
Program Aided Language Models
Data Skeptic
Generating Code with Limited Success
The speakers discuss their experiences with generating code using a LLM, noting that while they have had success with shorter snippets and improving existing code, generating repository level code is still a challenge. They explore the potential for incorporating longer context support decoder models and pre-training to enhance performance, and share specific examples of successfully generating code for tasks they were not familiar with.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.