HubSpot AI Essential for Partners Answers

HubSpot AI Essential for Partners Answers

Exam

• There are 60 questions.
• The exam takes about 1 hour to complete.
• You must answer 45 questions correctly to pass.
• You must wait 12 hours between attempts.
• You have 3 hours to complete the exam.

 

 

 

Questions:

What is a key legal consideration when using customer data in AI systems?

• Compliance with data privacy laws (e.g., General Data Protection Regulation (GDPR)
• File size limits
• Model speed
• User interface design

 

Predicting whether a customer will churn (leave) is an example of which type of machine learning task?

• Clustering
• Regression
• Classification
• Reinforcement

 

Claude, ChatGPT, Gemini, and Grok are all examples of generative AI assistants built on an LLM.

• True
• False

 

In reinforcement learning, how does the system learn?

• By analyzing pre-labeled datasets
• Through trial and error with feedback in the form of rewards or penalties
• By observing humans
• By clustering unlabeled data

 

In machine learning, what is overfitting?

• A model that's too simple and underperforms
• A model that performs well on both training and unseen data
• A model that performs well on training data but poorly on unseen data
• A model that uses too much data

 

True or false? Deep learning is a type of machine learning that uses large amounts of data and layered neural networks to learn complex patterns.

• True
• False

 

Which of these platforms markets its model as particularly focused on being "helpful, harmless, and honest"?

• Claude
• Grok
• Gemini
• ChatGPT

 

Which principle ensures humans can understand and explain how AI makes decisions?

• Fairness
• Transparency
• Automation
• Generalization

 

Which of the following should be prioritized when advising clients on responsible AI?

• How much money the AI system will save
• Minimizing training time for the model
• Ensuring fairness, privacy, and transparency in how AI is applied
• Choosing the most complex algorithm available

 

Which of the following is an example of unsupervised learning?

• Classifying loan applications as approved or denied
• Grouping customers into segments based on behavior
• Predicting stock prices
• Translating text from English to Spanish

 

A client with no data science team wants to deploy a language translation feature. What’s the most practical approach?

• Hire AI researchers to build a model from scratch
• Use a cloud-based translation API like Google or Azure Translate.
• Install an open-source ML library and code the logic themselves.
• Avoid AI and use manual translation.

 

True or false? AI can be used to personalize product recommendations in B2B ecommerce platforms based on customer behavior.

• True
• False

 

In the context of generative AI, what is fine-tuning?

• Making small adjustments to a pre-trained model using domain-specific data
• Manual editing AI-generated content
• Reducing the size of a language model
• Creating more concise prompts

 

When explaining model training to a business stakeholder, which description is most appropriate?

• The process of teaching employees to use AI systems
• A technical process where AI learns patterns from historical data to make future predictions or decisions
• Creating physical models of business processes
• Developing organizational training materials

 

What is training data?

• Data used to back up an application
• A list of business rules codec into an AI system
• Historical data used to teach an AI model how to recognize patterns
• Data collected after a model is deployed

 

In the context of LLMs, what is a token?

• A security credential for accessing the API
• A fundamental unit of text that the model processes (word parts, punctuation, etc.)
• A form of cryptocurrency used to pay for AI services
• A complete sentence processed by the model

 

In the context of AI, what is an inference?

• The process of training an AI model with labeled data
• Using a trained model to make predictions or decisions on new data
• The development of AI hardware
• The extraction of features from raw data

 

Fill in the blank: ___ is when a machine learning model performs well on training data, but poorly on new data.

• Underfitting
• Normalization
• Inference
• Overfitting

 

Which technique helps protect sensitive data when training AI models?

• Real-time inference
• Data anonymization
• Using raw personal data
• Maximizing batch size

 

True or false? It is unethical to use AI-generated content in marketing without clearly indicating it was created by AI.

• True
• False

 

What is a model card in responsible AI development?

• A physical ID card for AI developers
• Documentation that describes a model's intended use, limitations, performance metrics, and ethical considerations
• A visual diagram of neural network architecture
• A credit card used to purchase AI services

 

In the context of machine learning (ML), what is a model?

