/images/portrate.jpg

Large Language Models and Expert Assistants

In the ever-evolving landscape of Artificial Intelligence, a groundbreaking category of technologies known as foundation models is reshaping our interaction with the digital world. These models, which encompass the broad and powerful class of Large Language Models (LLMs), are built on expansive datasets, enabling them to perform an astonishing array of tasks. From engaging in human-like dialogues and answering complex questions to composing music and generating video content, LLMs are rapidly becoming specialized powerhouses in AI. Their versatility allows for remarkable applications that stretch the bounds of creativity and problem-solving. Through this series of posts, we will embark on a journey to uncover the myriad applications of LLMs across different domains, showcasing their transformative potential in reshaping industries and enhancing human capabilities.

Can we imitate Nature's evolutionary abilities?

In the rich theatre of Nature, few creatures exemplify the power of evolutionary optimization as strikingly as the hummingbird. Similarly, in the field of artificial intelligence, Genetic Algorithms stand among the select few that have successfully harnessed the principles of evolution for optimization. In this post, I will discuss these techniques that solve complex problems through an evolutionary process, leading to optimal or near-optimal solutions.

Generative Models: Variational Autoencoders

Generative models are a class of statistical models that aim to learn the underlying data distribution from a given dataset. These models provide a way to generate new samples that are statistically similar to the training data. They have gained substantial attention in various domains, such as image generation, speech synthesis, and even drug discovery.

Retrieval-augmented generation for Chatbot

I developed a chatbot-like agent that utilizes Retrieval-Augmented Generation, integrating a pre-trained sequence-to-sequence model with a document database. This innovative synergy facilitates the creation of a question-answering application, which is capable of delivering contextually relevant and accurate responses by querying specific information.

Responsible AI Tools

Responsible AI has become a crucial aspect of every Machine Learning project, underscoring the need for tools that promote the creation of fair and ethically sound models. In this post, we delve into some key concepts and replicate IBM’s inFairness example solution to address bias in the ‘Adult Income’ dataset. Beyond the technical configurations adopted, the intent to algorithmically address unfairness and the statistical pursuit of both overt and hidden bias in data are particularly noteworthy. These are the key insights we hope readers will take away from this post.