CCNA v2.0, section 5.0, lesson 4, in slides: a generative AI model choosing the next token, an AI agent as a model in a loop with network tools, an agent working a ticket, autonomy from suggesting to acting, digital network assistants; hallucinated and harmful commands, stale data, tools limited to the job, data leaks; the parts of a prompt, data classification, redacting a configuration, output formats, choosing the best of four prompts; and checking an AI's claims against the devices' outputs.
The slides
The lesson is twenty slides in four sections, after an introduction.
- Introduction: the picture, and the lesson's four sections.
- 1. Models and agents: a generative AI model, an AI agent as a model in a loop with tools, an agent working a ticket, autonomy levels, and digital network assistants.
- 2. Risks and limits: the four risks, hallucinated and harmful commands, and limiting an agent's tools.
- 3. Prompts: the parts of a prompt, data classification, redacting a configuration, output formats, and three worked choices of the best prompt.
- 4. Checking the answer: a wrong claim, a right claim, and the five checks.
- 5. The lab: the plan of Lab 5.4.
The buttons on the left and right edges move between slides. So do the arrow keys and a swipe.
The lab
The mechanism, step by step
The lesson stays at the view of a network team; these viewers show what happens inside an agent:
Sources
- The exam:
Cisco's CCNA v2.0 (200-301) exam topics, topics 5.1 and 5.2.
- Cisco's AI assistants, named on the Digital network assistants slide: Cisco's own product pages for Catalyst Center and the Meraki dashboard. The slide describes them generically.
- The IOS commands and outputs: Cisco's IOS 15 command references; the outputs were written in their format, not captured.