The Story of OpenAI GPT3 - Improve Observability with Tracing
summary
OpenAI GPT-3 is a 175 billion parameter NLP API that handles slang and intent well and powers products like GitHub Copilot and Duolingo; experiments pair OpenAI (via Python) with Twilio to generate responses and a simple app that works with minimal code. Distributed tracing (Zipkin/Jaeger or New Relic SDKs) instruments the flow, 175 billion parameters and GitHub Copilot built on GPT-3 are highlighted, and tracing showed OpenAI layer causes most delay, so bottlenecks are remote API latency not local code.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:50The Reality of AI
- 2:47Are we there yet
- 4:32What is tracing
- 5:41Story of OpenAI
- 6:37Github Core Pilot
- 7:27What is OpenAI
- 8:45OpenAI GPT3
- 10:20How to get started
- 11:42OpenAI API
- 13:01Drawing Service
- 13:44Documentation
- 16:12Web Application
- 16:47Recap
- 17:18ApiDriven Architecture
- 18:57Example
- 21:00Conclusion
related talks
51mOpenAI and the AI that may overtake humans
24mHow OpenTelemetry Helps Generative AI
44mExploring ChatGPT for Improved Observability
38mExploring ChatGPT for Improved Observability
30mHow to train GPT-3 with your own dataset
23mExploring ChatGPT for Improved Observability
45mHow to use GPT in your projects: A Case Study
45mBringing the Power of AI to Your Application