Learning Machines Project Brief
Project Identity
Learning Machines: Text, Images, Video is a CC Fest creative AI / machine learning camp for educators, artists, students, and curious learners, framed as three core sessions + optional studio.
The camp is not a prompt engineering workshop. It is an investigation into how generative systems work across three modalities:
| Session | Focus | Core Question | Artifact |
|---|---|---|---|
| 1 | Text | How do language models generate text that feels meaningful, and what are they actually doing? | Text Experiment Board |
| 2 | Images | How do image models turn words into pictures, and what defaults appear when prompts are vague? | Image Default Test Board |
| 3 | Video | Why is video generation harder than image generation, and what does that tell us about how models work? | Video Test Report |
| Optional Studio | Make, Teach, Investigate, Explain, or Critique | What did we learn about generative AI by making, testing, teaching, explaining, or critiquing it? | Final artifact, classroom adaptation, model-behavior investigation, explainer, or critique |
The intellectual spine:
Text is sequential. Images are spatial. Video is spatiotemporal.
The repeated question:
What is the machine actually doing?
The tool principle:
Every tool should make something invisible visible.
Origin
The camp grows out of two bodies of work.
Generative AI as a Creative Collaborator
The Campbell Hall Gen AI course developed the core curriculum:
- ELIZA vs. LLM comparisons
- next-token prediction activities
- Default Tests for image generation
- diffusion and latent-space activities
- video generation investigations
- Temporal Telephone
- A/B/C prompt testing
- student posters and model-behavior investigations
The strongest student-facing activities were robust enough to adapt for adults, especially educators who want classroom-ready ways to discuss AI without reducing it to hype or panic.
CC Fest Coding Camp
The camp inherits the CC Fest model:
- free, virtual, community-centered workshops
- beginner-friendly but intellectually serious facilitation
- Notion workspace and recap workflow
- GitHub Pages tool hosting
- interactive single-file HTML tools
- guest speakers
- assignment tiers
- open studio and showcase
- asynchronous access through recordings and notes
Pedagogical Pattern
Every content session follows the same six-move structure:
- Experience the concept unplugged
- See the mechanism
- Choose a participation pathway
- Run a structured investigation
- Document evidence
- Reflect ethically and creatively
Participants should leave with a repeatable investigation protocol:
hypothesis -> controlled A/B/C test -> output comparison -> evidence-based claim -> ethical/pedagogical reflection
Launch MVP Surface
The first-cohort repo is organized around launch-ready browser tools across text, images, video, and optional studio work:
- Session 1 tools for tokens, next-token prediction, probability, temperature, and ELIZA-style pattern matching
- Session 2 tools for pixels, features, diffusion, visual defaults, and prompt guidance
- Session 3 tools for temporal drift, video failure modes, and frame-by-frame coherence
- Cross-session and studio tools for evidence boards, model cards, access audits, confidence checking, and recap walls
Remaining stretch tools stay visible as future directions, but the launch cohort does not depend on them.
Each content session has a concise facilitation note in docs/, plus a worksheet that supports direct AI use, observation/critique, teaching/design, build/code, and critical/no-AI pathways.
Ethics and Consent
Ethics is structural, not topical. It appears in every session, tool, worksheet, recap, and final project.
Core commitments:
- AI use is optional, visible, and discussable.
- Pre-generated examples are available when participants do not want to use AI tools directly.
- Low-AI and no-AI pathways are first-class, not fallback options.
- Public recaps and screenshots require consent and human review.
- Model defaults are treated as evidence of data, design, culture, and platform decisions.
Key line:
Opting out of direct AI use should never mean opting out of the camp.
Final Project Pathways
Participants choose one:
| Pathway | Final Artifact |
|---|---|
| Make | Creative text/image/video artifact with process evidence |
| Teach | Classroom-ready activity, handout, or facilitation plan |
| Investigate | Model behavior experiment or Default Test |
| Explain | Interactive tool, poster, zine, or concept bridge |
| Critical / No-AI | Critique, consent checklist, model card, policy, or unplugged activity |
All pathways require evidence of iteration and reflection on what the model did, what the human did, and what the system assumed.
Launch Positioning
Full description:
Learning Machines: Text, Images, Video is a beginner-friendly creative AI workshop for educators, artists, students, and curious learners. Across three core sessions, we explore how generative AI systems create text, images, and video. Participants experiment with prompts, compare outputs, investigate model defaults and failures, and reflect on the role of human judgment in AI-assisted creativity. An optional studio supports final artifacts, classroom activities, model-behavior investigations, explainers, and critiques based on what participants learned. No coding or machine learning background required.
Short description:
How do AI systems write, imagine, and generate motion? In this creative AI camp with three core sessions and optional studio, we explore text generation, image generation, and video generation through hands-on experiments and reflective making. Participants test prompts, compare outputs, investigate defaults and failures, and may create a final project, classroom activity, explainer, or critique to share in a closing studio. Beginner-friendly. No coding required. Great for educators, artists, students, and curious learners.