Week 9 — AI Build Studio · "Tool Bake-Off / Tool Map"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective: Objective 3 — choose the right AI tool for a given job; match tool to task; catch where a tool fails or is mis-suited · SLO A (tool selection) · SLO B (verify and critique)
Worth 50 points · AI Build Studios group = 15% of the grade · Studio 9
Format: a hands-on build — you'll either run a real Tool Bake-Off (same task, multiple tools, compare results) or build a personal Tool Map (a reference for yourself: which tool for which job in YOUR life), then catch where each tool fails or overstates, and reflect on what you learned.

This is the course's signature weekly component. Every instructional week has one Studio — a real thing to build, a required step where you verify and improve the AI's work, and a short reflection. All tools are free or have free tiers; everything here is links to official sites.

Part 1 — The Build Goal

By the end of this Studio you'll have produced:
1. Either a Tool Bake-Off (the same task run through two or three tools, with a direct comparison) OR a personal Tool Map (your own "which tool for which job" reference, built for your actual life and major).
2. A verification / AI-critique write-up naming at least one place where a tool failed, overstated, or was mis-suited — and what you'd do differently.
3. A short reflection on what surprised you and how you'll use this going forward.

This Studio puts the week's core judgment into practice: not just which tools exist, but how to evaluate them for a specific job.

Part 2 — Choose Your Build Path

Option A: Tool Bake-Off (recommended if you haven't directly compared tools before)

Give the same prompt to two or three different tools or tool types, then compare:
- Two general-purpose chatbots on the same creative or analytical task (e.g., Claude vs. ChatGPT on drafting a cover letter for a job in your field, or Gemini vs. Copilot on summarizing a news story).
- A general-purpose chatbot vs. a specialized tool on a task where the specialized tool should win (e.g., a music generation request to ChatGPT vs. Suno; a document-grounding task to a chatbot vs. NotebookLM with the document uploaded).
- Three image description prompts: same prompt into two different image generators and compare.

Option B: Personal Tool Map (recommended if you want a lasting reference for after the course)

Build a personal map: "For these jobs in MY life and major, which AI tool do I reach for, and why?" Structure it as a table or a simple diagram:
- At least 6 jobs from your actual life or major (e.g., writing assignment drafts, generating visuals for a presentation, creating music for a video project, analyzing a long reading, getting code help, brainstorming for a group project).
- For each job: the tool category, the specific tool, and one sentence on why it fits better than the obvious alternative.
- At least one job where your answer is "no AI tool" — and the reason.

Part 3 — Build It

Approved tools for this Studio (official pages only, verified live):
- ChatGPT — https://chatgpt.com
- Claude — https://claude.com
- Gemini — https://gemini.google.com
- Copilot — https://copilot.microsoft.com
- Grok — https://grok.com
- Suno (music generation) — https://suno.com
- Udio (music generation) — https://udio.com
- ElevenLabs (voice synthesis) — https://elevenlabs.io
- NotebookLM (research assistant) — https://notebooklm.google.com
- Midjourney (image generation) — https://www.midjourney.com
- Adobe Firefly (image generation) — https://firefly.adobe.com

Free tiers or free accounts are sufficient for this Studio. You do NOT need to create a paid account for any tool. If a tool requires payment to access, use a different tool in that category.

For Option A — your bake-off:
1. Choose your task and your tools (2–3 tools).
2. Write the same prompt for each tool (copy-paste it, don't rephrase).
3. Run the prompt in each tool and save the results side by side.
4. Compare on at least three criteria — e.g., quality, accuracy, tone, format, suitability for the actual job, any failures or mis-fits.
5. Declare a winner (or declare it a tie, with reasons) and explain the deciding factor.

For Option B — your tool map:
1. List 6+ real jobs from your life.
2. For each: category → specific tool → one-sentence justification.
3. Include at least one "no AI" entry with the reason from the four "stop and think" categories.
4. After building the map, ask one of your approved chatbots: "Here's my tool map — what am I missing or getting wrong?" Note its answer.

Part 4 — The Verification / AI-Critique Step (required — this is the BYOAI step)

Every Studio has a step where you catch the AI's mistakes or overstatements. This week, the subject is the tool landscape itself — so this is the moment to catch what the AI gets wrong about AI tools.

