Chapter 1

Inside the machine's mind

Chapter 2

The heavy metal of artificial intelligence

Chapter 3

The daily grind in the AI gold rush

Chapter 4

Rules, red teams, and the global guardrails

The battle between open-source and closed-source AI

sonia Open-source AI shares its underlying code and weights freely, while closed-source keeps them secret.

Oct 6

Quiz

What is the primary goal of establishing "guardrails" on an AI model?

Oct 6

Quiz

What does the term "frontier model" refer to in the AI industry?

Oct 6

Quiz

Which global region passed the "AI Act" in 2024, creating the first comprehensive legal framework?

Oct 6

Rules, red teams, and the global guardrails
BETA

sonia Sort AI safety rules to discover how red teams and global guardrails protect us.

Oct 6

The European Union AI Act and global rules

sonia The European Union AI Act is the world's first comprehensive legal framework for artificial intelligence.

Oct 6

Quiz

From which fields did the term "red teaming" originate?

Oct 6

Quiz

What is an example of an AI practice banned under the European Union AI Act?

Oct 6

Quiz

What is the main difference between open-source and closed-source AI models?

Oct 6

sonia AI Red Teaming and Guardrails

We break the system just to build it back

Oct 6

sonia The US AI Safety Institute monitors massive supercomputer training runs of systems like GPT-4. This oversight prevents catastrophic risks from technology that surpasses all existing human knowledge

Oct 6

Quiz

Why do some companies choose to keep their advanced AI models closed-source?

Oct 6

Quiz

What is the practice of intentionally attacking an AI system to find its flaws called?

Oct 6

Fun Fact

Red team roots: The term 'red team' comes from Cold War military simulations where the red team acted as the opposing force.

Oct 6

Glossary
red teaming (noun)
The practice of systematically challenging plans, policies, or systems using adversarial tactics to find vulnerabilities. The U.S. AI Safety Institute published guidelines on using these simulations to test large language models.

Oct 6

Gossip

Should all advanced AI models be completely open-source?

Oct 6