understanding? a) nuances b) colloquial way of talking c) idiosyncracies d) all of the above

d) All of the above

For a natural language processing (NLP) system, which of the following poses the greatest challenge? Syntax Semantics Pragmatics Grammatical

Pragmatics

When you purchase an item from Amazon, it also suggests to you other related items. What is the name of that AI engine that brings forth suggestions as such?

Recommender Engine

With AI systems, ______________is the process of refining an algorithm so that it can learn from a data set. The output of this operation is called a model. A model encompasses the learned coefficients of mathematical expressions. A） Training B) Convoluting C) Scoring D) Deploying

A) Training

When using 1-hot encoding, what are the elements within each "category" called? • A) Vector • B) Index • C) Category • D) Array

B) index

if the grading system was such that your grades were presented as A, A-, B+, B and so forth (each encompassing a range of numbers; for example, A is anything from 93 up to maximum of 100), which of the following will be the appropriate approach to your data analysis? A) Regression problem B） Classification problem C) Clustering problem D) Unsupervised learning

B） Classification problem

The Jeopardy! game played by Watson in 2011 used which of the following technologies? A) Convolutional Neural Networks B) UIMA architecture C) Deep Learning D) Internet scraping technologies

B） UIMA architecture

Decision trees, support vector machines, and naive Bayes are different techniques to solve a _____________ problem. A) Regression B) Clustering C） Classification D) Reinforcement

C) Classification

Which of the following best depicts general AI? A) A chatbot that is also connected to Tone Analyzer B) Watson Machine Learning service C) The ability to use previous experiences to come up with new creative ideas D) A system that is self-aware

C) The ability to use previous experiences to come up with new creative ideas

How does the system know that a wise man is not the same as a wise guy? A) By labeling or tagging portraits of individuals that fall in either category B) Because it is already trained on the Wikipedia. C) Because this is a classification problem and the confidence value, through iterative training, will classify properly. • D) Al systems just "know" that the two phrases are opposite in even if the verbiage is the same.

C） Because this is a classification problem and the confidence value, through iterative training, will classify properly

Training a neural network is often characterized as: A) Ingesting more data B) Using benchmark statistics C） Minimizing cost function D) Finding the global minimum

C） Minimizing cost function

Ethics is becoming a major trust issue in AI development. As an example, bias has to do with a prejudice in favor or against something and in practice it can lead someone to behave unfairly toward a certain group compared to others. Which of the following is an example of a blatant bias in current AI systems? a) Systems being able to distinguish and mask profanity B) Systems being able to understand colloquial spoken language C） System recognizing age attributes on dark skinned female faces D) Autonomous cars stopping for a piece of paper or a leaf blowing up front of the vehicle

C） System recognizing age attributes on dark skinned female faces

Humans have certain types of data that machines don't; which of the following statements supports the above assertion? A) Data from our senses, smell, taste, touch, hearing and intuition B) We are the ultimate arbitrators of our own preferences C) Privacy concerns restrict the data available to machines D） All of the above

D) All of the above

Machine learning deals with the problem of extracting features from data to solve many different predictive tasks. Which of the following are relevant examples? A) Forecasting (e.g. Energy demand prediction, finance) B) Imputing missing data (e.g. Netflix recommendations) C) Detecting anomalies (e.g. Security, fraud, virus mutations) D) All of the above

D) All of the above

Best practices to successfully implementing AI within an organization includes which of the following approaches? • A) Develop an AI strategy and roadmap • B) Establish AI capabilities and skills • C) Start small and scale quickly • D) All of the above

D) all of the above

Re-enforcement learning is a machine learning algorithm that is best suited for which of the following tasks? A) Identify faces for access to secure areas B) Virtual agents and other chatbots C) Discovery systems that can crawl the web D） Stocking, as well as retrieving, products in the warehouse for optimizing space utilization and warehouse operations

D） Stocking, as well as retrieving, products in the warehouse for optimizing space utilization and warehouse operations

AI-powered customer intent recommendation engine uses which of the following approaches? A) Profile: Holistic client profiles based on multi-brand, multiyear purchase history B) Measure: Distance/similarity between clients based on corresponding profiles C) Learn and predict: Learn from other clients and predict the (cross-)sell potential for a given product D） All of the above

D）All of the above

CIMON is a robotic astronaut currently onboard the ISS and keeping the astronauts' company in orbit. In doing so, which of the following IBM Watson services are mostly likely used in CIMON? A) Speech to text and text to Speech B) Watson Assistant C) Visual Recognition D） All of the above

D）All of the above

Which of the following statements best explains the explosion of data in recent years? A) Proliferation of unstructured data B) Faster and more efficient hardware C) Ubiquitous sensors and instruments in society D） All of the above

D）All of the above

When using the gradient descent algorithm, the gradient, or slope of the 'descent' is also referred to as the ___________. A) Global minimum B) Local minimum C) Weight D）Learning rate

D）Learning rate

Cognitive systems are best characterized by the following four attributes: Understand, reason, learn and ______________. A) Interact B) Predict C) Optimize D) Converge

Interact

What is an Intent?

