According to the slides, what is a major concern related to AI hallucinations?

They can lead to misinformation and unreliable AI decision-making.

Which of the following is an example of a data poisoning attack?

Introducing misleading training data to manipulate AI predictions.

Why are AI-enhanced virtual assistants a security risk?

They can be manipulated to extract private user information

What is the primary goal of AI safety research?

To ensure AI systems avoid unintended harmful consequences

What was the main issue with Microsoft’s Tay chatbot?

It learned offensive behavior from interactions with Twitter users

What is the primary role of Google’s AI Red Team?

To simulate attacks on AI systems and identify vulnerabilities

Which of the following factors can contribute to AI hallucinations?

Insufficient or biased training data

How do companies attempt to reduce AI hallucinations?

By using fact-checking mechanisms and retrieval-augmented generation (RAG)

What is jailbreaking in the context of AI chatbots？

Using prompts to bypass AI safety mechanisms and force unintended behavior

studbud

Week 4 Reading Quiz - Responsible AI

10 termsStudBud Administration34 plays

Imported from external platform

Flashcards10 cardsFPractice quizPHost liveL

Questions 10

Answers

1
According to the slides, what is a major concern related to AI hallucinations?
They are easy to eliminate with more training data.
They can lead to misinformation and unreliable AI decision-making.
They make AI systems more creative and innovative.
They occur only in generative AI models.
2
Which framework was introduced by Google to ensure AI security?
AI Cyber Protection Model (AICPM)
Google AI Security Protocol (GAISP)
AI Defense System (AIDS)
Secure AI Framework (SAIF)
3
Which of the following is an example of a data poisoning attack?
Using adversarial perturbations to confuse an AI model after deployment.
Stealing an AI model’s source code to replicate it.
Introducing misleading training data to manipulate AI predictions.
Encrypting an AI model’s dataset to make it inaccessible.
4
Why are AI-enhanced virtual assistants a security risk?
They collect too much personal data
They are expensive to develop
They can be manipulated to extract private user information
They only work with cloud-based APIs
5
What is the primary goal of AI safety research?
To make AI models more computationally efficient
To enhance AI’s creative abilities
To ensure AI systems avoid unintended harmful consequences
To replace human decision-making in all aspects of society
6
What was the main issue with Microsoft’s Tay chatbot?
It learned offensive behavior from interactions with Twitter users
It failed to generate meaningful responses
It stopped responding to user queries after 24 hours
It was unable to process natural language
7
What is the primary role of Google’s AI Red Team?
To develop new AI models for Google products
To replace traditional cybersecurity teams
To simulate attacks on AI systems and identify vulnerabilities
To create AI models that are resistant to all cyber threats
8
Which of the following factors can contribute to AI hallucinations?
Insufficient or biased training data
AI models being too transparent
AI always verifying responses before generating an answer
AI models using retrieval-augmented generation (RAG)
9
How do companies attempt to reduce AI hallucinations?
By training AI models on a single source of information
By using fact-checking mechanisms and retrieval-augmented generation (RAG)
By removing all generative capabilities from AI models
By allowing AI models to generate unrestricted content
10
What is jailbreaking in the context of AI chatbots？
Hacking into AI servers to steal proprietary information
Unlocking premium features of AI chatbots without paying
Modifying AI hardware to improve response time
Using prompts to bypass AI safety mechanisms and force unintended behavior

Week 4 Reading Quiz - Responsible AI Flashcards | StudBud