• A flowchart that maps out business logic
• A collection of training data
• A mathematical representation trained to recognize patterns or make predictions.
• A user interface for interacting with AI

 

Fill in the blank: Deep learning is a subset of____.

• human intelligence
• data engineering
• machine learning
• robotics

 

Which of the following best describes the difference between AI and traditional software?

• Traditional software automates task using logic, while AI systems adapt and learn from data.
• AI systems require no data, but traditional systems do.
• AI runs only on cloud platforms, while traditional software runs locally.
• Traditional software uses neural networks and AI does not.

 

A client’s support team is overwhelmed with incoming tickets. They want to automate triage and respond to common issues. Which AI solution should you recommend?

• Forecasting model
• Conversational AI chatbot integrated with a knowledge base
• Image recognition
• Recommendation system

 

What does facial recognition typically involve?

• Identifying speech patterns
• Extracting emotions from text
• Matching faces in images to known individuals
• Recommending products to users

 

Fill in the blank: Clustering techniques are used when you ____.

• have labels for your data
• want to group data points with similar features
•  want to train a chatbot
• are analyzing time-series data

 

True or false? A pre-trained AI model can often be customized with a client’s data to improve performance.

• True
• False

 

When building a machine learning model, what's the purpose of a test dataset?

• To generate training labels
• To improve model performance during training
• To evaluate the model's performance on new, unseen data
• To increase the size of the training dataset

 

True or false? The field of AI and regulatory guidance around it are static.

• True
• False

 

What's a key difference between traditional AI and foundation models?

• Foundation models sure self-supervised learning.
• Traditional AI can write entire documents.
• Foundation models are limited to narrow task.
• Traditional AI is trained on unlabeled data.

 

What is a common reason AI projects fail to deliver value?

• Lack of executive sponsorship and business alignment
• Too much data
• Use of cloud services
• Too much automation

 

Which of these is the least likely to benefit from AI?

• Repetitive data entry
• Personalizing marketing emails
• Prediction customer churn
• Replacing critical legal decision-making

 

A consulting client wants to personalize product recommendations for their customers. What AI capability would best support this?

• Optical character recognition
• Recommendation system
• Image segmentation
• Chatbot framework

 

True or false? Tools like Microsoft Copilot and Notion AI are examples of AI applications built on top of foundation models.

• True
• False

 

Your team wants to use ChatGPT to help draft internal documents. What critical limitation should you be aware of?

• It can only generate very short texts.
• It may confidently present incorrect information as factual.
• It can only work with predetermined templates.
• It cannot understand industry terminology.

 

What is Artificial General Intelligence (AGI)?

• AI that can write marketing copy
• AI that excels in processing visual data
• A still-theoretical form of AI that can understand and learn any intellectual task a human can
• A common term for no-code AI tools

 

What's a neural network?

• A type of data science project
• A complex mathematical equation for learning A to B mappings
• A hardware component used in AI
• A tool for automating tasks

 

A client wants to generate marketing copy at scale using AI. Which of the following is the most accurate statement about this use case?

• Generative AI can assist, but outputs should be reviewed for accuracy and tone.
• Generative AI generates factually correct information.
• This isn't a recommended use case for AI.
• Generative AI requires proprietary datasets to work.

 

Check all that apply. Which of the following is a potential application of AI?

• Boosting productivity in HR
• Customer service
• Cybersecurity
• Recommending products based on past purchases

 

Which of the following best defines artificial intelligence (AI)?

• A set of hardware components that simulate human movement
• A software system that mimics human intelligence to perform task
• A programming language used for automation
• A cloud service used to store machine data

 

What's a key difference between Retrieval-Augmented Generation (RAG) and traditional LLM implementation?

• RAG models are always smaller and faster.
• RAG incorporates external knowledge sources to supplement model responses.
• RAG only works for coding tasks.
• RAG doesn't require prompts.

 

Which type of machine learning involves training a model using labeled data?

