The Pickle Problem - A Security Nightmare in ML
Learn how malicious code can be embedded in model weights and how it can sabotage training processes.
Recent posts
In-Context Vectors - Controlling Language Models Through Latent Space Steering
In-Context Vectors represent a promising approach to controlling language model behavior through direct manipulation of hidden states. Talk about making In C...
Why zero-ing gradients isn’t enough?
Talks about why setting set_to_none = True makes a difference
Creating an Automated UI controller with GPT agents & GPT-4v
Talks about how we can build an automated UI controller with GPT agents
Rapid Prototyping with LLMs - The Journey from Notebook Musings to Real-World Enterprise Solutions
Talks about how LLMs enabled me to rapidly prototype and test ideas and how they are enabling a new era of Product Integration.