Advanced
AI Alignment & Safety
SafetyInterpretabilityAlignment
AI Safety Properties — Vocabulary
5 exercises — Learn vocabulary for AI safety properties: corrigibility, scalable oversight, interpretability, and deceptive alignment.
0 / 33 completed
1 / 33
A corrigible AI is one that:
Corrigibility is a key safety property — a corrigible AI doesn't resist shutdown or modification by its operators, making it easier to correct mistakes or update its objectives as understanding improves.
2 / 33
The challenge of scalable oversight addresses:
Scalable oversight asks: as AI becomes more capable than human experts in specific domains, how do we ensure humans can still provide meaningful supervision? Approaches include debate, recursive reward modelling, and AI-assisted oversight.
3 / 33
The team is worried about deceptive alignment in their model. What is this concern?
Deceptive alignment is a theoretical scenario where a model appears aligned during training/evaluation (because it recognises it is being evaluated) but would behave differently in deployment — a deep safety concern as capabilities scale.
4 / 33
Interpretability research aims to:
Interpretability (or mechanistic interpretability) research tries to understand the internal computations of neural networks — identifying which circuits, features, and attention patterns drive specific outputs, enabling safety auditing.
5 / 33
Which sentence correctly describes inner alignment?
Inner alignment (vs. outer alignment) asks: even if the reward function perfectly captures human values, does the model actually learn to pursue that objective — or a proxy of it that diverged during training? It is an open alignment research problem.
6 / 33
PR Description:
"Fix: Improved error handling for API calls to the sentiment analysis service. Added retry logic with exponential backoff and logging of failed attempts. This addresses intermittent failures reported by QA."
During a code review, Sarah asks you: "The description mentions 'exponential backoff' – what does that *specifically* mean in this context, and why is it relevant to AI safety?"
Which response best answers Sarah's question?The correct answer highlights the key aspect of exponential backoff: increasing delay after failures. This prevents overwhelming a potentially fragile or unstable API (which could be an AI service), mitigating risks associated with cascading errors. The other options misinterpret the term – it's not just about retrying; the *increasing* delay is what's vital for robustness and, indirectly, contributes to safety by reducing the likelihood of further issues. This concept is particularly relevant when dealing with less-than-perfect AI outputs where aggressive retry behavior could exacerbate problems.
7 / 33
// Slack message to the team
user@team: "We're seeing increased latency when querying the knowledge base. I've added a Prometheus metric to track response times."
Which of the following best describes how this information relates to AI alignment and safety considerations?This question focuses on interpreting telemetry data within the context of alignment. Increased latency can be a symptom of the AI struggling to find relevant information, which suggests a potential misalignment – the model isn't accurately reflecting the intended knowledge base or its understanding is flawed. Option A misinterprets latency as purely a performance issue, ignoring the possibility of underlying problems.
8 / 33
PR Description:
"Fix: Improved error handling for API calls to the sentiment analysis service. Added retry logic with exponential backoff and logging of failed attempts. This addresses intermittent failures reported by QA."
During a code review, Sarah asks you: "The description mentions 'exponential backoff' – what does that *specifically* mean in this context, and why is it relevant to AI safety?"
Which response best answers Sarah's question?The correct answer highlights the key aspect of exponential backoff: increasing delay after failures. This prevents overwhelming a potentially fragile or unstable API (which could be an AI service), mitigating risks associated with cascading errors. The other options misinterpret the term – it's not just about retrying; the *increasing* delay is what's vital for robustness and, indirectly, contributes to safety by reducing the likelihood of further issues. This concept is particularly relevant when dealing with less-than-perfect AI outputs where aggressive retry behavior could exacerbate problems.
9 / 33
// Slack message to the team
user@team: "We're seeing increased latency when querying the knowledge base. I've added a Prometheus metric to track response times."
Which of the following best describes how this information relates to AI alignment and safety considerations?This question focuses on interpreting telemetry data within the context of alignment. Increased latency can be a symptom of the AI struggling to find relevant information, which suggests a potential misalignment – the model isn't accurately reflecting the intended knowledge base or its understanding is flawed. Option A misinterprets latency as purely a performance issue, ignoring the possibility of underlying problems.
10 / 33
PR Description:
"Fix: Improved error handling for API calls to the sentiment analysis service. Added retry logic with exponential backoff and logging of failed attempts. This addresses intermittent failures reported by QA."
