Neural Networks

AI Generative AI

Why Does Diffusion Work Better than Auto-Regression?

This video is from Algorithmic Simplicity. Have you ever wondered how generative AI actually works? Well the short answer is, in exactly the same as way as regular AI! In this video I break down the state of the art in generative AI – Auto-regressors and Denoising Diffusion models – and explain how this seemingly […]

Read More
AI Neural Networks

Unveiling the Power of ONNX: The Keystone of Interoperable AI Models

The Open Neural Network Exchange (ONNX) format emerges as a pivotal innovation, fostering interoperability among AI models. As the AI landscape burgeons with diverse frameworks and tools, the challenge of model portability and efficiency in deployment becomes pronounced. ONNX, an open-source format, addresses this challenge head-on, enabling models trained in one framework to be exported […]

Read More
AI Deep Learning Neural Networks

Watching Neural Networks Learn

Emergent Garden provides this video about neural networks, function approximation, machine learning, and mathematical building blocks. Dennis Nedry did nothing wrong.

Read More
AI Data Driven Quantum Computing Virtual Reality

Rene Schulte on the Evolution of AI and Its Impact Across Industries

On the latest episode of ‘Data Driven,’ I had the pleasure of inviting Rene Schulte, a luminary in the fields of AI, AR, VR, spatial computing, and quantum computing. Rene shared fascinating personal insights, including his upbringing in former East Germany and his passion for working with innovation-driving communities at Reply. The future-forward chat explored […]

Read More
AI Google

Did Google fake their Gemini Video?

This video is from Yannic Kilcher noticed something off in the marketing video released alongside the Gemini model.

Read More
AI Neural Networks Research

Retentive Network: A Successor to Transformer for Large Language Models (Paper Explained)

This video is from Yannic Kilcher. Retention is an alternative to Attention in Transformers that can both be written in a parallel and in a recurrent fashion. This means the architecture achieves training parallelism while maintaining low-cost inference. Experiments in the paper look very promising. Paper:

Read More