Inflection AI secures $1.3B funding with Microsoft and Nvidia as lead investors.

On June 29, Inflection AI, based in Palo Alto, announced that it has raised $1.3 billion in funding. The funding round was led by Microsoft, Reid Hoffman, Bill Gates, Eric Schmidt, and Nvidia. A significant portion of the capital will be used to construct a 22,000-unit Nvidia H100 Tensor GPU cluster, which Inflection AI claims to be the largest in the world. These GPUs will be utilized for the development of large-scale artificial intelligence models. The company’s developers stated:

“We estimate that if we were to enter our cluster in the recent TOP500 list of supercomputers, it would be ranked 2nd and near the top, despite being optimized for AI applications rather than scientific ones.”

Inflection AI is also working on its own personal assistant AI system called “Pi.” The company describes Pi as “a teacher, coach, confidante, creative partner, and sounding board” that can be accessed directly through social media or WhatsApp. Since its establishment in early 2022, the company has received a total funding amount of $1.525 billion.

Despite the increasing investment in large AI models, experts caution that their training efficiency can be severely limited by current technological constraints. As an example, Singaporean venture fund Foresight raised the issue of a 175 billion parameter AI model that stores 700GB of data:

“Assuming we have 100 computing nodes, and each node needs to update all parameters at each step, each step would require transmitting about 70TB of data (700GB*100). If we optimistically assume that each step takes 1 second, then 70TB of data would need to be transmitted per second. This demand for bandwidth far exceeds the capacity of most networks.”

In addition to the above example, Foresight also warned that “due to communication latency and network congestion, data transmission time might far exceed 1 second,” which means that computing nodes could spend most of their time waiting for data transmission instead of performing actual computations. Foresight analysts concluded that, given the current limitations, the solution lies in small AI models that are “easier to deploy and manage.”

“In many application scenarios, users or companies do not need the more universal reasoning capability of large language models but are only focused on a very specific prediction target.”

Magazine: AI Eye: AI travel booking hilariously bad, 3 weird uses for ChatGPT, crypto plugins