Aegis-GPUTrustworthy AI
10% of the NCA-GENL exam. Read each objective's key points, open “What NVIDIA says” to see the source, then practise.
Core Machine Learning and AI Knowledge · Software Development · Experimentation · Data Analysis and Visualization · Trustworthy AI
5.1 Principles of trustworthy AI
What trustworthy AI means and NVIDIA's principles.
Key points
NVIDIA's Trust Center lists its guiding principles as privacy, transparency, nondiscrimination, and safety and security.
What NVIDIA says (1)
“We work to maintain our guiding principles of privacy, transparency, nondiscrimin”
NVIDIA defines trustworthy AI as an approach that prioritizes safety and transparency for the people who interact with it.
What NVIDIA says (1)
“Trustworthy AI is an approach to AI development that prioritizes safety and transparency for the people who interact with it.”
NVIDIA says trustworthy AI models are transparent, providing information such as accuracy benchmarks or a description of the training dataset.
What NVIDIA says (1)
“They’re also transparent — providing information such as accuracy benchmarks or a description of the training dataset”
Key terms: Trustworthy AI
5.2 Privacy and consent
Using data while protecting the people behind it.
Key points
NVIDIA says developers of models that rely on data such as a person's image, voice, artistic work or health records should evaluate whether individuals gave appropriate consent.
What NVIDIA says (2)
“should evaluate whether individuals have provided appropriate consent for their personal information to be used in this way.”
“Developers of AI models that rely on data such as a person’s image, voice, artistic work or health records”
Federated learning builds models with a consortium of data providers. NVIDIA says it reduces risk to data security and privacy because the data never leaves the individual sites.
What NVIDIA says (2)
“Federated learning is a way to develop and validate AI models from diverse data sources while mitigating the risk of compromising data security or privacy, as the data never leaves individual sites.”
“It enables AI models to be built with a consortium of data providers”
NVIDIA's privacy principle states AI should comply with privacy laws and regulations and meet societal norms for personal data and information privacy.
What NVIDIA says (1)
“AI should comply with privacy laws and regulations, and meet societal norms for personal data and information privacy.”
Key terms: Personally identifiable information Federated learning
5.3 Tools that make AI more trustworthy
Guardrails, model cards, confidential computing and explainability.
Key points
NeMo Guardrails lets enterprise developers set boundaries for their large language model (LLM) apps. Topical guardrails keep chatbots on specific subjects. Safety guardrails limit the language and data sources the app uses.
What NVIDIA says (3)
“Topical guardrails ensure that chatbots stick to specific subjects.”
“Safety guardrails set limits on the language and data sources the apps use in their responses.”
“keeps AI language models on track by allowing enterprise developers to set boundaries for their applications.”
A model card gives developers and users a clear, concise view of a model's capabilities. NVIDIA's Model Card++ adds four subsections: Bias, Explainability, Privacy, and Safety and Security.
What NVIDIA says (2)
“Four subsections detailing model-specific information concerning Bias, Explainability, Privacy, and Safety and Security.”
“They provide both developers and downstream users and beneficiaries with a clear understanding of an AI model’s capabilities in a clear and concise format.”
Confidential computing encrypts application code, models and data while in use, not just at rest or in transit, using hardware-rooted Trusted Execution Environments (TEEs).
What NVIDIA says (2)
“Confidential computing secures AI workloads by encrypting application code, models, and data in use—not just at rest or in transit.”
“It does this by creating hardware-rooted Trusted Execution Environments (TEEs) that are enforced by cryptographic attestation.”
NVIDIA defines explainable AI (XAI) as tools and techniques that help people better understand why a model makes certain decisions and how it works.
What NVIDIA says (1)
“is a set of tools and techniques used by organizations to help people better understand why a model makes certain decisions and how it works.”
Key terms: NeMo Guardrails Confidential computing Explainable AI Model card / Model Card++
5.4 Reducing bias
Where unwanted bias comes from and how to reduce it.
Key points
AI models are often trained on data limited in size, scope and diversity. NVIDIA says reducing unwanted bias is important so all people and communities can benefit.
What NVIDIA says (2)
“AI models are trained by humans, often using data that is limited by size, scope and diversity.”
“To ensure that all people and communities have the opportunity to benefit from this technology, it’s important to reduce unwanted bias in AI systems.”
NVIDIA says trustworthy AI developers look for clues and patterns that suggest an algorithm is discriminatory, and that synthetic datasets are one way to reduce unwanted bias in training data.
What NVIDIA says (2)
“trustworthy AI developers mitigate potential unwanted bias by looking for clues and patterns that suggest an algorithm is discriminatory”
“Synthetic datasets offer one solution to reduce unwanted bias in training data”
Key terms: Unwanted bias