In a bold move to push the boundaries of generative AI technology, Luma, the company behind a popular smartphone app, has announced its plans to leverage a compute cluster of ~3,000 Nvidia A100 GPUs for training new AI models. Alex Yu and Amit Jain, co-founders of Luma, aim to create AI models that possess the ability to “see, understand, show, explain, and eventually interact with the world.”
The first phase of their plan involves developing models capable of generating 3D objects from text descriptions. Luma has already launched one such model, named Genie, on its Discord server. The next step will be the creation of “next-gen” generative AI models to address the current limitations in generating realistic and coherent 3D objects, also known as the “uncanny valley” problem.
To support their ambitious vision, Luma recently secured $43 million in a Series B funding round, with participation from Andreessen Horowitz and other backers. This funding places the company’s valuation between $200 million and $300 million, significantly boosting its resources to bring their vision to life.
Although the field of 3D model-creating AI models is becoming increasingly competitive, Luma aims to stand out by prioritizing fidelity. Existing models in the market are trained on two-dimensional images, resulting in distorted and unusable outputs when generating scenes. Luma aims to overcome this limitation by incorporating advanced generative photorealistic technologies into its intuitive app.
While Luma’s improved version of Genie is launching today, the development of more capable generative AI models is still in progress. Nevertheless, the company is wasting no time, planning to double its workforce by the end of next year and assembling a model-running server cluster with “thousands” of GPUs.
Luma’s mission to build vastly more capable, intelligent, and useful visual models for its users is gaining momentum. With their growing team of experts in generative AI research, engineering, design, and product, Luma is well-positioned to revolutionize the field of generative AI technology.