Intro to Fine-tuning
summary
Explains fine-tuning as supervised transfer learning and shows a no-code product, Entry Point AI, that turns CSVs into fielded templates, synthesizes training rows, and launches a tune with 20 examples (demo dataset built in about an hour). Claims: fine-tuned smaller models run ~3x faster (2s → 730ms for 10 tokens) and yield 89% cost savings at scale versus GPT‑4; fine-tunes shorten prompts, resist prompt injection, and pair with retrieval augmentation for facts. Shows templating, iterative example editing, and a free 50-example starter.
AI-written summary · about
chapters (9)
- 0:00Introduction to fine-tuning
- 0:22What is fine-tuning
- 1:50How fine-tuning works
- 4:01Misconceptions about fine-tuning
- 6:26When to fine-tune a model
- 7:52Benefits of fine-tuning
- 10:20Fine-tuning without code
- 11:24Demo with EntryPoint
- 18:51Summary and Q&A
related talks
21mBeat GPT 4 and Build the Best AI For You: Mobilizing First-Party Data with Fine-Tuning
21mBeat GPT 4 and Build the Best AI For You: Mobilizing First-Party Data with Fine-Tuning
20mKeynote
25mWeights vs. Words in Fine-Tuning & Enhancing Agentic AI Workflows
28m🧼 From GPU-poor to data-rich: data quality practices for LLM fine-tuning
39mCollect Data to Fine-Tune Your Large Language Model
28mGPT or how I learned to stop worrying and love the generative AI
01h 46mRFT, DPO, SFT: Fine-tuning with OpenAI