Live Fireside Chat
summary
Covers concrete progress: Luma AI’s Dream Machine launches late June 2024, hits about 25 million users and cuts generation for a 10‑second clip from ~200 seconds to 19.2 seconds, backed by a NeRF capture app that supplies vast 3D data; Meta’s MovieGen publishes a detailed paper on scaling to ~30B parameters with character‑consistent personalization and multi‑modal outputs. Also covers evaluation (human eval, A/B testing, red‑teaming) and current safety tradeoffs.
AI-written summary · about
chapters (13)
- 0:00Generative video revolution
- 1:53Amit’s journey to Dream Machine
- 8:17Movie Gen’s generative video
- 10:27Quality in AI video models
- 15:43User feedback on Dream Machine
- 17:52Movie Gen’s beta and community
- 19:20Future of generative video
- 22:52Peter’s vision for Movie Gen
- 24:17Safety in generative video
- 28:53Technical challenges beyond safety
- 31:00Why generative video is still early and evolving
- 34:10Magic wand for generative video
- 36:40Closing thoughts
related talks
23mFireside Chat: How AI is Shaking up Game Development in 2024
35mFireside Chat
19mDream Machine: Scaling to 1m users in 4 days
25mHow Modern PyTorch Supercharges Multimodal Training and Inference at Luma AI
32mFireside Q&A
40mPannel Discussion | From Hype To Value: Real World Applications Of Generative AI And LLMs
24mFIRESIDE CHAT: The Future of AI Infra: The Opportunities and Challenges That Await Us On Our Journey
01h 03mIt's All About AI