A connection

How did Food Triage lead to Large Language Models?

8 steps · 7 domains · 302,020 years

Food Triagesince the species began · 300,000 BCThe hard-won knowledge of what is safe to eat and how cooking turns poison into food — the difference between a full belly and a fatal mistake.
que precisou dela — tornou-a fabricável ↑
Control of Fire~1,000,000 BC · 1,000,000 BCThe ability to keep, feed, and eventually start a fire at will — humanity's first external source of energy.
tornou-a fabricável ↓
Distillation lives or dies on gentle, steady heat — hot enough to boil the light fraction, never so hot it boils everything at once.
Distillation~800 AD (alembic systematized in the Islamic world) · 800 ADBoiling a mixture and catching the vapor, so that the parts that boil easiest come off first and condense back to a purer liquid. Separation by temperature — it makes nothing new, it only sorts.
que precisou dela — tornou-a possível ↑
Clear glass lets the distiller watch the boil and condensation and resists the acids and hot vapors that corrode metal and clay.
Glass~1,500 BC (vessels) · 1,500 BCSand melted with a flux of soda or plant ash until it fuses into a transparent solid — shapeable while hot, rigid and chemically inert forever after.
tornou-a possível ↓
Lenses~1286 AD (spectacles) · 1286 ADGlass ground into curved surfaces that bend light to a focus — correcting aging eyes, and, stacked in tubes, revealing worlds too far or too small to see at all.
tornou-a fabricável ↓
Photolithography1957 AD · 1957 ADPrinting with light: project a circuit pattern onto a light-sensitive coating on a silicon wafer, so billions of microscopic transistors can be made in a single exposure.
tornou-a fabricável ↓
Integrated Circuit1959 AD · 1959 ADAn entire circuit — transistors, resistors, wiring — manufactured as one piece on a single sliver of silicon.
tornou-a possível ↓
The Internet1969–1983 AD · 1983 ADA network of networks with no center: any computer can reach any other by passing packets through whoever is in between.
tornou-a possível ↓
The training corpus IS the internet; no other artifact contains enough recorded human language.
Large Language Models2018–2022 AD · 2020 ADNeural networks trained on internet-scale text to predict the next word — and which, at sufficient scale, turn out to translate, code, reason, and converse without being taught any of those skills directly.

see it on the map trace your own connection

Every link in this chain is a verified dependency — not a trivia association. Explore either end: Food Triage or Large Language Models.