The High Cost of AI Validation: Why Skepticism Boils Down to Slop Filtering
kareem_carr · x · 2026-08-01
Data scientist Kareem Carr expressed pragmatic skepticism regarding the reliability of current AI systems for complex tasks. He argues that the real-world utility of AI comes down to three simple questions:
- Low Hit Rate: Will I need to test thousands of problems just to find the 0.1% it does well?
- Resource Burn: Will I burn through all my credits immediately?
- Validation Overhead: Does verifying answers mean wading through thousands of slop outputs to find the right one?
He concludes that his AI skepticism fundamentally stems from these hidden, exorbitant costs of usage and verification.
More from AGI Musings
- AX and UAX: How AI Agents Are Redefining Product Experience — EdenEmarco177 · 2026-08-02
- Math and Cyber Are Existence Proofs for Superhuman Intelligence — nabeelqu · 2026-08-02
- Gary Marcus: Pure LLMs Are Stochastic Parrots; Astra Proves Need for Neurosymbolic AI — GaryMarcus · 2026-08-02
- The AI Quest: Today's Amazing Capabilities Are Just a Minor Sub-quest — yunta_tsai · 2026-08-02
- Opinion: AI Skill Distribution is Becoming Spiky; Verification Loops Dictate Progress — nbaschez · 2026-08-02
- AI Prompt Automates Unclaimed Property Search Across 48 States — ScamSchoolBrian · 2026-08-02