LLMs, mixture-of-depths, mixture-of-experts, MoD, MoE
Author: AJ Maren
Building Your Online Portfolio (A Collection of Useful Links)
One of the strongest things that we can do to position ourselves – for the next career move, and also for creating a new tier of powerful professional relations – is to build our online Portfolio. This post provides links to good Portfolio examples for three different cases: This post provides a collection of useful… Continue reading Building Your Online Portfolio (A Collection of Useful Links)
Generative AI: A Teenager Acting Out
Four years ago, tech journalist and old friend Lee Goldberg asked me to semi-/sort-of collaborate with him on his yearly April Fool’s (April 1st) day article. Basically, Lee had a really fun idea for writing about “artificial stupidity,” instead of “artificial intelligence,” and he asked me to give him some real-AI credibility to use in… Continue reading Generative AI: A Teenager Acting Out
It Might All Come Down to Rare Earths
Jensen Huang’s keynote talk at NVIDIA GTC last week was very likely the tip of the iceberg. Demand for processing units is going up. Going CRAZY up. NVIDIA’s new product releases and recent stock price upsurges reflect that. But NVIDIA is not the only chip-maker in the US. The Biden Administration has been investing …… Continue reading It Might All Come Down to Rare Earths
Your Generative AI Self-Study Plan
First, Make Sure that You Understand Discriminative Neural Networks Discriminative neural networks are the kinds of neural networks where you train the network with a known and pre-labeled training and testing data set. Essentially, you “have the answers in the back of the book” – and you train the network by constantly having it check… Continue reading Your Generative AI Self-Study Plan
AGI: Generative AI, AGI, the Future of AI, and You
Generative AI is about fifty years old. There are four main kinds of generative Ai (energy-based neural networks, variational inference, variational autoencoders, and transformers). There are three fundamental methods underlying all forms of generative AI: the reverse Kullback-Leibler divergence, Bayesian conditional probabilities, and statistical mechanics. Transformer-based methods add in multi-head attention and positional encoding. Generative AI is not, and never can be, artificial general intelligence, or AGI. AGI requires bringing in more architectural components, such as ontologies (e.g., knowledge graphs), and a linking mechanism. Themesis has developed this linking mechanism, CORTECONs(R), for COntent-Retentive, TEmporally-CONnected neural networks. CORTECONs(R) will enable near-term AGI development. Preliminary CORTECON work, based on the cluster variation method in statistical mechanics, includes theory, architecture, code, and worked examples, all available for public access. Community participation is encouraged.
If You’re Teaching Yourself Generative AI: Some Resources (Book Chapters)
Teaching yourself generative AI (“gen-AI”) has to be one of the hardest things in the world. The really important, classic papers – the ones that you WISH THAT YOU COULD READ – all presuppose that you have a lot of knowledge coming in, about all SORTS of things. The situation is the same as it… Continue reading If You’re Teaching Yourself Generative AI: Some Resources (Book Chapters)
CORTECONs: AGI in 2024-2025 – R&D Plan Overview
By the end of 2024, we anticipate having a fully-functional CORTECON (COntent-Retentive, TEmporally-CONnected) framework in place. This will be the core AGI (artificial general intelligence) engine. This is all very straightforward. It’s a calm, steady development – we expect it will all unfold rather smoothly. The essential AGI engine is a CORTECON. The main internal… Continue reading CORTECONs: AGI in 2024-2025 – R&D Plan Overview
CORTECONs and AGI: Reaching Latent Layer Equilibrium
The most important thing in building an AGI is the ability to repeated bring the latent layer to equilibrium. This is the fundamental capability that has been missing in previous neural networks. The lack of a dynamic process to continuously reach a free energy minimum is why we have not had, until now, a robust… Continue reading CORTECONs and AGI: Reaching Latent Layer Equilibrium
Learning Energy-Based Neural Networks
In order to read any of the classic (and important) papers on energy-based neural networks, we need to know the vocabulary and essential concepts from: In today’s associated YouTube video, we illustrate how these different terms – and their respective disciplines – are blended together, using the Salakhutdinov and Hinton (2012) paper as a reference… Continue reading Learning Energy-Based Neural Networks