The Challenge of Limited Context Length

2min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

The paper points out that there are actually different types of hallucinations. Extrinsic hallucinations, where one cannot verify the output correctness from the provided source. Limited Context Length is another well-known challenge for chat GPT models. Many open-source models, despite claiming to perform well with longer contexts, exhibits performance degradation.

NLW reviews a recent research paper which provides a comprehensive overview of the state of LLM development and organizes remaining challenges and problems to be solved -- along with prospective solutions. Before that on the Brief: OpenAI launches GPTBot to crawl the web for AI training purposes; Zoom has to back off ToS changes that would allow them to collected data for AI training purposes. Plus a terrifying new cyberattack that can tell what you typed just by listening to your keystrokes. Read the paper: https://arxiv.org/abs/2307.10169 Today's Sponsor: Giskard - the testing framework for ML models - https://www.giskard.ai/ ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI. Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/