The Unreliable Computer Revolution - Embracing the Double-Edged Sword of GenAI
summary
Generative models like ChatGPT and Gemini are growing at roughly 10x annual growth and intentionally hallucinate, which is the creative tradeoff that makes them useful and unreliable. Practical wins include an augmented reality Tetris built in 54 minutes and a publishable industry article in 40 minutes, but a public Google prototype produced dangerous false advice. Use structured prompting, retrieval‑augmented generation and knowledge graphs (semantic DBs beat SQL for complex answers, ~38.7% vs 0), the Rocks method and a reuse recommendation engine to reduce hallucinations.
AI-written summary · about
chapters (22)
- 0:00Introduction and speaker background
- 2:39The urgency of using GenAI carefully
- 5:28Exponential growth of AI capabilities
- 7:13GenAI as culture compressed data
- 9:43AI applications beyond chatbots
- 12:13From calculators to creative computers
- 13:38AI's rapid and unpredictable evolution
- 15:01AI hallucinations as creative necessity
- 17:06AI limitations and human-like flaws
- 18:10Ethical risks and AI bias
- 20:16Potential for mass harm misuse
- 21:23Dove Real Beauty campaign and AI diversity
- 22:27Fifth-generation programming languages
- 23:10Productivity gains with AI tools
- 24:34Generative motion and synthetic training
- 25:17Teaching AI your content model
- 27:27Embracing AI as a double-edged sword
- 28:11Structured AI workflows and skills
- 32:02Optimizing AI success with taxonomy
- 33:46Training AI with high-quality data
- 34:52GenAI as part of a digital intelligence system
- 35:57Q&A on AGI, singularity, and AI dangers
related talks
46mConversations with ChatGPT: illusion or reality?
27mChallenges of Generative AI
17mUnlock the power of Generative AI: From understanding its principles to utilizing its power
37mGenerative AI: Beyond (Just) Hype
32mHow generative AI is Changing the way people interact with computers
55mAI Showdown: The Good, The Glitchy, and The Risky
15mHarnessing the Power of Generative AI for Intelligent Applications
06mInterview with Spela Poklukar