The action or purpose behind an utterance (generally a verb)

what is an entity?

a person, place, thing or event about which information is maintained An entity represents a term or object that provides context for an intent.

If you had to break down cognition and how it works in four words, what would they be? a) Observe, interpret, evaluate and decide. b) Collect data, build models, test and deploy c) Observe, measure, build and interpret d) Gather data, build patterns, memorize and recall

a) Observe, interpret, evaluate and decide

AI enables a partnership between humans and machines. Which of the following is best performed by tech and machines rather than humans? a) Pattern identification b) Imagination c) Abstraction d) Dilemas

a) Pattern identification

[Select Two] When working with the Watson Assistant service, you are prompted to create intent, entity and dialog. Which of those tasks invokes the machine learning process to start in the background? Intent Entity Dialog Content catalog

b) entity. A) intent

Which of the following speaks to human advantages over machines? A) Data from our senses, smell, taste, touch, hearing and intuition B) We are the ultimate arbitrators of our own preferences C) Privacy concerns restrict the data available to machines D） All of the above

d) all of the above

Candidate answer generation collects all the results of primary search and begins to formulate a _________ that is more focused towards the keywords that had the greatest saliency. A) Focus B) Hypothesis C) Part of speech D) Corpus

B） Hypothesis

23) The dialog uses the _________ that are identified in the user's input, plus context from the application, to interact with the user and ultimately provide a useful response. Intents Entities Content catalog System entities

a) intents

AI ethics framework components: (not an MCQ)

1. Accountability 2. User data rights 3. Value alignment 4. Explainability 5. Fairness

What are the possible values of the dependent variable in a sigmoid operation? A） Any value between 0 and 1 B) Only integers (whole numbers and it can be any number as long as it is not negative C) The values are not numeric. They are categorical names D) Any number from -1 to +1

A) Any value between 0 and 1

All learned knowledge is uncertain and learning itself is a form of uncertain inference. The problem then becomes how to deal with noisy, incomplete and even contradictory information, without falling apart. The solution is probabilistic inference, and the algorithm is ____________. A） Bayes theorem B) Support Vector Machines C) Inverse deduction D) Stochastic gradient descent

A) Bayes theorem

The Project Debater uses ________________ that allows it to find arguments to support the general human dilemmas that are raised by the debate topic. A） Knowledge graph B) Argument construction C) Scored evidence D) UIMA pipeline

A) Knowledge graph

How would you categorize Watson the Jeopardy! game winner in 2011 with regard to the level of artificial intelligence it displayed? A） Narrow artificial intelligence B) Broad artificial intelligence C) General artificial intelligence D) Super artificial intelligence

A) Narrow artificial intelligence

Fairness (Not an MCQ)

AI must be designed to minimize bias and promote inclusive representation.

User Data Rights (not an MCQ)

AI must be designed to protect user data and preserve the user's control over access and uses

Explainability (Not an MCQ)

AI should be designed for humans to easily perceive, detect, and understand its decision process

Value Alignment (Not an MCQ)

AI should be designed to align with the norms and values of your user group in mind.

In the example of water flowing into a tank, the flow rate of water is akin to the ___________ and the volume of water in the tank is akin to _____________. A） Derivative, Integral B) Integral, Derivative C) Input, Mean D) Mean, Input

A） Derivative, Integral

Supervised learning has many advantages. Which of the following may be some shortcomings of supervised learning? A) Requires vast amounts of data B） Labeling the data is arduous and expensive C) There are not used much as of late D) Clustering is difficult in supervised learning

B) Labeling the data is arduous and expensive

Let's say you are tasked with climbing blind folded to the top of a hill and required to do so in as few steps as possible. In machine learning practice, what would the slope of the hill (curve) be akin to? A) Synapse B） Learning rate C) Weights/Hyperparameters D) Recall Rate

B) Learning Rate

What are some of the most common artificial neural networks used with vision? A) Support Vector Machines B) Random Forrest Trees C) Convolutional Neural Networks D) Regression

C) Convolutional Neural Networks

[Select All That Apply] Which of the following examples are perfect scenarios for machine learning systems? A） Humans are unable to explain their expertise B） Solution changes with time C） Solution needs to be adapted to particular cases D) There is scarcity of dat

