All Writings

The convergence toward Artificial General Intelligence: Autonomous agent architectures, instrumental convergence, mechanistic interpretability, and humanity's existential stewardship of synthetic minds.

Aligning raw autoregressive intelligence: From Reinforcement Learning from Human Feedback (RLHF) and DPO to diffusion models, Chain-of-Thought deliberation, and test-time compute scaling.

The empirical physics of intelligence: Kaplan and Chinchilla scaling laws, compute-optimal training regimes, and the emergence of zero-shot in-context reasoning in trillion-parameter...

How the 2017 discovery of the Transformer architecture dismantled decades of recurrent neural network hegemony, unleashing unbounded sequence modeling through parallelized multi-head...

How Demis Hassabis and Google DeepMind conquered the 3,000-year-old game of Go: Dual policy-value networks, Monte Carlo Tree Search, and the historic...

How the convergence of massive visual data, parallel GPU supercomputing, and deep convolutional networks ignited the modern artificial intelligence revolution in October...

The psychological Rubicon of artificial intelligence: IBM Deep Blue’s 480 custom chess chips, 200 million positions per second, alpha-beta pruning, and the...

How statistical machine learning, Vladimir Vapnik’s Support Vector Machines, and convex quadratic programming displaced connectionist networks and symbolic logic throughout the 1990s.

How Kunihiko Fukushima’s Neocognitron and Yann LeCun’s LeNet-5 introduced weight sharing, local receptive fields, and spatial subsampling to conquer visual pattern recognition.

How the Parallel Distributed Processing (PDP) research group broke through the multi-layer impasse in 1986: The multivariable chain rule, internal representations, and...