Hier is Vincent. Eén en al oor. [Closing Key note]
summary
Claims a Digital Twin built from 800+ Van Gogh letters, uses LLMs plus topic‑context emotion mapping to produce time‑sensitive responses, and links answers to an art database and geolocated museum spots. A pilot runs with expert validation, legal and copyright review, privacy‑by‑design controls, and multilingual output; balance of temperature and dialogue tuning aims to avoid hallucination. Watch if the concrete engineering, validation process, usage numbers, and museum integration matter.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:54Why create digital Van Gogh
- 3:00AI conversation project inspiration
- 4:26Building team and partnerships
- 6:00Legal and ethical AI heritage
- 7:27Vincent’s emotional profile
- 10:00Refining emotions with models
- 12:48Natural dialogue design
- 14:39Balancing historical language
- 16:04Adjusting model temperature
- 17:29Vincent’s digital twin features
- 19:55Pilot results and feedback
- 21:42Legal risks in AI heritage
- 23:30Expanding to other heritage
- 24:55Future plans and scaling
- 28:31Q&A: Paintings enrich responses
- 29:57Q&A: Data sources and validation
- 30:18Q&A: Ethical implications
- 32:23Q&A: Preventing nonsense output
- 33:06Q&A: Stress testing AI
- 33:49Closing remarks
related talks
14mMachine Hallucinations
49mGenerating the Future
01h 12mArtist Keynote
32mEverything you wanted to know about VLMs but were afraid to ask
26mReducing Hallucinations and Evaluating LLMs for Production
30mNeurosymbolic Architecture in Production with Neo4j, GDS, Node2Vec & LLM Grounding
40mPannel Discussion | From Hype To Value: Real World Applications Of Generative AI And LLMs
17mTechnologies Enabling Real-Time Interaction With Digital-Humans