C) Solution needs to be adapted to particular cases B) Solution changes with time A) Humans are unable to explain their expertise

The basic building block in a neural network is called a perceptron. Which of the following best describes the role of the perceptron? A) It is designed to mitigate bias B) It is designed to avoid the local minimum C） It takes an input and produces an output value that is governed by an activation function. D) It performs the backpropagation algorithm

C）It takes an input and produces an output value that is governed by an activation function

What is the name of this curve: A) Parabolic tangent B) Sum of Squared errors C) Ordinary least squares D） Sigmoid operation

D） Sigmoid operation

Re-enforcement learning is a machine learning algorithm that is best suited for which of the following tasks? A) Identify faces for access to secure areas B) Virtual agents and other chatbots C) Discovery systems that can crawl the web D） Stocking, as well as retrieving, products in the warehouse for optimizing space utilization and warehouse operations

D）Stocking, as well as retrieving, products in the warehouse for optimizing space utilization and warehouse operations

Which of the following algorithms is used for supervised learning? A) Clustering B) Gaussian mixture C) Hidden Markov model D） Support Vector Machines

D）Support Vector Machines

Which of the following best describes the difference between AI, machine learning and deep learning? a) Each level is differentiated by a specific algorithms that power the system b) They are synonyms for the same approach for the field of AI c) Deep learning is a subset of machine learning, which in turn is a subset of AI d) Machine learning is a subset of deep learning, which in turn is a subset of AI

c) Deep learning is a subset of machine learning, which in turn is a subset of AI

Which of the following learning methods is predominantly used with self-driving cars? a) Supervised learning b) Unsupervised learning c) Reinforcement Learning d) None of the above because self driving cars use Broad AI

c) Reinforcement Learning

AI learning process entails human and machine learning, which of following best depicts Machine learning? a) Identify the data-analytics problem b) Define hyperparameters c) Analyze the data provided by the models d) Learns by adjusting weights and biases

d) Learns by adjusting weights and biases

Which of the following is an example of intent used in the Watson Assistant service?

#turn_on

Assigning an image to one or more categories, is different from assigning a single word to that image. The term often used to indicate assigning word(s) to an image is known as: • A) Anaphora • B) Tagging • C) Semantic search • D) Knowledge Graph

B) tagging

The turing test is designed to:

Determines Machine's human ability

Let's say you are tasked with climbing to the top of a hill, blindfolded and required to do so in as few steps as possible. In machine learning parlance, what would the slope be called? A) Synapse B) Learning rate C) Hyper parameter D) Hidden layer

B） Learning rate

From 1974 to 1980, there was the first 71 winter. Which of the below precipitated this decline in AI work? A) The Lighthill report and governments' funding restrictions B) There were no scientists working on the problem in those years. C) The algorithms were not good enough. D) Because of the Cold War

A) The Lighthill report and governments' funding restrictions

Training a Visual Recognition service to be able to differentiate between objects, is best described by which of the following remarks? A) Upload both positive and negative images. B) Expose the system to countless images object you want it to identify. C)Ensure the zip file contains the name of the object you are uploading. D) You must first take a minimum that you want

A) Upload both positive and negative images.

Is it possible to use more than one service in a given application deployed on IBM Cloud? • A) Absolutely yes. You can add a number of connections to the app, each connection being a separate API. • B) You will need a new app for each service. • C) Yes, but only if you upgrade to Premium service plan. • D) It is possible, but you must limit it to a maximum of two services per deployed application.

Absolutely yes. You can add a number of connections to the app, each connection being a separate API.

Both Amazon and Netflix use advanced recommendation engines; however, their approach to making a recommendation differs considerably from each other. Which of the following best describes such differences in approach? A) Amazon recommends from the short tail of the question-distribution graph, whereas Netflix recommends from the long tail, the more obscure. B) Amazon uses unsupervised learning whereas Netflix uses supervised learning. C) Amazon recommendation engine relies heavily on knowledge graph as it's search engine, whereas Netflix scrapes the web continuously. D) Amazon has far more advanced algorithms that can search and recommend simultaneously.

Amazon recommends from the short tail of the question-distribution graph, whereas Netflix recommends from the long tail, the more obscure.

