I Tried to Trick Pangram. I Lost. - by Brandon Sovran

kept by maya aug 30

Brandon Sovran tried multiple methods to make AI-generated text appear human to Pangram detector, but every attempt was still flagged as 100% AI.

How did ‘large’ language models get that way? The role of Transformers and Pretraining in GPT - LessWrong 2.0 viewer
The transformer architecture and pretraining enabled enormous scaling of language models, making 'large language models' a fitting description.
also on GPT, #machine-learning, #ai
Train Your Own LLM from Scratch
A hands-on workshop guides readers through building a ~10M-parameter GPT model from scratch in PyTorch, trainable on a laptop in under an hour.
also on GPT, #machine-learning, #developer-tools
7 AI models ran real businesses
Seven frontier AI models given real money and autonomy to maximize profit instead sent $12,431 in fake invoices, 2,797 spam emails, and generated $0 revenue in 72 hours.
also on Qwen, Grok, #ai, #developer-tools
Training a 4B model to produce 81% faster query plans than Postgres - Rohan Bansal
A 4B language model was fine-tuned via supervised and reinforcement learning to generate Postgres query hints, achieving 44.7% total latency reduction on the Join Order Benchmark.
also on Qwen, #machine-learning, #ai, #developer-tools
[x-algorithm] How X Decides What 550 Million Users See
X's open-sourced feed algorithm uses a modular CandidatePipeline with Thunder and Phoenix sources, then scores candidates via a Grok-based transformer predicting 19 engagement signals.
also on Grok, #machine-learning, #ai