Unlearning means taking one thing back out, after it has already been mixed in. The video shows a baker asked to remove the honey from a finished cake.
A short explainer video, under a minute. The baker and the honey seller are an illustration of the idea.
At a glance
What Unlearning Is
The baker tries. But the honey is in every crumb, and she cannot pick it out.
How It Works
The video shows two ways to try. Bake the cake again without the honey, which is slow and costs a lot.
Why It Matters
The video turns this into one question: how would you know it is really gone?
Our report on a complaint filed by six publishers against Microsoft and OpenAI shows the idea in a legal filing. The report says the complaint’s prayer for relief asks for an injunction “requiring Defendants to remove all copies of Registered Works from all GPT or other LLM models and training sets.” It also says the complaint alleges that model parameters “encode retrievable copies of many of those training works.”
From our news coverage
The report presents all of this as the plaintiffs’ allegation and says the prayer does not say how removal from a trained model would be done. This is the lesson’s question in a filing: once something has been mixed in, taking it back out is hard, and so is knowing it is gone. This publication verified none of the allegations independently.
The Business Engineer Lens
Where this idea sits in the Business Engineer library:
The Business Engineer framework maps where each part of the AI stack sits and who holds the leverage. These pills are short lessons from the same library.
Unlearning builds on Parameters. Training mixes every source into the same trained numbers, which is why taking one source back out is so hard.
Data Quality contrasts with Unlearning. Filtering examples before training is something you can check; taking one back out afterwards means retraining or hoping.
Unlearning contrasts with Consent. Consent is asked before something is recorded; unlearning tries to take data back out after it was trained in.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
