[ML News] Microsoft to spend 100 BILLION DOLLARS on supercomputer (& more industry news)

Microsoft plans to invest $100 billion in building a supercomputer for open AI models, raising discussions on achieving Artificial General Intelligence. Industry updates include the resignation of Stability AI’s CEO, improvements in Twitter’s Grock model, and Open AI’s advancements in synthetic voice technology and large language models.

Microsoft is planning to invest $100 billion in building a supercomputer to power open AI models, sparking discussions about the potential of achieving Artificial General Intelligence (AGI). The partnership between Microsoft and open AI limits commercialization rights to technologies not reaching AGI, creating uncertainty around the purpose of the large investment. Meanwhile, Stability AI’s CEO, Emad Mustak, has resigned, leading to speculation about the company’s future direction as it transitions to find a new CEO. Additionally, Twitter has announced the release of Grock 1.5, an improved model with enhanced reasoning capabilities and a context length of 128,000 tokens.

Open AI has introduced a voice engine for creating custom voices, showcasing experiments in synthetic voice technology. The blog post on navigating synthetic voices highlights the model’s ability to generate natural-sounding voices based on text input and a 15-second audio sample. However, the post is seen as an advertisement framed in the language of safety. Open AI’s Sora model has also been featured, showcasing its use in creating artistic content like movies, although some limitations in shot variety have been noted. Open AI is exploring partnerships to test usage-based earnings with GPT models, aiming to create a rewarding ecosystem for builders who contribute to the platform.

Open AI’s latest model, with 1.8 trillion parameters and requiring 30 billion quadrillion flops to train, was announced by Jensen W at GTC, emphasizing the scale of modern language models. Thomas Wolf has released a talk on building large language models in 2024, covering topics such as training data evaluation, parallelism, and effective training methods. The industry is witnessing advancements in AI technologies, with key players like Microsoft, Stability AI, Twitter, and Open AI making significant strides in the field. Despite uncertainties surrounding AGI and the commercialization of AI technologies, the rapid development of language models and synthetic voice technology continues to shape the future of AI research and applications.