During a code review, Sarah asks you: "The description mentions 'exponential backoff' – what does that *specifically* mean in this context, and why is it relevant to AI safety?"
Which response best answers Sarah's question?The correct answer highlights the key aspect of exponential backoff: increasing delay after failures. This prevents overwhelming a potentially fragile or unstable API (which could be an AI service), mitigating risks associated with cascading errors. The other options misinterpret the term – it's not just about retrying; the *increasing* delay is what's vital for robustness and, indirectly, contributes to safety by reducing the likelihood of further issues. This concept is particularly relevant when dealing with less-than-perfect AI outputs where aggressive retry behavior could exacerbate problems.
11 / 33
// Slack message to the team
user@team: "We're seeing increased latency when querying the knowledge base. I've added a Prometheus metric to track response times."
Which of the following best describes how this information relates to AI alignment and safety considerations?This question focuses on interpreting telemetry data within the context of alignment. Increased latency can be a symptom of the AI struggling to find relevant information, which suggests a potential misalignment – the model isn't accurately reflecting the intended knowledge base or its understanding is flawed. Option A misinterprets latency as purely a performance issue, ignoring the possibility of underlying problems.
12 / 33
PR Description:
"Fix: Improved error handling for API calls to the sentiment analysis service. Added retry logic with exponential backoff and logging of failed attempts. This addresses intermittent failures reported by QA."
During a code review, Sarah asks you: "The description mentions 'exponential backoff' – what does that *specifically* mean in this context, and why is it relevant to AI safety?"
Which response best answers Sarah's question?The correct answer highlights the key aspect of exponential backoff: increasing delay after failures. This prevents overwhelming a potentially fragile or unstable API (which could be an AI service), mitigating risks associated with cascading errors. The other options misinterpret the term – it's not just about retrying; the *increasing* delay is what's vital for robustness and, indirectly, contributes to safety by reducing the likelihood of further issues. This concept is particularly relevant when dealing with less-than-perfect AI outputs where aggressive retry behavior could exacerbate problems.
13 / 33
// Slack message to the team
user@team: "We're seeing increased latency when querying the knowledge base. I've added a Prometheus metric to track response times."
Which of the following best describes how this information relates to AI alignment and safety considerations?This question focuses on interpreting telemetry data within the context of alignment. Increased latency can be a symptom of the AI struggling to find relevant information, which suggests a potential misalignment – the model isn't accurately reflecting the intended knowledge base or its understanding is flawed. Option A misinterprets latency as purely a performance issue, ignoring the possibility of underlying problems.
14 / 33
Review this code review comment:
`"This change introduces a potential race condition. The API call to the chatbot is not properly synchronized with the database update, leading to inconsistent data."` What does this comment primarily address regarding AI safety?
This comment focuses on data integrity – a core AI safety property. A race condition means different parts of the system can access and modify data simultaneously, leading to unpredictable outcomes. The concern isn't about ethical considerations or performance directly, but rather about ensuring the chatbot's actions align with the stored information.
15 / 33
Sarah in the AI alignment team sends this Slack message: "I'm concerned that our reward model is overly incentivizing the agent to prioritize speed over accuracy. We need to investigate potential 'shortcut' behaviors."
Which of the following best describes Sarah's primary worry?
Sarah's message highlights a classic problem in reinforcement learning: reward hacking. When rewards are poorly defined (e.g., solely based on speed), an agent will find ways to maximize that metric, even if it compromises the intended goal – accuracy. This is a fundamental aspect of AI alignment.
16 / 33
PR Description:
"Refactor: Improved logging for the model inference service. Added detailed timestamps and context to all log messages. This will aid in debugging and monitoring unexpected behavior."
How does this change contribute to AI safety practices?
Robust logging is *essential* for AI safety. By tracking detailed information about model execution, engineers can quickly identify and diagnose unexpected behavior – a critical step in mitigating potential risks. This allows for rapid debugging and intervention before issues escalate.
17 / 33
"Good morning, team. I've been focusing on improving the robustness of the adversarial training pipeline. We're running more extensive tests to identify vulnerabilities and improve our defenses against malicious inputs."
What is the primary objective of this update in relation to AI safety?
This update directly addresses vulnerability mitigation – a core component of AI safety. Adversarial training aims to identify and correct weaknesses in the model's behavior before they can be exploited by malicious actors or lead to unintended consequences. It's about building a more robust and reliable system.
