Lesson #3
Under the Hood: Models, Meaning & Speed
- 1
Machine Learning
Fundamentals“muh-SHEEN LURN-ing”
2/5 · Getting thereA way of teaching a computer by showing it lots of examples instead of writing out every rule.
Think of it like… Teaching a new hire by letting them watch a hundred jobs, not by handing them a rulebook.
Construction: A system that learns which bids you usually win and flags the best ones to chase.
Business: Software that learns which customers are likely to cancel next month.
Why this matters: Almost every AI tool you buy is machine learning under the hood — knowing that helps you ask better questions.
Open full entry → - 2
Neural Network
Models“NOOR-ul NET-wurk”
3/5 · TrickierA math structure loosely modeled on the brain, made of layers that pass signals along to spot patterns.
Think of it like… A crew passing a measurement down the line, each person refining it a little before the next.
Construction: Software that spots cracks in bridge photos taken by a drone.
Business: A system that reads scanned invoices and pulls out the totals.
Why this matters: Neural networks are the engine behind modern AI — the reason it handles messy, real-world input.
Open full entry → - 3
Token
Models“TOH-kun”
2/5 · Getting thereA small chunk of text — roughly part of a word — that an AI reads and writes one piece at a time.
Think of it like… Bricks in a wall: the AI doesn't see the wall, it lays one brick after another.
Construction: A long spec sheet costing more to process because it's simply more tokens.
Business: Your AI bill going up because every chat message is priced per token.
Why this matters: Tokens are how AI is measured and billed — they explain both cost and length limits.
Open full entry → - 4
Embedding
Data“em-BED-ing”
4/5 · AdvancedTurning words or documents into numbers so a computer can tell which ones mean similar things.
Think of it like… Pinning every tool in your shop on a map so similar tools end up sitting near each other.
Construction: Searching your job archive for 'water damage' and also getting 'moisture intrusion' results.
Business: A help desk matching a new ticket to past tickets that meant the same thing.
Why this matters: Embeddings are why AI search finds meaning, not just matching words — a huge upgrade for your files.
Open full entry → - 5
Vector Database
Infrastructure“VEK-tor DAY-tuh-bays”
4/5 · AdvancedA place to store embeddings so an AI can quickly find the most similar pieces of your content.
Think of it like… A filing cabinet organized by meaning instead of by alphabet.
Construction: Storing ten years of site reports so an AI can pull the three most relevant in a second.
Business: A company handbook an AI can search and quote accurately.
Why this matters: It's the storage layer that lets AI answer from your own documents instead of guessing.
Open full entry → - 6
Multimodal
Models“muhl-tee-MOH-dul”
3/5 · TrickierAn AI that can handle more than one kind of input — text, images, audio, or video together.
Think of it like… A foreman who can read the plans, look at the site, and listen to the crew.
Construction: Uploading a site photo and asking the AI to write the punch list.
Business: Sending a screenshot of an error and getting written steps to fix it.
Why this matters: Most real work isn't just text — multimodal AI can finally use the photos and calls you already have.
Open full entry → - 7
Temperature
Models“TEM-pur-uh-chur”
3/5 · TrickierA setting that controls how predictable or creative an AI's answers are.
Think of it like… The thermostat on an oven — low and steady for bread, hot and fast for a sear.
Construction: Keeping it low so safety instructions come out the same way every time.
Business: Turning it up when brainstorming marketing taglines.
Why this matters: One slider decides whether AI is reliable or inventive — matching it to the task avoids surprises.
Open full entry → - 8
Benchmark
Fundamentals“BENCH-mark”
2/5 · Getting thereA standard test used to compare how well different AI models perform on the same task.
Think of it like… A load test on materials — same weight, same conditions, honest comparison.
Construction: Testing two estimating tools on the same ten past jobs.
Business: Comparing two support AIs on the same 50 real customer questions.
Why this matters: Vendors all claim 'best AI' — a benchmark on your own work is how you actually find out.
Open full entry → - 9
Latency
Infrastructure“LAY-ten-see”
2/5 · Getting thereThe delay between asking an AI something and getting the answer back.
Think of it like… The pause between flipping a switch and the light actually coming on.
Construction: A field app that stalls when checking specs on a weak signal.
Business: A chat widget losing customers because replies take eight seconds.
Why this matters: Slow AI gets abandoned — speed often matters more to users than a slightly smarter answer.
Open full entry → - 10
Human in the Loop
Agents“HYOO-mun in the LOOP”
2/5 · Getting thereA setup where a person reviews or approves what the AI does before it counts as final.
Think of it like… An apprentice doing the work and the foreman signing off before it ships.
Construction: AI drafting the change order; the PM approves before it goes to the client.
Business: AI writing the customer refund email, a rep hitting send.
Why this matters: It's the safest way to adopt AI now — you get the speed without handing over the final call.
Open full entry →