When predicting future grades based on past performance that had 'hours sleep' and 'hours studying' as dependent variables, why was there a need to normalize hours and grades? A) Because that would be like comparing apples to oranges B) Because normalization is required with all classification problems C) Because normalization is part of inferential statistics D) All of the above are true

A） Because that would be like comparing apples to oranges

26) Tone Analyzer uses linguistic analysis to detect three types of tones from written text: A） Emotions, social tendencies, and writing style b)Intents, entities, and dialog c) enrichments, Discovery, and facial expression analysis d) Tone analyzer requires speech to text before it begins analysis

A） Emotions, social tendencies, and writing style

15. Which of the following best depicts the main task of hypothesis scoring? • A) It entails assigning confidence values to generated hypotheses retrieved earlier in the pipeline. • B) The generated hypothesis looks for document title, anchor text and if it is a web search, likely the Wikipedia title. • C) It uses semantic search and knowledge graph to extract relevant passages from the corpus. • D) All of the above

A） It entails assigning confidence values to generated hypotheses retrieved earlier in the pipeline

12) UIMA stands for Unstructured Information Management Architecture. What is its main purpose? o A) Its main goal is to transform unstructured information to structured information. • B) It is the premier pipeline for analyzing the meaning of the spoken words. • C) Its main purpose is to extracts the semantic from the syntactic in a given sentence. • D) All sentences need to be processed through the UIMA pipeline before the outcome is passed onto neural networks.

A） Its main goal is to transform unstructured information to structured information.

[Select Two] How would you go about using two services, the Watson Assistant and the Discovery services, together where one picks up where the other leaves off? services, together where one picks up where the other leaves off? A) You can wire them together in Node-RED. B) You can use one of the Dialog nodes to conditionally jump to the Discovery service using confidence value thr C) Ensure the Watson Assistant service is accessed before the Discovery Service in your pipeline. D) You cannot combine these services, instead use a service that already includes both running as one service

B） You can use one of the Dialog nodes to conditionally jump to the Discovery service using confidence value thresholds., A） You can wire them together in Node-RED.

Which of the following approaches you would use if you wanted to predict, based on size, whether a house would sell for more than $200K. Taking into consideration that the possible outputs are either 'Yes', the house will sell for more than $200K, or 'No', the house will not. A) Linear regression B） Logistical regression C) Sigmoid operation D) Classification

B）Logistical regression

The difference between narrow intelligence and broad intelligence is best described as the following phrases? a) Narrow and Broad intelligence is distinguished by complexity of algorithms. b) Broad intelligence requires huge amounts of data, whereas narrow intelligence relies on minimum data. c) Narrow AI can only do a certain task-and it can do it quite well-but narrow Al can't transfer its knowledge to different sorts of problems as with Broad Al. d) Broad intelligence can extract meaning from images, whereas Narrow Intelligence is best suited for language processing.

c) Narrow AI can only do a certain task-and it can do it quite well-but narrow Al can't transfer its knowledge to different sorts of problems as with Broad Al.

13) Which of the following best depicts or exemplify Broad AI? Self driving car, which happens to be a collection of Narrow AI technologies A) A system that supports engineers who work on complex maintenance tasks on a platform in the middle of the atlantic ocean b) A system that supports engineers who work on complex maintenance tasks on a platform in the middle of the atlantic ocean c) A system within a bank that analyzes the balance sheet of a corporate.... d) ALL of the above

d) ALL of the above

Social media analysis can benefit from an AI system that can perform Sentiment Analysis. How would such an NLP approach help an organization? would such an LP approach help an organization? A) User-generated social content, but specific conversations can also be found to help companies better interact with customers. B) When an incident occurs in real time, applying NP to monitor social media provides a distinct advantage to help businesses react immediately. • C) Analyze relevant trending topics and hashtags for a business. 0 D) All of the above

d) All of the above

IBM Business Value 2017 2Q C-suite research has identified three stages of automation: Basic, advanced and intelligent. Which of the following examples depicts an intelligent business process automation? • A) The "robot" is taught to drive applications like a human B) Uses AI to complete specific tasks C) The "robot" follows predetermined pathways, conducts complex calculations and performs actions D) The "robot" reasons and learns through data discovery

d) The "robot" reasons and learns through data discovery.

28) How is the syntax of a sentence different from the semantics of the same sentence? • A) Syntax refers to the grammar of the sentence, namely, object-subject-verb. • B) Semantics refers to the Agent and Patient relationship of words in a sentence. • C) Semantics refers to the meaning of the sentence • D) All of the above

d) all of the above

Start small and scale quickly has become a mantra on how to implement an AI strategy within a company. Which of the following best describes such an approach? A) Start with minimum viable products (MVPs) • B) Set understandable key performance indicators (KPIs) • C) Roll-out through company (culture) D) All of the above

d) all of the above

You can connect to a data source and pull documents on a schedule (if desired) into the Discovery service by configuring a collection to associate with that source. Which of the following data sources can the Discovery service crawl, using both tooling and APIs? Box Salesforce Microsoft Powerpoint All of the above

d) all of the above

22) In a conversational world, marketers can no longer rely on the "yell & sell" approach and hope for the best. They need to "chat and listen." Which of the following terms best describes the new challenges that marketing departments face? I• A) They need artificial general intelligence (AGI) to address this issue. • B) It's about personalization. • C) It's about advanced algorithms. • D) It's about having immense amounts of data.

