AI: Actual Intelligence
summary
Gives concrete examples: an fMRI study that flags a dead salmon due to multiple statistical tests, chatbots that learn abusive behavior and get shut down, convolutional neural nets that colourize black‑and‑white photos and beat earlier feature‑engineering approaches, and biometric systems fooled by recorded voice or poor fingerprint scans. Notes predictive‑sentencing models and automation bias where clinicians defer to algorithmic reads of medical scans. Emphasizes validation, dataset bias, and that some prototypes work while many fail or mislead.
AI-written summary · about
chapters (31)
- 0:00Intro
- 0:45Overview
- 1:49Human, uniquely human
- 4:33Do the maths (properly)
- 5:27Frankenstein's monster
- 7:24Vitruvian man
- 8:05Masahiro Mori's Uncanny valley
- 9:00Human, all too human
- 11:36Penrose
- 12:52Prove it's you
- 16:07Fingerprints
- 18:04Business Intelligence
- 19:55Anonymised?!
- 20:51Snoopers
- 22:36Data science?!
- 24:27Really big data
- 25:41Smart != Intelligence
- 29:06Define intelligence
- 30:38How do you spot intelligence?
- 33:11Formalised thinking
- 35:04Credo, ergo sum
- 36:06Drawing conclusions
- 37:09Goodness...
- 37:34Algorithms
- 40:06Automate everything?
- 43:12BIBO: bias in, bias out
- 43:54I have created a monster
- 45:34Automation bias
- 47:09Deep...
- 48:45Make the machines play the game
- 51:52Take-aways
related talks
19mAI: The current state of play
56mArtificial Intelligence? - more like Artificial Stupidity!
38mAI in action: How to use practically in iGaming beyond the hype?
01h 03mAI is Your New Design Material
39mAI is everywhere!
23mAI in the Life Science Sector
03mInterview
20mDecision Intelligence: The Next Phase in AI