• Unsupervised learning
• Reinforcement learning
• Supervised learning
• Deep learning

 

What is named entity recognition (NER) used for?

• Extracting names, locations, or organizations from text
• Recognizing specific objects in videos
• Identifying images in a dataset
• Translating documents

 

What’s the main benefit of using a platform like OpenAI’s API when building tools?

• It builds full applications automatically.
• It provides access to general-purpose language models for a wide range of text-based tasks.
• It only works for image classifications.
• It requires advanced data science knowledge to use.

 

What's a defining feature of generative AI models like GPT?

• They classify images.
• They generate new content such as text or images.
• They compress large files.
• They only perform regression.

 

True or false? AI systems trained on biased data can unintentionally produce unfair or discriminatory outcomes.

• True

• False

 

How is machine learning (ML) different from traditional programming?

• ML uses hard-coded rules to make decisions.
• ML writes its own software from scratch.
• ML learns from data to make predictions or decisions without being explicitly programmed.
• ML only works for image processing tasks.

 

True or false? Prompt engineering is a specific skill used by technical developers, not consultants.

• True
• False

 

A client’s AI model is unintentionally producing biased results for a specific demographic. What should you recommend?

• Stop using AI entirely.
• Remove the demographic data to make it fair.
• Ignore the issue unless complaints are received.
• Conduct a bias audit and retrain the model with more diverse data.

 

True or false? In supervised learning, the model learns from data without any labels or target outputs.

• True
• False

 

True or false? Generative AI tools like ChatGPT can NOT help consultants draft client reports, proposals, and marketing content.

• True
• False

 

A client is unsure whether AI is appropriate for their process automation needs. As a consultant, what’s the best first step?

• Recommended an AI tool.
• Ask about the available data and the problem's predictability.
• Begin coding a prototype.
• Suggest hiring more staff.

 

What is a deep fake?

• A complex AI algorithm
• A synthesized video/image of someone doing something they didn't
• A type of adversarial attack
• A biased AI output

 

A client asks you if AI can help improve their monthly sales forecasting. Which machine learning technique should you recommend?

• Clustering
• Classification
• Regression
• Reinforcement learning

 

Why is prompt engineering important when using LLMs?

• It allows you to reduce data size.
• It improves model training speed.
• It helps get more accurate and relevant outputs from the model.
• It makes the model learn new knowledge.

 

True or false? Every AI project should begin with a clear understanding of the business goal it's trying to solve.

• True
• False

 

True or false? Generative AI models like GPT can only summarize existing content and can NOT create anything new.

• True
• False

 

Which of the following is NOT a typical use case for conversational AI?

• Internal knowledge base access
• Customer support
• FAQ automation
• Image classification

 

Which of the following best describes ChatGPT and its competitors like Gemini and Claude?

• Spreadsheet analysis tools
• Rule-based expert systems
• Large language models designed for natural language understanding and generation
• Data encryption platforms

 

True or false? Hugging Face is a tool for creating low-code AI dashboards.

• True
• False

 

True or false? Natural language processing allows AI systems to understand, interpret, and generate human language.

• True
• False

 

Which task would best be handled using Natural Language Processing (NLP)?

• Predicting weather patterns
• Detecting anomalies in iOT sensor data
• Classifying images of cats and dogs
• Analyzing customer support transcripts

 

What should businesses do to protect their competitive advantage in AI?

• Outsource their data.
• Protect their data.
• Share their models openly.
• Ignore transparency.

 

What makes conversational AI different from simple rule-based bots?

• It uses NLP and machine learning to understand and generate text dynamically.
• It's unable to understand natural language.
• It uses fixed responses only.
• It's limited to answering yes or no questions.

 

A B2B SaaS client wants to improve how their SDR team prioritizes leads. Which AI capability should they use?

• Image classification
• Lead scoring using predictive analytics
• Facial recognition
• Anomaly detection

 

In the context of AI, what makes a system intelligent?

• It uses a user interface.
• It can store large amounts of data.
• It mimics human reasoning and decision-making.
• It connects to the internet.