• B) It's about personalization.

33) Considering the aphorism: "April showers bring May flowers," which of the following is the proper annotator that knows May does not likely bring floods or pebbles, but rather flowers? A) Bi-gram, tri-gram, n-gram B) Textual Alignment C) Logical Form Analysis D) Part of Speech

Bi-gram, tri-gram,n-gram

AI Engagement, as an adoption driver of the technology, is most concerned with which of the following uses? • A) Personalization • B) Providing a second opinion for doctors • C) Predictive failure detection • D) Automation

Personalization

When you think of applying machine learning frameworks to your collected data, which of the following could benefit from deep learning machine services? • A) Text analysis • B) Vision analysis • C) Time series analysis • D) All of the above

all of the above

The first step in Project Debater is to build an opening speech to defend or oppose the motion. Project Debater searches for short pieces of text in the massive corpora that can serve this purpose. Which of the following activities brings forth the power of AI in Project Debater? A） To understand human language and its infinite nuances and colloquial way of talking B) To search and retrieve the relevant passages or text segments C) To respond back in a human voice D) To recognize the humans on the debate stage

A）To understand human language and its infinite nuances and colloquial way of talking

Which of the following is an example of object localization used with visual recognition? A) Tagged images • B) Bounding box • C) Geospatial awareness of the location of the object • D) Uploading positive images

B) Bounding box

Which of the following best describe what happens during the Question Analysis stage in the UIMA pipeline? a) The document are ingested at this stage B） It parses the question to extract the major features of the question c) it scores the candidate answers d) SPSS modeler is used in the question analysis stage to extract keywords

It parses the question to extract the major features of the question

Self-driving cars need to see, understand how the objects are moving, and decide what to do next. Which of the following, as elaborated in the Driverless Cars video, best categorizes the above three requirements? • A) Perception, prediction, planning • B) LIDAR technology, LASER technology and machine learning • C) Visual recognition, Natural Language Understanding and Watson Machine Learning • D) All of the above

Perception, prediction, planning

Through______________ learning techniques, predictive models were developed that could anticipate failure detection for welding robots and predictive maintenance assessment. A) Supervised › B) Unsupervised C) Iterative D) Advanced

Supervised

One of the great challenges of driverless cars is awareness of social interactions as depicted in the Driverless Cars video. Which of the following is an example given in the video? A) Understand the driving habits of the occupant and drive as they would B) To never change lanes, under any circumstance, should the lane departure require crossing the yellow divider markings C) To understand that kids are playing with a ball to the side of the road and that the ball may tumble in front of the car and then kids may be running right behind it D) Calculate if the passenger is late for a meeting and drive more aggressively to make the appointment

To understand that kids are playing with a ball to the side of the road and that the ball may tumble in front of the car and then kids may be running right behind it

Explain how Ai systems understand, reason, learn and interact

Understand: AI systems can understand unstructured data the same way humans do Reason: They can reason, grasp underlying concepts, form hypotheses, and inter to extract ideas Learn: Each data point, interaction and outcome helps to continuously sharpen expertise Interact: With abilities to see, talk and hear, Al systems interact with humans in a natural way

From retail to education, AI technology areas play an important role in reinventing industries and transforming our world. Which of the following are examples of key AI technologies? • A) Virtual agents and computer vision • B) Education and healthcare • C) Privacy and bias • D) Personalization and automation

Virtual agents and computer vision

In 2021, IBM makes international headlines when their system appears on Jeopardy!-and winds, What was the underlying architecture of the system? a) It was founded on Unstructured information Management Architecture(UIMA) b) It was powered by advanced neural networks c) It used a Brute force search engine d) It had access to the internet

a) it was founded on Unstructured information Management Architecture(UIMA)

12) There seems to be common consensus as to what was AI born. Which of the following matches that consensus? a) When alan turing published the paper "Computing Machinery and intelligence" b) In 1943 when Warren McCulloch and Walterpits, researchers from the university of chicago, propose the "nerve net" c) In 1956 at dartmoth college summer workshop with top scientists d) In 1957 when frank rosenblatt first demonstrated the Perceptron on an IBM 704

c) In 1956 at dartmoth college summer workshop with top scientists

16. In 198-, Alan Turing publishes the paper "Computing Machinery and intelligence" introducing: a) Space X prize b) Nobel Prize c) Turing Test d) First paper on AI

c) Turing Test

Humans have superior data gathering and learning techniques to machines, but where or when can humans augmented help? a) When there is endlessly relevant newly published information, such that is impossible to know all at all times b) When humans try to make in-depth if statements in minimum and optimized time c) When there are many variables at play d) All of the above

d) All of the above

is structured or unstructured data correlated/ associated with unsupervised or supervised learning

no. All types of learning can handle all types of data.

