Training Data
322 episodes · Page 4 of 14
#6147: Hebrew AI: Ears, Mouth, and Brain
Open Hebrew speech recognition beats the paid services. Hebrew text-to-speech is a different story — and the language models are a third.
#6131: Judging the Judges: How AI Evaluates Creative Writing
LLM judges match human raters — but the human ceiling is low. What that means for evaluating creative writing.
#6129: How to Build an AI Preference Dataset
Pairwise comparison, forced choice, or just "this one or that one" — the engineering behind collecting human preferences on AI outputs.
#6126: Can You Train a Machine to Be Funny?
Comedy fights everything next-token prediction is built for. Turns out you can train a model to land a joke — and then measure it.
#6110: The Chatbot OpenAI Almost Shelved
Instruction models existed for months. Then a rumour about Anthropic pushed OpenAI to ship a chat box in two weeks.
#6105: Не только Китай: ИИ в России, Индии и Японии
Китай занимает почти всё внимание в разговорах об ИИ, но Россия, Индия и Япония тоже создают мощные региональные модели с необычной архитектурой.
#6088: What 5,964 Episodes Reveal About One Show
Nearly 6,000 episodes, sorted and mapped. AI and geopolitics spike, then fade — but a third of all "nearest neighbours" cross category lines.
#6076: How AI Actually Predicts Elections
Bayesian models, "silicon sampling," and agent societies — what actually works for forecasting votes, and what's just fashion.
#6070: The Papers That Moved AI: From AlexNet to This Year's Fights
The classics all said 'learn it from data and scale it up'. This year's most-discussed AI papers are auditing what scaling took for granted.
#6068: Chinese AI Models: Refusing vs. Reframing
New studies show Chinese AI models have shifted from refusing sensitive questions to answering them — in the Party's voice.
#6062: Who Actually Shapes Chinese AI Thinking?
Beyond the lab CEOs: the researchers and ideas defining how China builds — and argues about — artificial intelligence.
Also available in:中文
#6055: How to Fine-Tune ASR on Noisy Audio
A small noise library mixed into clean speech beats collecting hours of noisy recordings. Here's the ratio that actually works.
#6049: Claude Code by the Numbers: What's Measurable, What's Hype
Anthropic won't publish a user count. Here's what revenue, surveys, and 400K sessions actually tell us about Claude Code's rise.
#6039: Who's Actually Building Frontier AI Outside the US and China?
Mistral sits 20 points off the frontier. Korea runs a bake-off. The UAE open-sources everything. What's actually driving sovereign AI?
#6032: Blending Datasets vs Merging Weights for Fine-Tuning
Merging weights happens in parameter space. Blending datasets happens in data space — and the messy part is worth it.
#6030: Who Do You Actually Hire to Train a Model?
Research engineers and data engineers are diverging. Here's who you actually hire to train a model — and why the org chart keeps changing.
#6023: How Much Voice Data Is Enough?
Two hours of phone recordings took Hebrew code-switching from 20% to 60%. But when does more data stop mattering?
#6022: Fine-Tuning for Code-Switched Hebrew on 2 Hours of Audio
Two hours of one man's voice beat models trained on 120,000 hours — but only for code-switched Hebrew.
#6014: Why Every AI Model Is Secretly a Decoder
Encoder, decoder, or both? The three transformer shapes, why one won, and why BERT still refuses to die.
#6008: Why Codex Writes Code for a Dead API
Models don't hallucinate your API — they quote real docs from the wrong year. Here's why, and what actually fixes it.
#5994: Hebrew in Latin Letters: A TTS Pipeline Fix
A token classifier that finds Hebrew words written in Latin letters so TTS stops mangling them — built for under $3.
#5982: Teaching a Voice Engine to Say "Yalla" Right
A small token classifier that spots Hebrew words hiding in English sentences — and why finding the word is the easy half.
#5975: Merging vs Distillation: Can Two Models Become One?
Merging and distillation get lumped together, but they're opposites. One composes existing capability; the other has to teach it.
#5962: How Word Error Rate Lies to Us
WER is the number everyone reports for speech recognition — and the one that agrees with human judgment least. Here's what it actually measures.