Submissions by trott

Will We Run Out of Data? Limits of LLM Scaling Based on Human-Generated Data

AI Agents: too early, too expensive, too unreliable

AI's learning ineffectiveness and how to fix it

Are You Smarter Than An LLM? (Quiz based on the most popular LLM benchmark)

Training LLMs using 8-bit numbers

Using scaling laws to predict when a scaled-up GPT-4 becomes superhuman

Learning Logic Programs by Explaining Failures