AIAcademy update cohort · archived

2021 Q4 research-frontier bridge

The archived 2021 Q4 bridge preserves the research frontier just before the 2022 assistant stack: CLIP and DALL-E text-image learning, Codex and HumanEval code generation, WebGPT browsing with citations and human feedback, and DeepMind Gopher/RETRO scale-plus-retrieval evidence.

5 updates · 5 source-linked · 0 product-created

Key questions

• Which 2021 research ideas became everyday multimodal, coding, and retrieval lessons later?
• Which claims should stay historical because they predate instruction-tuned chat assistants and product guardrails?
• Which source-grounded answering and evaluation habits were visible before current AI products made them routine?

Updates

• Text-image research baseline archived
  CLIP and DALL-E stay visible as the 2021 baseline for text-image representation and zero-shot generation before generated images became everyday creator and family media.
• Codex and HumanEval baseline archived
  The first Codex evaluation evidence remains visible so coding-agent lessons can compare unit-function correctness with later repository and autonomous-agent benchmarks.
• Source-grounded answering baseline preserved
  WebGPT stays archived as the older browsing, citation, and human-feedback pattern before current assistants made source display and retrieval quality an everyday learner expectation.
• Scale and retrieval alternatives archived
  Gopher and RETRO preserve the 2021 contrast between scaling transformer language models and adding explicit retrieval before Chinchilla shifted compute-optimal training guidance.
• Research-to-product gap made explicit
  The 2021 cohort flags that multimodal, coding, browsing, and retrieval research existed before the product, policy, classroom, and household practices learners now need.