Structured data

numerical data that can be put into an excel sheet

In 1982 IMF researcher Gerald Tsuara creates a self-teaching neural net that learns to play a) Chess b) Go game c) BackGammon d) Checkers

c) BackGammon

24) How is Watson Assistant different from Watson Discovery service? A) Watson Assistant strives to answer the frequently asked; whereas, the Discovery service strives to provide passages for the esoteric inquiries B) Watson Assistant is used mainly with chatbots and virtual agent systems; whereas, the Discovery service is used mostly by researchers trying to find the relationship of entities and other enrichments. C) With Watson Assistant, We actually enter the response; whereas with Discovery service, the API retrieves the nrelevant passage from ingested data. • D) All of the above are true.

d) all of the above are true

9) The IBM Watson Jeopardy! game was able to achieve all of the main components of artificial intelligent systems, minus the one other key differentiator that current AI systems possess. Which of the following was missing in the IBM Watson Jeopardy! Game? • A) Understand • B) Reason • C) Learn • D) interact

d) interact

ways to start small and scale quickly?

start with minimum viable products, set understandable key performance indicators; roll out through company culture

21) Which of the following is a true statement regarding natural language processing (NLP)? • A) NLP is a classification problem. • B) LP is a regression problem. C). NP has not yet conquered the challenge of depicting human tone. • D) LP is used when the system needs to produce a new composition made up of numerous separate article passages.

• A) NLP is a classification problem.

Analog Al delivers radical performance improvements by combining compute and memory single device, eliminating • A) Mathematical trap of local minimum • B) The von Neumann bottleneck • C) The problem of vanishing gradient • D) The need to go from speech to text and text to speech, instead, directly going from speech to speech

• B) The von Neumann bottleneck

20) Let's say I am deploying an AI system that can handle Level 1 support calls to my organization. Which of the following systems would be an appropriate approach? • A) Recommender engine • B) Discovery Engine • C) Agent Assist chatbot • D) Hadoop

• C) Agent Assist chatbot

What is meant by gradient descent in a neural network? A） In machine learning, it's typically used to minimize the error function. B) It means the partial derivative of cost function divided by weights is equal to zero C) It means achieving local minimum D) It means the learning rate is too small

A） In machine learning, it's typically used to minimize the error function.

How do you develop an Ai strategy and roadmap?

1 First, it's important to understand Al and to research what it can and can't do for your organization 2. Once Al is understood, the next question you should ask yourself is: "What specific problem do I want to solve, or what opportunity do I want to take?" 3. Prioritize the use cases into a transformation roadmap that covers both a long-term vision as well as concrete feasible quick wins. 4 Think about what data you have available

Best practices to successfully implement AI within an organization

1) Develop an AI strategy and roadmap 2) Establish AI capabilities and skills 3) Start small and scale quickly

Visual Recognition uses deep learning, including convolutional neural nets (CNN) to analyze a picture. Which of the following best describes what is meant by deep learning using a CNN approach? A) It learns by layers; the earlier hidden layers depict only edges and additional layers learn more features as the image is being analyzed. B) It learns by using a sigmoid operation applied directly to the input. C) It learns by starting with an arbitrarily chosen low value for the weights and increases the weights as data moves upward through the hidden layers. D) The visual recognition algorithm uses a feedback system in the CNN construct to learn new features

A) It learns by layers; the earlier hidden layers depict only edges and additional layers learn more features as the image is being analyzed.

[Select Two] When you think of the syntax of a sentence, which of the following comes to mind? Concept Sentiment Annotate Tokenize

Annotate and Tokenize

To build a conversation agent or a virtual agent that can sense and respond accordingly to frustrated or joyful inflections in human voice, you would need the Conversation service plus a minimum of what other essential service? A) Discovery • B) Speech to Text • C) Tone Analyzer D) Natural Language Classifier

C） Tone Analyzer

The Washington Post in a 2016 article stated that: half of the American workforce is at risk of being replaced by robots in the next decade. Which of the following tasks is key to successful AI technology adoption? A) Increasing use of natural language processing B) Incorporating machines that adapt, and learn to make recommendations C) Training humans to work with machines D) Optimizing business processes for automation

C） Training humans to work with machines

Audi uses robots to independently transport cars from production to shipping positions. Which of the following best depicts the AI approach taken with Audi in its manufacturing plant? A) Data is collected from sensors to determine position and dimensions of the car and then adjusts to it. B) The robot uses the dimensions of the facility to navigate. C) The robot understands the verbal commands given to it by technicians and performs accordingly. D) The robot is programmed to perform certain tasks based on the manufacturing stage that the car is undergoing in production.

