Why NVIDIA?

Introduction

I'm Minh, an engineer with more than 7 years of experience in Automotive embedded software. Currently, I'm test manager and CI developer for a inter-platform communication framework call IPCF. I'm passionate about AI, test driven development and automation test.

I want to join NVIDIA because I want to grow my career in AI field, encourage test driven development mindset and for bigger picture is to make sure AI is safe for human.

Mr. Jensen Huang's Vision

• AI as a Foundational Infrastructure
• From Science to Application
• Physical AI and Robotics
• Massive Growth in Computing Power
• Democratization of AI / Access
• Ethics, Sovereignty, Oversight
• Building the Ecosystem: Hardware + Software + Platforms

Ms. Elena Shao

• Data center, cloud, HPC, and AI offerings.
• Datacenter SW/FW, CUDA, HPC, GenAI, NIM/Blueprints, DGX Cloud, Aerial

Questions

• What is your perspective on the current challenges our software QA team is facing? And you as a leader want to lead the team to solve it?
• In your point of view. What is the current challenges that our team as a software QA is facing?
• Standards
  • I read a know that NVIDIA made a commit about Trust Worthy AI. Are we trying to follow any standards like ISO of IEEE? Or we are going to make our own standards?

My vision

• AI do the implementation, Human handle requirements, validate and verify.
• Tester is about keeping doubting about the truth and asking the right questions.
• Shift left strategy, Test driven development.
• Faster testing, Faster development iteration.
• Make sure human safety.
  • AI right now is still 0 or 1. Like if we fix the temperature and seed for random, we can get the same result.
  • Just my guess, but I think the mystery about life force is the key different between Human and AI.
  • I just have a guess that the randomness based on probability in quantum physics is the key to understand life force, consciousness, soul, afterlife.
  • Our existence can be threatened if AI can generate randomness like human brain. Maybe it can apear in the combination of quantum computer and NVIDIA GPU.

My mission

• Grow up with my children.
• Practice to ask the right questions.
• Understand AI enough to use it effectively and not be afraid of it.
• Contribute to the community by encouraging everyone to have a tester mindset.

My choice

• To be a teacher follow Waldorf Steiner education method.
  • It emphasizes the holistic development of children—head, heart, and hands—by balancing intellectual, artistic, and practical activities.
  • I was admitted by a Waldorf school in Vietnam. Actually, I quit my current job for this purpose.
• To join NVIDIA as a software QA.
  • I don't know if I would be accepted or not.
  • Even if I am accepted by NVIDIA and my inner balance lean on NVIDIA job. I will still need to bring this up with the school council. If the school say they need me, then I will keep my word to be a teacher.
• I'm a bit kind of a dreamer. But that's always the thing that drive me forward up until now.
• So I will wait for the answer from NVIDIA and discuss it with the school council.

NVIDIA's approach to Trustworthy AI is a comprehensive framework focused on making AI development and deployment more safe, transparent, and responsible.

Here's a summary of their key principles and technological solutions:

Core Principles

Trustworthy AI, for NVIDIA, is about prioritizing the safety and transparency for people who interact with AI systems. They recognize the real-world impact of AI and aim to channel its power responsibly. This means going beyond legal compliance to proactively address potential risks.

Key Technological and Platform Initiatives

1. Safety and Guardrails:

   • NVIDIA NeMo Guardrails: This software is used to set boundaries for enterprise generative AI applications. It helps keep AI language models on track by implementing topical guardrails (ensuring the model sticks to specific subjects) and safety guardrails (setting limits on the language and data sources used).

2. Privacy and Security:

   • Confidential Computing: Built into their hardware like the NVIDIA H100 and H200 Tensor Core GPUs, this technology protects sensitive data while it's in use. It uses hardware-based security to ensure unauthorized entities can't view or modify data or applications while they are running.
   • Federated Learning: To help institutions like hospitals and banks train robust models while keeping data private, NVIDIA provides technology like NVIDIA FLARE. This allows researchers to train AI models on decentralized data from multiple organizations without the confidential information ever leaving the company's private servers.

3. Transparency and Explainability:

   • Model Cards (Model Card++): On their software hub, NGC, NVIDIA uses a detailed format called Model Card++ to provide extensive information about each AI model. This includes details on the datasets used, training methods, performance measures, licensing information, and specific ethical considerations.
   • Retrieval-Augmented Generation (RAG): This technique enhances transparency in generative AI by connecting models to authoritative external databases. This allows the models to cite their sources for the information they provide, leading to more accurate and verifiable answers.

4. Mitigating Bias:

   • NVIDIA encourages developers to look for patterns that suggest an algorithm is discriminatory or involves the inappropriate use of certain characteristics, actively working to mitigate potential unwanted bias beyond just following anti-discrimination laws.
   • Tools like NVIDIA Omniverse Replicator can be used to generate synthetic data, which can help developers create diverse and balanced datasets to train AI models, further helping to reduce bias.

In essence, NVIDIA's strategy is to integrate trustworthy principles across their entire end-to-end AI development stack, providing developers and enterprises with both the hardware and software tools necessary to build and deploy AI systems that are not only powerful but also reliable and ethical.