AIAcademy update cohort · archived

2020 Q4 few-shot foundation roots

The archived 2020 Q4 baseline preserves the pre-assistant foundation-model roots: transformer architecture as the reusable substrate, GPT-3 few-shot prompting as the first widely visible general model pattern, and MMLU-style benchmark literacy before instruction tuning, chat products, and system cards existed.

3 updates · 3 source-linked · 0 product-created

Key questions

• Which 2020 research claims became durable foundations for later assistant lessons?
• Which few-shot and benchmark claims should stay historical rather than current product advice?
• Which transformer and scaling ideas explain later AI progress without replacing current source-review habits?

Updates

• GPT-3 few-shot baseline archived
  GPT-3 remains visible as the 2020 baseline for in-context learning, few-shot prompting, and the shift from task-specific systems toward general-purpose language-model behavior before assistant products arrived.
• Early benchmark literacy baseline preserved
  MMLU stays archived as the 2020 broad-knowledge benchmark reference so later system-card and evaluation lessons can compare benchmark claims with real-world competence limits.
• Transformer roots linked to few-shot scale
  The 2020 archive connects the transformer architecture root to GPT-3-scale behavior so learners can see what was already durable before retrieval, instruction tuning, and consumer chat interfaces changed the curriculum.