4 ms·

In 2020, GPT-3 175B was already outperforming humans on SAT questions by ~20%. There are far more formal benchmarks for models now. https://lifearchitect.ai/iq

In 2020, GPT-3 175B was already outperforming humans on SAT questions by ~20%. There are far more formal benchmarks for models now. https://lifearchitect.ai/iq-testing-ai/ https://lifearchitect.ai/iq-testing-ai/

bagels 4y ago
Surely outperforming some humans? I don't see how it can outperform perfect scores by 20%?
- cypress66 4y ago
  Obviously performing better than the average SAT (human) result.