ZML releases free product to speed inference across AI chips (techcrunch.com)

ZML, a French AI startup backed by Turing Award winner Yann LeCun, has launched ZML/LLMD, an inference-performance software designed to enhance the execution speed of open-source large language models across various AI chips, including those from Nvidia, AMD, Google, and Intel. The tool aims to dismantle existing technological silos and alleviate vendor lock-in, thereby enabling organizations to optimize AI inference at unprecedented speeds. Founder Steeve Morin emphasizes the significance of this move in a landscape where optimizing inference — the processing of prompts — is becoming increasingly crucial as AI applications permeate everyday life. By allowing enterprises to mix and match chips, potentially reducing costs and energy consumption, ZML/LLMD could disrupt the AI market amidst growing concerns over expensive AI solutions. The startup has ambitious plans to collaborate with various chip manufacturers, particularly from Europe, marking a significant shift in the industry. Although ZML/LLMD is being offered as a free product initially, the startup is focused on gathering usage data before transitioning to a monetization strategy. The product's reception and potential market impact are being closely watched by key figures in the AI sector, solidifying ZML’s position as a significant player in the burgeoning field of AI inference.