Why It’s Cheap and Easy to Mimic OpenAI’s Reasoning Model
The word of the moment in artificial intelligence is distillation.
Distillation is the technical term for transferring knowledge from one large language model, which acts like a teacher, to another model that acts as a student. And it’s likely one of the key ways a Chinese quant fund replicated OpenAI’s o1 reasoning model and set the industry on fire with DeepSeek, OpenAI said publicly Wednesday.
Distillation involves the student model learning from the answers the teacher gives. In the case of o1 as a teacher model, it means learning from the model’s “thought process”—a block of text, also called a chain of thought, which OpenAI’s model spits out to give people an understanding of the logical steps it took to arrive at an answer.
Related Knowledge Center: Artificial intelligence
On top of a failing business model, newsrooms must navigate the powerful — and not fully reliable — new capabilities of artificial intelligence and the deployment of large language models.
State of the business model | Copyright and intellectual property