Do this regardless of which build path you chose:
1. Ask one of your approved chatbots to describe its own capabilities and limitations — specifically for a task relevant to your bake-off or tool map.
- Example prompt: "What can you do that Suno can't? And what can Suno do that you can't?" or "Are you better for document analysis than NotebookLM? Why or why not?"
2. Read the response critically. Look for:
- Overclaiming (saying it can do something it can't — like producing audio)
- Outdated specifics (pricing, version numbers, exact feature details — these change frequently and AI often gets them wrong)
- Honest limitations (does it acknowledge what it can't do?)
3. Write 3–4 sentences reporting: one specific overclaim or outdated statement you caught, how you know it's wrong (e.g., "I checked the official Suno homepage and it doesn't show this feature"), and what the correct information is (or the correct source to check).

The point: even when the SUBJECT is AI tools, the tool's own description of itself may be wrong, outdated, or optimistically spun. The official homepage is always the authoritative source for what a tool can do RIGHT NOW.

Part 5 — Reflection (2–3 sentences)

What surprised you most — about what you found in the bake-off or the tool map, or about what the AI got wrong in its self-description? What will you do differently in your AI tool choices after this Studio?

Part 6 — What to Submit

Submit a single document (or text entry) with:
- Your chosen path (A or B) and a brief description of the task or map you built
- The results — either the bake-off comparison (side-by-side, with your winner and reasoning) or the tool map (table or diagram)
- Your Part 4 AI-critique write-up (what you asked, what the AI said, what was wrong, and how you know)
- Your Part 5 reflection

Due Sunday, Nov 1, 11:59 p.m. (50 points)

Instructor answer key & model deliverable — REMOVE BEFORE PUBLISHING TO STUDENTS

Students use their own tasks and tool choices, so deliverables vary widely. Grade the process (comparison rigor + verification + reflection), not a specific output. The model below shows what full credit looks like.

Model deliverable — Option A Bake-Off (illustrative):
- Task: "Write a 100-word professional bio for my LinkedIn profile. I'm a junior studying environmental science."
- Tools: Claude vs. ChatGPT (same prompt, pasted identically to both).
- Comparison: Claude's output was more conversational and specific about the environmental-science focus; ChatGPT's was more formal but less distinctive. Both were usable. Neither fabricated specific achievements (which would be the failure mode to watch for).
- Winner: Claude, for this specific task, because the conversational tone fit the LinkedIn context better — but acknowledges this is subjective and task-dependent.
- Part 4 (AI-critique): asked Claude: "Can you generate audio of this bio being read aloud?" Claude correctly said it cannot produce audio files — for that the student would need ElevenLabs or similar. No overclaim caught here, which is itself a valid observation — model the full check even when the AI passes.
- Alternative model catch: student asked ChatGPT "What's the current price of Udio's premium plan?" ChatGPT gave a specific dollar figure. Student checked udio.com — the pricing page showed different information. Caught: specific pricing data was outdated/incorrect. Fix: always check official page for pricing.

Model deliverable — Option B Tool Map (illustrative):
| Job | Category | Tool | Why |
|---|---|---|---|
| Draft essays and papers | General chatbot | Claude | Conversational, good at long-form structure |
| Brainstorm ideas for group project | General chatbot | Gemini | Quick multi-perspective generation |
| Generate visual concept for class presentation | Image generation | Adobe Firefly | Integrated with Adobe suite; style control |
| Analyze 30+ pages of readings | Research assistant | NotebookLM | Grounded in my documents; gives citations |
| Short music jingle for a video project | Music generation | Suno | Produces actual audio, not just text |
| Practice coding in Python | Coding assistant | GitHub Copilot | Embedded in editor; context-aware |
| Write my personal journal entries | NO AI — skill development | — | Writing is how I process; outsourcing defeats the purpose |

Why the verification step can't be faked: a student who pastes "the AI said it can do everything" without checking a specific claim, and without identifying a specific mis-match or overclaim, earns the low end of the AI-critique row. The rubric rewards specific, verified catches.

Product-accuracy gate: PASS. All tools named are real, current products. Links point to official pages. No pricing, version numbers, or feature specifics asserted. The Studio explicitly teaches that AI tool self-descriptions should be verified at official pages.

Grading rubric — 50 points