Machine Learning can't do the thinking
summary
The core claim is that no AGI yet and urgent attention belongs to data and ethics not speculation about sentient machines The talk names concrete systems and outcomes showing why: a billboard with camera facial detection that outputs nonsense counts and provokes outrage, the Amazon hiring model that is quietly discarded, a hospital algorithm using cost as a proxy that reproduces racial disparities, and the COMPAS risk score trade offs between individual fairness and group parity Deep neural nets and ML pipelines predict associations not causation and those deployments break or harm in specific, fixable ways
AI-written summary · about
chapters (8)
- 0:00AI versus early trains
- 3:22Role of data in ML
- 10:34AI misuse and ethics
- 15:58AI ethics and ownership
- 21:27Fairness in AI trade-offs
- 26:25Causality vs correlation in AI
- 31:06Human responsibility in AI
- 31:28Q&A
related talks
49mAn AI with an Agenda
56mArtificial Intelligence? - more like Artificial Stupidity!
29mThe Political Roots of AI Policy
30mHow to Lie with AI: Understanding Bias, Ethics, and the Hidden Risks in ML
16mMachine Intelligence With Humans in Mind
26mReflective, not Objective: Intentionality in AI
52mSome pitfalls in AI
46mAI and ML, separating hype from what actually matters