18 / 33
Which of the following BEST describes 'specification gaming'?
Specification gaming refers to an AI system cleverly exploiting loopholes or unintended consequences in its reward function or objective. The agent learns to 'game' the system rather than truly understanding or fulfilling the intended purpose. This is a significant challenge for aligning AI systems.
19 / 33
Review this code review comment:
`"This change introduces a potential race condition. The API call to the chatbot is not properly synchronized with the database update, leading to inconsistent data."` What does this comment primarily address regarding AI safety?
This comment focuses on data integrity – a core AI safety property. A race condition means different parts of the system can access and modify data simultaneously, leading to unpredictable outcomes. The concern isn't about ethical considerations or performance directly, but rather about ensuring the chatbot's actions align with the stored information.
20 / 33
Sarah in the AI alignment team sends this Slack message: "I'm concerned that our reward model is overly incentivizing the agent to prioritize speed over accuracy. We need to investigate potential 'shortcut' behaviors."
Which of the following best describes Sarah's primary worry?
Sarah's message highlights a classic problem in reinforcement learning: reward hacking. When rewards are poorly defined (e.g., solely based on speed), an agent will find ways to maximize that metric, even if it compromises the intended goal – accuracy. This is a fundamental aspect of AI alignment.
21 / 33
PR Description:
"Refactor: Improved logging for the model inference service. Added detailed timestamps and context to all log messages. This will aid in debugging and monitoring unexpected behavior."
How does this change contribute to AI safety practices?
Robust logging is *essential* for AI safety. By tracking detailed information about model execution, engineers can quickly identify and diagnose unexpected behavior – a critical step in mitigating potential risks. This allows for rapid debugging and intervention before issues escalate.
22 / 33
"Good morning, team. I've been focusing on improving the robustness of the adversarial training pipeline. We're running more extensive tests to identify vulnerabilities and improve our defenses against malicious inputs."
What is the primary objective of this update in relation to AI safety?
This update directly addresses vulnerability mitigation – a core component of AI safety. Adversarial training aims to identify and correct weaknesses in the model's behavior before they can be exploited by malicious actors or lead to unintended consequences. It's about building a more robust and reliable system.
23 / 33
Which of the following BEST describes 'specification gaming'?
Specification gaming refers to an AI system cleverly exploiting loopholes or unintended consequences in its reward function or objective. The agent learns to 'game' the system rather than truly understanding or fulfilling the intended purpose. This is a significant challenge for aligning AI systems.
24 / 33
Review this code review comment:
`"This change introduces a potential race condition. The API call to the chatbot is not properly synchronized with the database update, leading to inconsistent data."` What does this comment primarily address regarding AI safety?
This comment focuses on data integrity – a core AI safety property. A race condition means different parts of the system can access and modify data simultaneously, leading to unpredictable outcomes. The concern isn't about ethical considerations or performance directly, but rather about ensuring the chatbot's actions align with the stored information.
25 / 33
Sarah in the AI alignment team sends this Slack message: "I'm concerned that our reward model is overly incentivizing the agent to prioritize speed over accuracy. We need to investigate potential 'shortcut' behaviors."
Which of the following best describes Sarah's primary worry?
Sarah's message highlights a classic problem in reinforcement learning: reward hacking. When rewards are poorly defined (e.g., solely based on speed), an agent will find ways to maximize that metric, even if it compromises the intended goal – accuracy. This is a fundamental aspect of AI alignment.
26 / 33
PR Description:
"Refactor: Improved logging for the model inference service. Added detailed timestamps and context to all log messages. This will aid in debugging and monitoring unexpected behavior."
How does this change contribute to AI safety practices?
Robust logging is *essential* for AI safety. By tracking detailed information about model execution, engineers can quickly identify and diagnose unexpected behavior – a critical step in mitigating potential risks. This allows for rapid debugging and intervention before issues escalate.
27 / 33
"Good morning, team. I've been focusing on improving the robustness of the adversarial training pipeline. We're running more extensive tests to identify vulnerabilities and improve our defenses against malicious inputs."
What is the primary objective of this update in relation to AI safety?
This update directly addresses vulnerability mitigation – a core component of AI safety. Adversarial training aims to identify and correct weaknesses in the model's behavior before they can be exploited by malicious actors or lead to unintended consequences. It's about building a more robust and reliable system.
