GPT-4 Finished Training 4 Years Ago, Underwent 7 Months of Safety Evals
koltregaskes · x · 2026-08-09
OpenAI executive Greg Brockman revealed that GPT-4 finished training four years ago (around 2019). A commenter pointed out that this means GPT-4 was already developed before ChatGPT (initially powered by GPT-3.5) was even launched.
The analysis highlights that GPT-4 underwent 7 months of safety evaluations, reinforcement learning, red teaming, and alignment before its public release (not counting the early Sydney release by Microsoft). Given this historical context, the industry's current impatience over having to wait an extra 30 days for safety checks seems rather unjustified.
Related event: GPT-4 Training Anniversary Sparks Retrospect and Scaling Debate(8 posts)→
More from Companies & People
- DHH: Starship, Starlink and Grok should have been Europe's Sputnik moments — georgemillo · 2026-10-16
- Tinyfool sparks debate over whether Tesla Roadster 2 counts as 'useless' innovation — tinyfool · 2026-10-16
- Tinyfool: does Xiaomi's foldable released days before iPhone count as innovation? — tinyfool · 2026-10-16
- US universities go on an AI philosopher hiring spree as philosophy departments add tenure-track AI roles — AnnaCiaunica · 2026-10-16
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- Ex-Amazon Engineer: AI Has Killed Bezos's Six-Page Memo Culture — bibryam · 2026-10-16