Autonomous Agents — #AI

In deep learning, training dynamics often mislead, and loss behaviors can appear counterintuitive. A decreasing training loss does not…

Here is a burning question. “Why can’t the AI models (LLMs) adhere to minimum word counts yet?”

Every time I talk about causal inference in genomics, people ask, ‘But how?’ How do we move from observing correlations in massive genomic…

Unseen Vulnerabilities

Balancing Stochasticity and Determinism

In the previous parts, I discussed why large language models alone fall short of achieving AGI and why I believe a paradigm shift is…

Language is an emergent property of Mathematics. Not the other way around.

I believe that within the next seven years, the fundamental nature of how autonomous agents communicate will change. The current reliance…

In Part-1 of Why LLMs Will Never Lead to AGI, I argued that LLMs are static in nature and lack real-time learning, lack internalized…

The Mathematical and Biological Barriers

Reinforcement Learning in Mathematical Proofs: Aligning Verifiability with NP-Hard Problem Solving

Chain-of-thought (CoT) reasoning has significantly advanced the capabilities of large language models (LLMs) by enabling them to handle…

OpenAI’s recent release of the OpenAI o1 model, also known as the Strawberry AI model, represents a significant leap forward in the field…

A Mathematical Exploration of Rapid Adaptation

When it comes to building robust deep neural networks (DNNs), the importance of loss function design cannot be overstated. The choice of a…

I was chatting up with a few advisee companies last month, who wanted to know what goes into foundational AI models. The conversation…

In a recent conversation with postdoctoral math grads, we discussed the capabilities and limitations of Large Language Models (LLMs) in…

In this paper I want to present a mathematical comparison between the Mamba (Selective Structured State Space Model) and my research on…

In recent years, there has been significant interest in leveraging machine learning techniques to solve complex partial differential…

In the quest to build Artificial General Intelligence (AGI) models, one of the most pressing challenges is to endow machines with the…

A General Summary Without Any Math.

The Imperatives of Discretization and Resolution Invariance