Conversations with ChatGPT: illusion or reality?
summary
Explains how ChatGPT and similar systems (GPT-3.5, GPT-4) are transformer large language models trained on web data with human feedback, noting the rapid user adoption claim. Demonstrates a live-style example where the model can speed up prototyping and provide StackOverflow-like coding help but also hallucinates (fabricates a repository and outputs), makes arithmetic and reasoning errors, produces biased job text and potential malware, and has high compute and opaque data-use costs; frames the development as an evolutionary advance rather than a solved reasoning system and mentions emerging techniques like auto-prompt chaining.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:25ChatGPT popularity
- 5:40What is ChatGPT
- 10:52Buzz origin and access
- 11:56Live demo with ChatGPT
- 17:36ChatGPT limits and errors
- 21:46How ChatGPT works
- 24:55ChatGPT risks and issues
- 30:59Bias and harmful content
- 32:26AI ethics and responsibility
- 33:55Environmental impact and data
- 34:37Impact on developer jobs
- 37:48Summary and future
- 40:19Q&A
related talks
41mThe Unreliable Computer Revolution - Embracing the Double-Edged Sword of GenAI
01h 01mHow ChatGPT Works
02h 00mChatGPT & Prompt Engineering
01h 01mKeynote: Transformers: The Rise of ChatGPT
35mChatGPT is lying, how can we fix it?
50mAgency
32mHow generative AI is Changing the way people interact with computers
51mOpenAI and the AI that may overtake humans