Evolution of AI: From Language Models to Engineering
Early chapters cover the evolution of AI and foundation models, which gives you the big picture.
Why Begin With a Map
Most learners want to jump straight to the lesson that answers their question. This article asks you to slow down first, because it is a map of the course rather than a lesson on AI itself. It tells you what kind of material you will meet, how that material is grouped, and how to move through it without getting lost.
What do you think happens?
Suppose you want to learn how retrieval-augmented generation works. Which part of the course should you check first: the entry chapter or a later chapter?
Reveal answer
A later chapter, because the course goes deeper into topics such as retrieval-augmented generation in its later chapters.
The entry chapter prepares you for the rest of the course instead of teaching a single technical concept in depth. Use the course layout to pick a starting point rather than assuming you must read in strict order.
Where Getting Started Fits in the Sequence
The Getting Started chapter is the entry point of the course, and this article is the first stop inside it. Its purpose is to prepare you for the rest of the material. It does not teach one technical concept in depth, and it does not ask you to build a skill yet. Instead, it gives you a clear picture of how the course is arranged before you commit time to any single topic.
The course moves from broad to specific. Early chapters cover the evolution of AI and foundation models, which gives you the big picture. Later chapters go deeper into topics such as evaluating model performance, prompt engineering, retrieval-augmented generation, finetuning, and building AI applications. Each chapter assumes that the earlier ones have supplied the vocabulary it relies on, so the order carries meaning even when you choose your own starting point.
The source material gives a concrete taste of the history that the early chapters build on. The model it describes was a recurrent neural network called LSTM (Long Short-Term Memory). LSTM was the dominant architecture for natural language processing in deep learning before the transformer architecture took over in 2018. A related 2020 study showed that pairing a pre-trained language model with a retrieval system could dramatically improve its performance on factual questions. You do not need to master these ideas yet. Notice only that the later topics, such as retrieval, are built on this history.
Reading the Three Levels of the Course
The course is organized into three nested levels. Each level narrows the scope of the one above it, which is why the structure is described as moving from broad to specific. Once you can tell the levels apart, every title in the course becomes a clue about what the material covers.
Tracing a Question Through the Layout
The course is designed so you can choose a starting point rather than reading in strict order by default. The trick is to narrow your search one level at a time: first the chapter, then the module, then the sub-topic. At each level you check whether the title fits your question, and you move on if it does not.
Locating Retrieval-Augmented Generation
You want to understand retrieval-augmented generation. Use the course layout to decide where to look first.
- 1Rule out the entry chapter
Getting Started prepares you for the rest of the course and does not teach a single technical concept in depth, so it is not the place for an in-depth treatment.
- 2Scan the later chapters
The later chapters go deeper into topics such as retrieval-augmented generation, so they are the likely area to check.
- 3Check the module titles
Read each module title in the later chapters carefully. Longer titles often reveal the scope of a module, so you can confirm whether it covers retrieval before opening it.
- 4Confirm at the sub-topic level
If a sub-topic title names retrieval-augmented generation directly, you have found your starting point. If none fits, move to another likely chapter and repeat the scan.
The entry chapter gives the preparation, and a later chapter holds the in-depth material. Narrowing from chapter to module to sub-topic lets you find the right starting point without reading in strict order.
Mistakes Learners Make With the Layout
Common mistake Skipping the entry chapter and then feeling lost when a later lesson uses an idea you have not met
Why it failsEach chapter assumes that the earlier ones have given you the vocabulary it relies on, so missing the orientation makes later references hard to place.
Do this insteadSpend a few minutes on the layout first. It makes every later lesson easier to place.
Common mistake Reading a module title only as a label and ignoring the rest of it
Why it failsModule titles are longer than they first appear, and each part tells you something about the scope.
Do this insteadRead the full title and ask whether it matches your question before you spend time inside the module.
Common mistake Assuming the course must be read strictly in order
Why it failsThe course is laid out so you can choose a starting point, not only to be read from the first page onward.
Do this insteadUse the chapter, module, and sub-topic titles to pick a starting point, and return to the earlier chapters when a term is unfamiliar.
Common mistake Treating the entry chapter as an in-depth lesson on AI
Practicing the Course Map
You have a question about finetuning and want to know whether to start with an early chapter on the evolution of AI or a later chapter. Using only the chapter, module, and sub-topic logic from this article, decide which level of the course you would check first and explain why the entry chapter is or is not the right stop.
Hints
- Recall which chapters cover the big picture and which go deeper into specific topics such as finetuning.
- Think about whether the entry chapter teaches a single technical concept in depth or prepares you for the rest of the course.
Key Takeaways
- The Getting Started chapter is the entry point of the course. It prepares you for later material rather than teaching one technical concept in depth.
- The course moves from broad to specific: chapters contain modules, and modules contain sub-topics.
- Early chapters cover the evolution of AI and foundation models, while later chapters go deeper into evaluation, prompt engineering, retrieval-augmented generation, finetuning, and building AI applications.
- Use chapter, module, and sub-topic titles to narrow your search, and choose a starting point rather than assuming strict reading order.