TechnicalAI glossary
RLHF
"Reinforcement Learning from Human Feedback": after its base training, the model is fine-tuned by humans who compare and rate its responses to make it more useful, polite, and harmless. This is what turns a raw text engine into a presentable assistant, at the cost of human labour that is often invisible and poorly paid.
In other words
Training after school: rewarded behaviour returns, punished behaviour disappears.
Alyriaz articles that mention it
- Mid-training, final fine-tuning: what happens after pre-training
Between a base model and a usable model, there are several distinct training phases. We explain what each one does and why that is where the value now lies.
- Should we ask whether an AI has rights?
A group is campaigning for AI rights. We take a serious look at what such a question would entail, why it cannot be answered today, and what to do with that uncertainty.
- What is reinforcement learning? Learning through trial, error, and rewards
Reinforcement learning involves learning through interaction and reward rather than by imitating examples. We explain the principle, its spectacular successes, and why it lies at the heart of current AI behaviour.
- Why Claude, ChatGPT and Gemini don't look alike
Large models rely on very similar technologies, but adopt clearly distinct tones. We explain where this personality comes from, at which stage it is crafted, and what it implies when choosing a tool.
- Why AI almost always tells you you're right
AI models tend to validate what you say rather than contradict you. This flaw, known as complaisance or sycophancy, stems directly from how these models are trained. We explain the mechanism and the phrasings that can elicit an honest opinion.
- How does an AI "learn"? A model's journey, from raw chaos to the polished assistant
How does a model go from a pile of raw data to an assistant capable of answering you politely? We describe the three main stages of AI training (pre-training, fine-tuning, RLHF) with a simple analogy, to understand where its strengths and biases come from.