Scribd Logo
Upload

Welcome to Scribd!

  • 264 views

    Respostas Corretas

    Uploaded by

    gilbertogavioli2
    The document provides answers to multiple choice questions about AI topics such as large language models, foundation models, self-supervised learning, bias in AI models, semantic search, domain adaptation, and techniques used by LLMs. The questions cover key concepts around responsible and ethical AI development and deployment.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Respostas Corretas

    Uploaded by

    gilbertogavioli2
    264 views3 pages
    The document provides answers to multiple choice questions about AI topics such as large language models, foundation models, self-supervised learning, bias in AI models, semantic search, domain adaptation, and techniques used by LLMs. The questions cover key concepts around responsible and ethical AI development and deployment.

    Original Title

    RESPOSTAS CORRETAS

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document provides answers to multiple choice questions about AI topics such as large language models, foundation models, self-supervised learning, bias in AI models, semantic search, domain adaptation, and techniques used by LLMs. The questions cover key concepts around responsible and ethical AI development and deployment.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    264 views3 pages

    Respostas Corretas

    Uploaded by

    gilbertogavioli2
    The document provides answers to multiple choice questions about AI topics such as large language models, foundation models, self-supervised learning, bias in AI models, semantic search, domain adaptation, and techniques used by LLMs. The questions cover key concepts around responsible and ethical AI development and deployment.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 3

    RESPOSTAS CORRETAS

    Questão 1
    What is the name of the architecture choice that allows for the most control over data?

    Fully private

    What is the purpose of prompt engineering?


    Prompt engineering involves designing and refining inputs (prompts) to effectively guide a large
    language model in performing specific tasks or generating desired responses

    Which of the following is a challenge when deploying large language models for knowledge
    search in the enterprise?
    All of the above

    ____________ is a complete enterprise-ready studio to train, validate, tune, and deploy AI


    models and a library of curated and trained IBM and open-source foundation models.
    watsonx.ai

    What could be the key to ensuring AI is used responsibly?


    Ethical guidelines

    What is the benefit of using deterministic answers from large language models in knowledge
    search?
    All of the above

    What does a new hire to IBM Consulting want to use Sidekick AI for?
    To get detailed information about how they're adhering to all the standards and equality

    What is the main advantage of using prompt tuning over fine tuning when tailoring a large
    language model to a specific task?
    Prompt tuning is more efficient as it avoids the need for extensive retraining and can be
    implemented more quickly

    What does IBM's technology offer in terms of data management?


    All of the above

    ____________ is a data store built on open lakehouse architecture optimized for governed data
    and AI workloads.
    watsonx.data

    To ensure that solutions are targeted to real world business scenarios, IBM will focus on the
    following AI use cases:
    IT processes, Customer service, and HR and talent management
    What are the three key characteristics of foundation models?
    They are trained on a large amount of data, they often use unsupervised learning, and they are
    trained on a specific task

    What is self-supervised learning?


    Self-supervised learning is a type of learning where a model is trained on a large amount of
    unlabeled data, enabling it to learn patterns and structures from the data itself without explicit
    labels.

    What could be a reason behind hallucinations in LLMs, as discussed in the information


    provided?
    Noisy and inconsistent training data

    Which factor must you consider when choosing afoundation model?


    Your use case

    How does Sidekick handle sensitive information?


    It abstracts the models and allows for integration of future advancements
    It enables the use of guardrails for safety and to reduce bias
    It provides a way for users to save and reuse frequently used prompts
    It flags it with e-information warning

    What is bias in AI models?


    The tendency of models to be fair and unbiased
    The tendency of models to make incorrect predictions
    The tendency of models to make accurate predictions
    The tendency of models to reflect inherent biases in the data

    Which of the following options correctly explains why a common risk of using Large Language
    Models (LLMs) is that of a “flagrantly false narratives"?
    AI doesn’t know how to tell authentic stories
    AI only guesses at the next best syntactically correct word and cannot infer meaning
    Large Language Models are created by hackers
    IBM is not governing the rules of AI in all situations

    How can semantic search help reduce the number of documents needed for large language
    models in knowledge search?
    By using cosine similarity
    By using Azure cognitive search
    By using Watson Discovery
    By using embeddings
    What is the Sidekick AI code analyzer assistant trained to do?
    To create personas
    To summarize calls
    To generate code
    To analyze code quality

    What is domain adaptation in AI?


    Domain adaptation in AI refers to the administrative tasks involved in AI model management
    Domain adaptation in AI is focused on enhancing the computational speed and efficiency of AI
    models
    Domain adaptation in AI involves training AI models solely on generic, non-specialized data
    Domain adaptation in AI is the process of modifying and fine-tuning AI models to perform
    effectively in particular domain-specific tasks or contexts

    What is the purpose of the "Assistant" feature in Sidekick?


    To provide a way for users to save and reuse frequently used prompts
    To abstract the models and allow for integration of future advancements
    To enable the downstream fine-tuning of API models
    To provide a way for users to chat with the AI model

    What does IBM's Technology organization offer in terms of AI?


    AI-powered commentary for highlight reels
    Models created by IBM for specific industries
    A platform for accessing open-source models
    B and C

    Which method used by LLMs might favor generic words with higher probabilities over specific
    words with lower probabilities?
    Sampling
    Reinforcement learning
    Maximum likelihood estimation
    Beam search

    You might also like