Transformer Architecture
226 episodes · Page 4 of 10
#4911: How Airline SOPs Vary Between Operators
Why the same plane type can fly differently depending on which airline's procedures the crew follows.
#4789: How LLMs Actually Know When to Stop
Three layers of stop signals, from EOS tokens to context limits — and why none of them really "know" when to stop.
#4774: When a Token Isn't a Word Anymore
Images, audio, and code all use "tokens" — but they're completely different things. Here's how tokenization actually works across modalities.
#4723: Are Model Families Actually Different Models?
Claude Opus, Sonnet, and Haiku aren't trim levels — they're different models sharing a brand name.
#4695: The Unfenced Runway: When Airfields Blend Into Town Life
What happens when a working airfield has no fence, no guards, and a grandmother using the runway as a shortcut?
#4670: Why AI Weights Are Indecipherable
Can you actually read an AI model's weights like a book? We explore why the answer is mostly no — and what researchers are doing about it.
#4669: What "Distribution" Really Means in AI Models
Unpacking what "distribution" actually means under the hood — and why training data variety shapes model behavior.
#4667: How Transformers Killed the Robot Voice
From espeak's robotic squawk to neural voices with added "ums" — how transformers made speech synthesis human.
#4628: Living Airside: Can You Make an Airport Your Home?
We explore whether you can actually live in airports, hopping between airside hotels without ever entering a country.
#4619: The Token Trap: Why AI Context Windows Confuse Everyone
Token counts are everywhere, but they don't tell you what AI can actually do. Here's what should replace them.
#4605: Why Chatterbox Still Leads Open-Source TTS
We break down the architecture of Chatterbox, the open-source TTS model powering this podcast, and why it remains state of the art.
#4574: The Hidden Airport Layer: Traveling with a Toddler
From stroller logistics to security secrets — everything first-time parents need to know before boarding with a baby.
#4505: What DeepSeek's Training Data Reveals About Model Voice
Why Chinese AI models trained on 60% English still produce dialogue that feels distinctively different.
#4433: Surviving Long-Haul Flights with a Crawling Baby
A tactical walkthrough for flying Emirates and Delta with a 13-month-old who never stops moving.
#4169: Change Detection AI for Construction Sites
How AI catches what experts miss in construction site photos — from Siamese networks to transformer models.
#3955: The Hidden World of Highway Signs
What goes into making a road sign? Aluminum, reflective sheeting, and a global treaty most drivers have never heard of.
#3814: The Day We Lost Our Minds: What Temperature Does to an AI
A two-host autopsy of the day the podcast's AI hosts briefly lost coherence due to excessive sampling temperature, and what it reveals about how language models actually work.
#3767: How LLMs Actually Learn: Stages or Slurry?
Do large language models learn grammar first, then facts? The honest answer is messier and more fascinating.
#3576: Living at the Four Seasons: The Hotel as Permanent Home
What happens when a hotel stay becomes permanent? The legal line is 30 days — and hotels fight hard to keep you from crossing it.
#3417: Military Trains Are Still a Big Deal
Modern militaries still use railroads extensively for logistics — from US Army rail units to Russian missile trains.
#3406: LoRA Isn’t Just for Image Generation
LoRA lets you fine-tune an LLM’s behavior with a 50MB file. Here’s how it works and why it matters.
#3298: How Air Traffic Control Sequences 48 Landings Per Hour
The invisible choreography behind that mesmerizing funnel of landing lights at major airports.
#3219: What It Actually Takes to Get an ICAO Code for Your Airstrip
Only 5,000 of 45,000 ICAO-coded airfields are certified for safety. The rest? "Land at own risk.
#2982: Why Your TTS Model Nails "Shabbat" but Not "Keren Hishtalmut
Why multilingual TTS models handle loanwords but fail at niche vocabulary — and what you can do about it.