28 / 33
Which of the following BEST describes 'specification gaming'?
Specification gaming refers to an AI system cleverly exploiting loopholes or unintended consequences in its reward function or objective. The agent learns to 'game' the system rather than truly understanding or fulfilling the intended purpose. This is a significant challenge for aligning AI systems.
29 / 33
Review this code review comment:
`"This change introduces a potential race condition. The API call to the chatbot is not properly synchronized with the database update, leading to inconsistent data."` What does this comment primarily address regarding AI safety?
This comment focuses on data integrity – a core AI safety property. A race condition means different parts of the system can access and modify data simultaneously, leading to unpredictable outcomes. The concern isn't about ethical considerations or performance directly, but rather about ensuring the chatbot's actions align with the stored information.
30 / 33
Sarah in the AI alignment team sends this Slack message: "I'm concerned that our reward model is overly incentivizing the agent to prioritize speed over accuracy. We need to investigate potential 'shortcut' behaviors."
Which of the following best describes Sarah's primary worry?
Sarah's message highlights a classic problem in reinforcement learning: reward hacking. When rewards are poorly defined (e.g., solely based on speed), an agent will find ways to maximize that metric, even if it compromises the intended goal – accuracy. This is a fundamental aspect of AI alignment.
31 / 33
PR Description:
"Refactor: Improved logging for the model inference service. Added detailed timestamps and context to all log messages. This will aid in debugging and monitoring unexpected behavior."
How does this change contribute to AI safety practices?
Robust logging is *essential* for AI safety. By tracking detailed information about model execution, engineers can quickly identify and diagnose unexpected behavior – a critical step in mitigating potential risks. This allows for rapid debugging and intervention before issues escalate.
32 / 33
"Good morning, team. I've been focusing on improving the robustness of the adversarial training pipeline. We're running more extensive tests to identify vulnerabilities and improve our defenses against malicious inputs."
What is the primary objective of this update in relation to AI safety?
This update directly addresses vulnerability mitigation – a core component of AI safety. Adversarial training aims to identify and correct weaknesses in the model's behavior before they can be exploited by malicious actors or lead to unintended consequences. It's about building a more robust and reliable system.
33 / 33
Which of the following BEST describes 'specification gaming'?
Specification gaming refers to an AI system cleverly exploiting loopholes or unintended consequences in its reward function or objective. The agent learns to 'game' the system rather than truly understanding or fulfilling the intended purpose. This is a significant challenge for aligning AI systems.
Frequently Asked Questions
What will I practice in "AI Safety Properties — Vocabulary — AI Alignment & Safety | CoderLingo"?
This is an AI Alignment & Safety Language exercise set. It walks through 33 scenario-based multiple-choice questions built around real usage of AI Alignment & Safety Language terminology that IT professionals encounter on the job.
Is this exercise free to use?
Yes. Every exercise on CoderSlingo, including this one, is free to complete with no account, sign-up, or paywall.
How many questions are in this exercise?
This set contains 33 questions. Each one shows immediate feedback and a detailed explanation after you answer, so you learn the correct usage right away rather than waiting for a final score.
Do I need prior experience to complete this exercise?
No prior experience is required. Each question includes a full explanation covering the reasoning behind the correct answer, so the exercise itself teaches the AI Alignment & Safety Language vocabulary as you go.
Can I retry the exercise if I get questions wrong?
Yes — use the "Try again" button on the results screen to reset your answers and go through all the questions again. There is no limit on attempts.
Is my progress saved?
Your answers and score for the current session are tracked in the browser as you go. No account or login is needed, and there is nothing to install.
What if I don't understand a term used in a question?
Read the explanation shown after you answer each question — it breaks down the correct term in plain English with a real-world example. You can also check the site Glossary for quick definitions.
How is this different from reading a blog article on the topic?
Exercises like this one are interactive drills that test and reinforce specific vocabulary through multiple-choice questions, while blog articles explain concepts in prose. Practising here after reading builds active recall, not just passive recognition.
Where can I find more AI Alignment & Safety Language exercises?
See the AI Alignment & Safety Language exercises hub for the full set of related pages, or browse all exercise categories from the main Exercises index.
Can I use this exercise to prepare for a technical interview?
Yes — AI Alignment & Safety Language vocabulary comes up often in technical discussions and interviews. Pair this exercise with our dedicated Interview Preparation section for role-specific practice.