Author: ritikaverma
-
AI Governance and Explainable AI – Addressing Algorithmic Bias
Artificial intelligence (AI) is rapidly transforming industries and reshaping decision-making processes across sectors. However, as AI systems become more embedded in critical areas such as finance, healthcare, and marketing, concerns about algorithmic bias became impossible to ignore. While the conversation on bias isn’t new, its relevance remains pressing. My recent read from Harvard Business School…
-
Securing AI – Addressing Web-Based Attacks on Large Language Models
Securing AI involves two key aspects: first, protecting the models themselves—ensuring that the data they are trained on is safe and resilient against manipulation—and second, safeguarding the underlying application layer, including the APIs that LLMs interact with. In this post, we’ll explore some common web-based attacks on LLMs and introduce how frameworks like NIST Dioptra…
-
Ensuring Precision in RAG Systems: Evaluation
In my journey of creating multiple Retrieval-Augmented Generation (RAG) systems, I’ve encountered the common challenge where the RAG responds with “I do not have the context” or provides partial data from the document. This experience has underscored a crucial insight: while creating a RAG is no longer a significant challenge, developing a high-performing RAG—one that…
-
One-Pixel Attack: A Subtle Yet Potent Adversarial Technique
Generated by DALL-E Introduction How easy is it to cause a deep neural network to misclassify an image by modifying just one pixel? Surprisingly, it’s quite simple. An attacker can manipulate the network to return any desired answer. Recently, I discovered the concept of the One-Pixel Attack on Deep Neural Networks. It’s fascinating how altering…