Hugging Face Launches ML Intern, a Budget-Capped ML Agent

A new assistant from Hugging Face autonomously orchestrates machine learning experiments while adhering to a predefined budget. A demonstration mentions six hours of computation for less than $0.50. This release comes as Nvidia acquires the company, promising to keep the platform open and neutral.

Announced Costs and Tracking: Six Hours for Less than $0.50

Hugging Face reports that an experiment showcased in a video required approximately six hours of computation at a cost of under $0.50. Each training session features a dashboard that allows users to track the progress of the work.

From Brief to Execution: Conversation, Research, and Guaranteed Budget

The user begins by describing their idea in a conversation with the assistant. It then searches the Hugging Face Hub, GitHub, and the web for suitable models, datasets, and tools. Before launching the experiment, ML Intern estimates the computational costs and proposes a budget not to exceed. Once this budget is approved, the system operates autonomously.

Launch, Functional Scope, and Nvidia Context

ML Intern, integrated into the Hugging Face chatbot, enables machine learning experiments without prior expertise. The assistant can create datasets, train models, monitor work, upload results to the Hub, write reports, and build demonstrations. This launch occurs as Hugging Face is being acquired by Nvidia, whose CEO Jensen Huang has stated a desire to keep the platform open and neutral regarding hardware.