What AI is, what a model is, what all these words mean.
02
Usage
Use the tools daily, and with intent.
PHASE 2
Practical AI
Use it well, then look behind every tool you used.
03
Mechanism
Open every box you were handed.
PHASE 3
Technical core
How an LLM actually works, token by token.
04
Build
Write the thing yourself. No framework.
PHASE 4
Making your own
Build an agent from zero and ship it.
One rule runs through all four: no magic. Any black box you cannot explain is a gap, not a feature.
Framing · ECE analogy
You might already know some of these concepts
A neuron
=
a MAC unit followed by a nonlinearity
Gradient descent
≈
the LMS adaptive filter from your DSP course
A GPU
=
a very large parallel DSP farm
Quantization
=
quantization. Same word, same idea.
When we discuss each one in detail, we will show you these links again.
Phase 1
Orientation
What is AI, and what do all these words mean?
1.1 Buzzwords1.2 What AI is1.3 Models & shells1.4 Agents & harnesses1.5 Limits
1.1 · The buzzword parade
“Here is every word the AI world throws at you. We are not explaining these today. This is the map of what you will understand by the end of the course — when a term's turn comes, we will open it properly.”
Awareness only. Zero teaching.
1.1 · The buzzword parade
Every word you will meet online
PRODUCTS
ChatGPT
claude.ai / Claude
Codex
Claude Code CLI
MODELS & TIERS
model
parameters
Haiku / Sonnet / Opus
OPENNESS
open source
open weight
THE LOCAL STACK
llama.cpp
Ollama
LM Studio
GGUF · MLX
safetensors
quantization (Q5_K_M)
TurboQuant · KV cache
THE AGENT STACK
agent
harness
tools
skills
MCP
LoRA
Twenty-two terms. By the end of the course, zero of them are pending.
Session mechanic
Rate yourself on all 22 words
0
Never heard it
1
Heard it
2
Can explain it
Keep your sheet. We run the exact same exercise on the last day. That delta is the course.
1.2 · What AI actually is
Machines imitating things humans do
Language, recognition, judgment — performed without anyone writing an explicit rule for each case.
1.2 · What AI actually is
AI is not just machine learning
ARTIFICIAL INTELLIGENCE
MACHINE LEARNING
The machine fits its own rules from examples.
Example: a spam filter trained on a million labelled emails.
DEEP LEARNING
LLMs live here
One way to build AI. Today, the most common one.
ALSO AI, WITH NO LEARNING AT ALL
Rule-based systems
An engineer writes if / then rules: “if cell temperature > 45 °C, stop charging.”
Search
Route finding in a maps app: try the possible paths, keep the shortest one.
Expert systems
MYCIN (1970s): about 600 rules from doctors, used to suggest antibiotics.
Hand-written game players
Deep Blue beat Kasparov in 1997 with fast search and hand-built scoring rules.
AI contains machine learning. But you can have AI without machine learning.
1.2 · What AI actually is
An “AI” with no learning: a battery charger
# Rule-based Li-ion charger controller. Every rule written by a human.
V, I, T = read_sensors() # volts, amps, °C
if T > 45or T < 0:
stop_charging() # unsafe temperature
raise_fault("TEMP")
else if V < 3.0:
charge(current = 0.1 * C) # deep discharge: trickle
else if V < 4.2:
charge(current = 1.0 * C) # constant-current stage
else if I > 0.05 * C:
hold_voltage(4.2) # constant-voltage stage
else:
stop_charging() # full
show("Battery full")
INPUT
Three sensor readings
Voltage, current, temperature.
RULES
Written from the datasheet
An engineer chose every threshold.
LEARNING
None
Nothing was fitted from data.
Humans wrote every rule. No learning happened, yet it imitates a human decision: the one a careful technician would make.
1.2 · What AI actually is
The difference is who writes the rules
SOFTWARE YOU HAVE WRITTEN
A human writes every rule
You decide each case in advance, by hand — if subject contains "lottery". You maintain that list forever, and it fails on the case you did not foresee.
MACHINE LEARNING
Nobody writes the rules
Show it a million labelled emails and the rule is fitted from the examples. It fails on the case that looks unlike anything it was shown.
Now try to hand-write the rules for “is there a cat in this photo?” The moment you cannot, you have found the reason this field exists.
1.2 · Untangling the umbrella terms
AI, machine learning, deep learning
ARTIFICIAL INTELLIGENCE
MACHINE LEARNING
DEEP LEARNING
LLMs live here
AI
The goal: human-like behaviour from a machine. Includes hand-written rule systems.
