Leveraging AI for Video Generation: A Journey from Simulating Reality to Market Disruption
Welcome to Luma AI, where I'm Amed, Co-Founder and CEO, and we're revolutionizing multimodal general intelligence by building world-class video models.
- 1. The worst thing that can happen to someone who is creating something is apathy from the audience.
- 2. The second worst scenario is when everyone is happy with what you created because there is nothing left to do.
- 3. Amed is one of the co-founders and CEO of Luma AI, a company building multimodal general intelligence starting with the world's best video models.
- 4. Luma AI has raised $200 million from investors including Andreessen Horowitz, Amplify Partners, Matrix Partners, Nvidia, AMD, and Amazon.
- 5. Amed grew up in India with a strong interest in physics, eventually studying math and physics in college.
- 6. He started building iOS apps alongside his physics studies, with his first app being a differential equation solver.
- 7. After college, Amed joined friends at their startup instead of pursuing graduate school for physics.
- 8. Later, while working at Apple on the Shortcuts/Workflows project, he learned about the Vision Pro and joined the team working on 3D capturing of the world with it.
- 9. In 2020, Amed became interested in large language models and Nerf neural radiance fields (Nerf) for 3D instruction.
- 10. He experimented with various methods to simulate the dynamic world based on Nerf and Dali's image generation capabilities.
- 11. In 2022, Amed left Apple and founded Luma AI with a focus on building a 3D generative model called Genie.
- 12. The company also developed large-scale infrastructure for training models and data collection systems.
- 13. In 2024, Luma AI released its first video model called Dream Machine.
- 14. This release was preceded by Jaing joining the team as Chief Scientist, with his expertise in image and video generation from Nvidia.
- 15. The new graphics cards from Nvidia were capable of handling large-scale data and algorithms for 3D generative models.
- 16. Dream Machine took four and a half months to train and was released after a year and a half of work on encoders, learning models, and infrastructure.
- 17. Open Sora's announcement in February 2024 allowed Luma AI to scale its video efforts significantly.
- 18. The first Dream Machine model received much attention from media outlets like Good Morning America and CNN.
- 19. Iterating quickly is important when developing new technologies, even if it means sacrificing some stability or scalability in favor of faster development and testing.
- 20. Amed's experience with the Vision Pro at Apple inspired him to create Luma AI and focus on multimodal general intelligence.
- 21. Multimodal general intelligence involves models that let people create worlds by combining different data types, such as text, video, and audio.
- 22. Luma AI focuses on solving multimodal general intelligence rather than just building image or video generation models.
- 23. Amed suggests trying various things and going deep into them to discover if the depth of a problem increases your excitement and motivation.
- 24. The key to finding a worthwhile problem is whether it genuinely interests you, not just in the moment but over time.
Source: EO via YouTube
❓ What do you think? What is the most innovative or impactful way you have seen apathy manifest, and how did you respond to it in your own entrepreneurial journey? Feel free to share your thoughts in the comments!