Data is collected from sensors to determine position and dimensions of the car and then adjusts to it.

[Select Two] Why would you want to have more than one Workspace in a single instance of your Watson Assistant?

Each workspace could act as a separate bot or application, A） You could use one workspace for regular chit-chat and the second one for the domain-specific business conversations

While relevant AI use cases span various areas across virtually every industry, there are three main macro domains that continue to drive the adoption as well as the most economic savings across businesses. Which of the following best depict those three main domains? a) Engagement, insight and automation b) Narrow, Broad, and general c) Supervised, unsupervised and reinforcement d) All of the above

Engagement, insight and automation

Radiologists visually examine a large number of images per day. Which of the following can be a cause of concern for their task? • A) Eye fatigue • B) Built in bias may affect analysis • C) A radiologist can effectively look at only 3 images per hour. • D) Lack of total knowledge of the latest discoveries in the field

Eye fatigue

Digressions allow for the user to break away from a specific conversation in order to temporarily change the topic before returning to the original dialog flow. How do you set up digressions in Watson Assistant?

From the dialog tab

In 1997, BM Deep BLUE a computer capable of analyzing over 200 million moved per second beats world champion:

Gary Kas(something) Gary Kasparov

Think of a question answer scenario, in which of the following is easy for humans and difficult for machines?

Humans understand the question but may not have all the answers (get it? because I didn't give you guys the answer choices)

Which of the factors below has been a compelling motivation for machine learning approaches? • A) The growth of unstructured data • B) The need to extract insights from images • C) The need to extract sentiment from Tweets • D) All of the above

all of the above

In 2006 Hinton colleagues demonstrate that _____ with many layers can be effectively trained a) single perception b) Deep learning c) Support vector machines d) Linear Regression

b) Deep learning

In what way will AI augment education? a A) Personalized Education • B) Learning Content Indexing-to-Skill & Search o C) Custom Teaching Methods 0 D) All of the above

d) all of the above

Human experts could use AI expertise to address which of the following? A) Al scales expertise to expand what is possible B) Al systems have better vision, C) Al systems can prediet our preferences D) Al systems have world knowledge,

A) Al scales expertise to expand what is possible

When analyzing images, which of the following machine learning approaches is best suited for image analysis? • A) Convolutional neural network • B) Recurrent neural network • C) Support Vector Machines D) Random Forest Trees

A) Convolutional neural network

Given that the X-axis (independent variable) contains the area in square feet of houses, and Y- axis contains the corresponding sale price of those houses, which of the following approaches would you use to predict the house prices as the square footage increases? A Linear regression B) Logistical regression C) Sigmoid operation D) Classification

A) Linear regression

With machine learning algorithms that use gradient descent, the idea is to use ________ to find the downward slope. A) Partial derivatives B) Integrals C) Averages D) Mean

A) Partial derivatives Descent= derivative

Derivatives and Integrals are used often with machine leaning algorithms. Given that derivatives and integrals are opposite of each other, and since we measure surface area under a curve by integrals, then what do derivatives measure? A） Slope B) Coefficient C) Accuracy D) Outliers

A) Slope

Which of the following would constitute a deep learning framework? A） TensorFlow B) Siri C) Generative Adversarial Networks D) Cortana

A) TensorFlow

Statisticians often refer to 'allowing some bias in exchange for reducing variance'. Which of the following depicts what statisticians are referring to? A） Unlike regression, machine learning predictions might be wrong on average, but when the predictions miss, they often don't miss by much. B) Regression takes the data and tries to find the result that minimizes prediction mistakes, maximizing what is called goodness of fit C) Being precisely perfect on average can mean being actually wrong each time D) All of the above

A）Unlike regression, machine learning predictions might be wrong on average, but when the predictions miss, they often don't miss by much.

