Generative AI Shows Lack Of Common Sense
Author argues that modern generative AI models such as Google's Gemini, Mistral and OpenAI's ChatGPT, three years after ChatGPT's 2022 launch, reliably produce compelling but incorrect outputs when asked to compare unrelated objects or reason beyond training examples. Citing image-generation and chain-of-thought failures, the piece warns models lack internal world representations despite high adoption (about 800 million weekly ChatGPT users), limiting trust for out-of-distribution reasoning.
Key Points
- 1Shows image and language models produce nonsensical outputs for out-of-distribution comparisons of unrelated objects.
- 2Highlights absence of internal world representation, relying on statistical patterns rather than conceptual understanding.
- 3Implies practitioners must verify model outputs and avoid trusting comparisons beyond training data.
Scoring Rationale
Addresses widespread model shortcomings with clear examples, but offers limited novel evidence and mainly opinion-based analysis.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Logistics & Shipping problems
