Code Vulnerabilities Explored

Ajay discusses the vulnerabilities inherent in neural code and language models, emphasizing how slight changes in input can dramatically alter their behavior. He highlights the importance of data diversity and architectural adjustments to enhance model robustness. Additionally, he mentions the potential for using data augmentation techniques, such as recompiling code, to improve performance, particularly for fine-tuning models with smaller datasets.

In this clip
From this podcast
Super Data Science: ML & AI Podcast with Jon Krohn
711: Image, Video and 3D-Model Generation from Natural Language — with Dr. Ajay Jain
Related Questions
- What are adversarial attacks on machine learning models?
- Are there people using this commercially now, or are there open source implementations that are easy for people to use if they want to be doing that kind of fine tuning on their own enterprise data with contra code, can they do that today?

Code Vulnerabilities Explored

In this clip

From this podcast

Super Data Science: ML & AI Podcast with Jon Krohn

711: Image, Video and 3D-Model Generation from Natural Language — with Dr. Ajay Jain

Related Questions

What are adversarial attacks on machine learning models?

Are there people using this commercially now, or are there open source implementations that are easy for people to use if they want to be doing that kind of fine tuning on their own enterprise data with contra code, can they do that today?