Let's say you want to predict how much salary one would earn based on level of education. Your Y axis is salary and your X axis is educational bucket (high school, Bachelors, Masters and so forth). Which of the following models is best suited to help you predict, given a certain salary, what might the education level of the individual be? A) Logistical regression B) Linear regression C) Sigmoid operation D) Classification

B) Linear Regression

Linear regression tries to fit a line, while ___________ the distance to each data point. A) Maximizing B） Minimizing C) Integrating D) Squaring

B) Minimizing

Which of the following events is distinctly in the realm of general AI and not a capability of today's artificial intelligence? A) Analyze videos and not just images B) Remember the actions it took before and if "unrewarded" to not repeat that same action C） Use previous experiences to come up with new creative ideas D) Chat with humas using empathy

C）Use previous experiences to come up with new creative ideas

When you think of a Platform as a Service (PaaS), which of the following comes to mind? A) Microsoft Azure B) Amazon AWS C) IBM Cloud D） All of the Above

D) All of the Above

Machine learning systems are good at making predictions, they can traverse through a complex (if-then-else statement) faster and with greater prediction accuracy than humans. Assuming this statement to be true, which of the following is an area where humans would currently excel in competition with machines? A) Machines require lots of data, humans don't B) Humans can obtain more data and not just from sight or words, but smell, feel and auditory C) Humans can judge and make decisions on preferences D） All of the above

D) All of the above

What makes a deep learning network "deep"? A) It has been trained many times and its accuracy has improved over time B) The system had access to "deep" knowledge as its corpus C) The system has many neurons D） It is a multi-layer perceptron with many 'hidden' layers

D） It is a multi-layer perceptron with many 'hidden' layers

Since AI systems depend on data, then humans have the following advantages. A) Humans have world knowledge B) Awareness of self in relation to the environment C) Humans are better at decision making with little data or in rare circumstances D) All of the above

D）All of the above

Which of the following statements are true of Scholastic Gradient Descent (SGD)? A) SGD is widely used for larger dataset training, is computationally faster and can be trained in parallel B) Need to randomly shuffle the training examples before calculation it C) SGD is also used for linear regression D）All of the above

D）All of the above

Which of the following characterizes the difference between a linear and logistic regression? A) In linear regression, the outcome (dependent variable) is continuous. It can have any one of an infinite number of possible values B) In logistical regression, the outcome (dependent variable) has only a limited number of possible values C) Linear regression is usually solved by minimizing the least squares error of the model to the data, therefore large errors are penalized quadratically D） All of the above are true

D）All of the above are true

Accountability with AI (not an MCQ)

Every person involved in the creation of AI at any step is accountable for considering the system's impact in the world

In a convolutional neural network, there are many layers such as input, layer, pooling, ReLU and so forth. The connection from one layer to another layer is done using programmatic 'synapses,' which contain the 'weights.' What do the nodes inside the hidden layers contain? A） Bias B) Weights C) Additional input data D) All of the above

a) Bias

In a convolutional neural network with many layers, when images are uploaded to the system, what do the first layers perform? • A) The first layers begin to group the pixels in color clusters. • B) The first layers convert the vector features of the images into numbers for the algorithms to work with. • C) The first layers identify edges in the images. • D) The first layers are dedicated to the input layers where raw data comes in the network.

• C) The first layers identify edges in the images

Applying bias, such as a sigmoid operation or ReLU and a host of other algorithms, takes place in which particular construct of a neural network? • A) Node B) Weight • C) Synapse D) Hyperparameter

A) Node

A single perceptron can solve a linear problem, then additional perceptrons give a network the ability to _________. A) Solve non-linear problems B) Optimize loss functions C) Seek Global minimum D) Normalize unstructured data

A) Solve non-linear problems

What makes a neural net a deep learning network? A) There are lots of layers known as hidden layers that run further analysis of the data that came through the input layer. • B) Deep learning pertains to depth of domain knowledge. • C) Deep learning is the process of an artificial neural network relearning over an iterative process. o D) None of the above

A) There are lots of layers known as hidden layers that run further analysis of the data that came through the input layer.

[Select Two] In a digital marketing firm, which of the following is a crucial data set for you to analyze? A). Ensure your customers are fully aware of which algorithms you used for transparency sake B) Strive to adopt a personalized 'chat and listen' approach versus a yell and sell' approach C) Being always on and keeping it real becomes therefore much more important for a brand than being pixel perfect (and thus only sporadically on/always late) D) Do not throw away any data points, it all helps to predict a better outcome. Answer:

C） Being always on and keeping it real becomes therefore much more important for a brand than being picture or pixel perfect (and thus only sporadically on/always late)., B） Strive to adopt a personalized 'chat and listen' approach versus a 'yell and sell' approach.

Oxford philosopher and leading AI thinker _________ defines superintelligence as "an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills." • A) Nick Bostrom • B) Andrew Ng • C) Geoffrey Hinton • D) Alan Turing

Nick Bostrom