GPU Shortage Solved: Scientists discover how to train LLMs on ambient anxiety

In a frantic bid to bypass the global silicon shortage, a team of rogue researchers has successfully trained a 70B parameter model purely on the ambient anxiety of junior developers.

The breakthrough occurred when Dr. Sarah Jenkins noticed that her intern’s nervous pacing generated a consistent electromagnetic field. ‘We realized that the sheer existential dread of a bootcamp grad trying to center a div is a form of pure, unadulterated energy,’ Jenkins explained.

By wiring up a co-working space in San Francisco, the team managed to simulate the compute power of 4,000 Nvidia H100s. ‘It’s sustainable, renewable, and as long as JavaScript frameworks keep changing every week, we have infinite fuel,’ said the project lead. Critics argue that this method is unethical, but stakeholders are thrilled about the reduced cloud bills.