Correct Answer:
B. Reinforcement Learning
Correct answer: Reinforcement Learning. ChatGPT is fine-tuned with reinforcement learning from human feedback, which rewards helpful and safe answers.
Background: After training on text, the model is improved by human trainers ranking its answers, so it learns which replies people prefer.
- Machine working: Machine working is not a type of learning.
- Linear Programing: Linear programming is a maths method, not how ChatGPT learns.
- Supervision Learning: Supervised learning is used in training, but the fine-tuning that shapes ChatGPT is reinforcement learning.
Did you know? GPT stands for Generative Pre-trained Transformer, and ChatGPT was released in November 2022.
Memory trick: ChatGPT learns from human feedback - reinforcement from people.
Leave a Reply
You must be logged in to post a comment.