 

What's one practical way a consultant can help a client increase transparency in an AI system?

• Use proprietary models.
• Recommend regular data deletion.
• Document how the model makes decisions and provide explanations for users.
• Limit who can see outputs.

 

In the context of AI, what is hallucination?

• AI generating creative content
• AI generating incorrect answers with confidence
• AI processing data very quickly
• AI having emotional responses

 

Why is accountability important in responsible AI?

• It designates human oversight for AI actions and outcomes.
• It ensures AI will always be accurate.
• It reduces data storage costs.
• It eliminates the need for AI ethics.

 

Which of the following is NOT a commonly used AI tool or framework?

• PyTorch
• TensorFlow
• Hugging Face
• Python

 

True or false? Claude is an open-source AI tool anyone can download and run locally.

• True
• False

 

You've been tasked with leading your department's first AI initiative. What should be your first step?

• Purchase and AI platform.
• Restructure your department to focus on AI.
• Hire data scientists.
• Define the specific business problem you're trying to solve.

 

Which AI platform offers access to open-source foundation models and a large model hub, often used for fine-tuning or experimentation?

• Salesforce Einstein
• GitHub
• Hugging Face
• Webflow

 

Which of these is a common limitation of generative AI models?

• They can generate content in multiple formats.
• They struggle to cite sources in a verifiable way.
• They may generate plausible-sounding but incorrect or biased information.
• They're incapable of learning from prompts.

 

What is a key feature of Artificial Narrow Intelligence (ANI)?

• It can understand and perform any intellectual task a human can.
• It specializes in a single task or domain.
• It's more powerful than human intelligence.
• It can replicate human emotions.

 

Which step involves putting an AI model into use in a real-world environment?

• Training
• Validation
• Deployment
• Feature extraction

 

Which of the following best improves a prompt to generate a detailed response from an LLM?

• Break the prompt into multiple unrelated questions.
• Keep the prompt extremely short.
• Include emojis.
• Give context and ask for step-by-step reasoning.

 

How can AI bias be reduced?

• By using more a complex algorithms
• By using more historical data without changes
• By avoiding cloud services
• By diversifying training data and testing model fairness

 

True or false? Bias in AI doesn’t occur as long as the model itself is technically accurate.

• True
• False

 

True or false? All AI models are interpretable and can easily explain how they make decisions.

• True
• False

 

What is algorithmic bias?

• An error in coding
• A problem found in robotics
• A tendency for an AI system to produce skewed or unfair outcomes
• A feature of deep learning models

 

A client asks you to build a model that selects job applicants based on historical hiring data. As a consultant, what’s your first ethical concern?

• Whether the model can scale across departments
• If historical data may contain bias that could be perpetuated by the model
• If the client prefers a cloud or on-prem solution
• Whether it's faster to use a rule-based system

 

Which of the following is NOT an example of Artificial Narrow Intelligence (ANI)?

• A chatbot that answers FAQs
• A self-driving car trained for road navigation
• An AI that generates email responses
• An AI that can independently invent scientific theories and understand human emotions

 

True or false? ChatGPT can write code, answer questions, summarize text, and simulate conversation.

• True
• False

 

True or false? Microsoft and Google offer AI APIs that allow businesses to analyze language, extract key information, and translate text.

• True
• False

 

Which of the following is true of Artificial General Intelligence (AGI)?

• It has already been achieved and deployed.
• It specializes in single tasks.
• It's capable of learning and performing any intellectual task a human can.
• It's the same as robotic process automation.

 

What’s one key question to ask when assessing whether a task is suitable for AI?

• Can this task be completed with no human involvement at all?
• Does this task require constant manual rule writing?
• Is there enough data to learn patterns from?
• Does it involve high-performance hardware?

 

What are the three modes of AI consumption?

• Embedded AI, API calls, platform model
• Cloud-based AI, on-premise AI, hybrid AI
• Free AI, paid AI, subscription AI
• Research AI, development AI, production AI