Encounters of the Third Kind With the Multimodal Generative AI Revolution
summary
Claims that generative models recreate humanlike system one creativity get concrete here: recent projects include dead Jennifer's and the Lost landscape project that produce high‑res latent visualizations. Describes Dali 2 using clip to encode text, a prior translating modalities, and diffusion generators trained to denoise Gaussian noise back to images. Notes scale improvements (more neurons/parameters yield qualitative jumps), concrete limits like concept bleed and inconsistent object counts, and rapid, working demos of text→image and early text→video.
AI-written summary · about
chapters (14)
- 0:00Age of super subconscious
- 1:38Generative AI revolution
- 6:16DALL·E 2 and human cognition
- 11:30AI journey to essence
- 15:40AI demystifying creativity
- 18:51Super subconscious age
- 23:02AI volume and architecture
- 25:31Market winners in AI
- 26:53Copyright and responsible AI
- 28:58Practical AI applications
- 33:57AI future of AR/VR/3D
- 35:19AI limitations and questions
- 38:24Future of human creativity
- 39:05Q&A on prompts and copyright
related talks
27mGenerative AI: What's Next?
53mGenerative AI: Overview and use cases
45mHey, Look, I'm an Artist Now © ML Image Generator
32mHow generative AI is Changing the way people interact with computers
45mGenerative AI for Image and Video Production
22mGenerative AI & its Transformation of Business, Research, and Society
26mGenerative AI: Get a Grip … and a Look Under the Hood
40mPannel Discussion | From Hype To Value: Real World Applications Of Generative AI And LLMs