High level notes:
Technical working session #1
Tue, 18 Nov 25
Technical Discussion on AI Creation Tools
- Technical workflow challenges for creators
- Current process requires multiple platforms ($50+ across different services)
- Need single application for end-to-end content creation
- API integration complexity main barrier for non-technical users
- Service engineering vs. user experience priorities
- Backend can use agents to connect multiple APIs
- Frontend needs intuitive interfaces for creators without technical skills
- Intent expression as core design challenge
- Language models parse user intent better with specificity
- Need alternatives to text prompts (biometric signals, eye tracking)
- Golf analogy: choose broad or narrow context scope based on creation stage
Content Personalization and Customization
- Blockchain-enabled personalized content creation
- Mood manipulation through customized imagery
- Real-time biometric feedback for content generation
- Research applications (cannabis studies using mood-based content)
- Customized video material understanding individual users
- Context windows in language/diffusion models for personalization
- SEO landing page personalization as current example
- Black Mirror-style future of completely personalized media feeds
Attribution and Rights Management
- Open source vs. copyrighted content handling
- 50,000+ public domain books used for training
- Tag system for attribution to original creators
- Spin-offs model: 10% royalty to original authors for style-based generation
- Cameo-style consent models for IP
- Sora app’s cameo feature as template for on-chain IP representation
- Voice, video, and AI model licensing opportunities
- Monetization chains around individual IP consent
Future of Human vs. AI Content
- Value hierarchy predictions
- AI-generated content becomes commodity ($12 → free)
- Human-created content becomes luxury ($12 → $100)
- Proof systems needed to verify human creation
- Detection and watermarking challenges
- Current fingerprinting works but can be removed
- Industry collaboration (Sony, Canon) on content tagging
- Frequency-based signatures in AI-generated audio/video
- Market dynamics
- “Something better than nothing” philosophy for imperfect AI solutions
- Attribution as language system (using symbolic rather than deterministic mapping)
- Agents as backend for connecting complex tool ecosystems
Chat with meeting transcript: https://notes.granola.ai/d/6b710799-1160-4977-ab7a-ba030b5f64e9