Machine learning
One way to get there: fit parameters to data instead of writing rules.
Deep learning
ML with many stacked layers of those parameters. Everything in this course sits here.
1.2 · Testing for imitation
The Turing test
WHAT IT IS · TURING, 1950
The “imitation game”. A judge chats by text with a human and a machine, both hidden. If the judge cannot tell which is which, the machine passes.
WHY IT MATTERS
“Can machines think?” has no clear answer. Turing swapped it for a question you can test.
TODAY
Modern chatbots can fool many people in a short chat. That shows good imitation. It does not show understanding.
Magic-killer
It is not a brain in a computer
It is pattern-fitting at scale that imitates human-like behaviour. Impressive imitation is still imitation.
1.2 · What AI is not, yet
AGI and superintelligence
WE ARE HERE
TODAY · NARROW AI
Today’s AI
Very strong at some tasks: chat, code, images. Still fails at simple things outside its patterns.
→
NOT BUILT YET
AGI
General intelligence: matches people at nearly every mental task, and learns new ones the way you do.
→
NOT BUILT YET
Superintelligence
Goes past the best humans at nearly everything.
If a machine could imitate 100% of what a human brain does, that would be AGI. Nobody has built it yet. Even the best models you use today are not AGI.
1.2 · A short history
Four eras, each fixing the last one's bottleneck
01
Rule-based systems
Humans write every if-then. Breaks the moment reality is messy.
02
Statistical ML
Learn from data — but a human still hand-designs the features.
03
Deep learning
The network learns the features too. Needs data and compute.
04
Transformers & LLMs
One architecture that scales, applied to nearly every modality.
1.2 · Why now · Part 1
The maths was always there
These methods existed for decades. Backpropagation, the method that trains today’s models, is from 1986.
1.2 · Why now · Part 2
What changed is scale
INPUT 1
Data
THEN · 199860,000 digit images
NOW · 2024~15 trillion words
The internet became a training set. Bars not to scale: the real gap is far bigger.
INPUT 2
Compute
THEN · CPUa few cores
NOW · GPUthousands of MACs
A GPU is a parallel MAC farm. Training runs use thousands of GPUs at once.
INPUT 3
One architecture
THEN · LENET~60 thousand weights
NOW · TRANSFORMERhundreds of billions
The transformer (2017) keeps improving as you make it bigger, for text, images, audio and code.
Same maths+scale=today’s AI
1.2 · ECE analogy
AI is already inside your syllabus
01
Earbud noise cancellation
A small network separating voice from noise, in real time, on a tiny DSP.
02
Camera image signal processors
Denoise, demosaic and night mode are learned, not hand-tuned filters.
03
5G channel estimation
Networks predicting channel state faster than classical estimators.
04
Predictive maintenance
Vibration and current signatures classified before a machine fails.
AI is not only a computer-science topic. It is already in your own subject.
1.2 · Beyond ECE
It is in every field, and in your day
01
Medicine
Reads X-rays and eye scans to flag disease early.
02
Agriculture
A phone photo of a leaf identifies the crop disease.
03
Finance
Flags a fraud card payment in milliseconds.
04
Maps
Predicts traffic and your arrival time.
05
Phone cameras
Night mode and portrait blur are computed by a network.
06
Translation
Live captions and translation between dozens of languages.
You probably used AI today before this class started.
Why this course exists
AI is not optional
Almost every task you do can be made faster or better with AI. It has become a built-in part of how work gets done.
1.2 · Where AI already sits
AI is already inside everyday work
Writing & email
Drafts, rewrites, summaries of long threads.
Coding
Writes, explains and debugs code, including firmware.
Search
Answers with sources instead of ten blue links.
Design
Diagrams, slides and images from a description.
Data analysis
Cleans a CSV of lab readings and plots it.
Circuit & chip design
PCB auto-routing, layout and verification help in EDA tools.
Documentation
Reads a 400-page datasheet and finds the register you need.
Meetings
Transcribes, then lists the decisions and action items.
Study
Explains a topic at your level, then quizzes you on it.
Why this course exists
You can’t skip AI
AI is the new internet: a personal-computer moment. ECE already depends on computers and the internet. It will depend on AI the same way.
1980s
Personal computer
Circuit simulation and CAD moved onto your desk.
1990s – 2000s
Internet
Datasheets, forums and open-source tools a search away.
NOW
AI
A tool that reads, writes, codes and designs with you.