DISTILLATION

 

Distillation was familiar to me when I had been a student of Chemistry.

But an editorial in the Times of india of 25th February 2026 teaches me 'Distillation' is an important term in the IT segment apart from Chemistry.

Excerpts from the editorial;

 Anthropic doesn't want rivals cloning Claude for free, but that is what three Chinese firms, including last year's newsmaker DeepSeek, allegedly tried doing.  Anthropic says they used Claude as a teacher for their "student" models. Hardly two weeks earlier, Anthropic's bigger rival OpenAI had accused DeepSeek of extracting its model.  And last year, DeepSeek's market shaking debut was clouded by similar allegations.

Distillation is a decade old idea that was rejected when first presented at a conference. The way it works is that a rookie AI poses millions of questions to a leading AI model like ChatGPT. It seeks not only final answers but also steps used to arrive at them.  This reveals the larger model's "thinking,"  which the new model copies to deliver pretty good answers most of the time, using a fraction of hardware and energy.

Distillation isn't always a bad thing. AI firms distil their own models for speed and efficiency regularly. But it is unfair when rivals use distillation to catch up. It can also be dangerous. Recall that Claude was reportedly used in America's Venezuela operation to extract Maduro.  If Chinese firms figure out Claude's reasoning, but strip it of all safeguards, the resultant AI could be used to cause havoc. AI firms will have to improve their capabilities to detect and thwart unauthorised distillation attempts. As AI aspirations and capabilities grow, its models might also face such attacks. It is evident AI defences should be built starting now.