It AI-n't What You Think!
summary
Claims that large models are accelerated inference not reasoning are backed by examples: asking for a Java stream to count occurrences produces usable code, asking for the most‑repeated character yields nonsense, while the model reliably spots a C++ evaluation‑order bug. Real‑world failures include a United chatbot disclaimer and an Air Canada refund example that ends in court. Conclusion: AI excels as a great investigator and idea generator but is alternative information—inconsistent, trained on humans’ bad code, and unsuitable as a sole source for production solutions.
AI-written summary · about
chapters (11)
- 0:00Introduction and welcome
- 3:33AI excitement and fear
- 6:20AI as magic and Clark's laws
- 13:12Lessons from history
- 20:47AI stands for alternative information
- 24:19AI as an investigator and code generator
- 31:35AI for ideas not solutions
- 38:22AI as a black swan
- 43:39AI as the future compiler
- 46:09Adapt or get swept away
- 49:53Q&A on ethics and skill levels
related talks
01h 00mAI: It AI-n't what you think!
56mAI: It AI-n't What You Think!
59mIt AI n't What You Think!
01h 01mKeynote: It AI-n't What You Think!
32mBeyond the Buzz: Navigating AIs Myths and Realities
52mArtificial, yes; Intelligent, no.
26mArtificial Intelligence? - More like Artificial Stupidity!
39mBack to Basics: Cutting Through the AI-Generated Code Noise