Vintage Large Language Models
SMRTR summary
Researchers are developing "vintage" large language models trained exclusively on historical data up to specific cutoff dates, ranging from recent years like 2019 back to ancient periods like 200 AD. These time-constrained AI systems could enable backtesting of forecasting abilities against known events like the pandemic, simulate conversations with people from past eras, and test whether AI can "reinvent" major discoveries like quantum computing or Newton's laws using only knowledge available before those breakthroughs occurred.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article