Every CompTIA SecAI+ Objective Explained | EP 45

Okay, foundations first.

Domain one, that's the 17% one on the exam.

This is your, it's the vocabulary domain.

It's the building blocks so you understand everything else that sits on it.

CompTIA wants you to know, do you actually understand what these AI systems are? Not just, "I've used ChatGPT." Yeah? So you need to be able to explain how an AI works, where roughly, what makes deep learning different from regular machine learning.

How do AI models actually learn from data? Not super in-depth, but enough so you've got a foundational understanding.

The exam covers all of this.

The, the difference between AI, how you'd train it, and how you'd build it if you happen to have $100 million spare.

And domain one is really asking, if someone hands you an AI system, do you actually know how it works, how it would be trained? But the exam doesn't just stop at vocabulary.

The exam tests, like prompt engineering, and this is worth paying attention to.

Say you ask an LLM to classify a security log entry with zero examples.

That's called zero-shot prompting.

Now, give it three example entries first.

This one is a brute force attempt.

This one is normal traffic.

This one is, uh, privilege escalation.

Then ask it to classify the fourth.

That's called few-shot prompting.

These are not just SecAI terms, they're industry terms.

So the quality difference then changes.

It improves, actually.

So you can try this tonight, actually.

Install something called LM Studio, download the latest Llama model, and test it yourself.

That's testable content you can practice in 10 minutes.

The other big concept is RAG, retrieval-augmented generation.

Okay, foundations first.

Domain one, that's the 17% one on the exam.

This is your, it's the vocabulary domain.

It's the building blocks so you understand everything else that sits on it.

CompTIA wants you to know, do you actually understand what these AI systems are? Not just, "I've used ChatGPT." Yeah? So you need to be able to explain how an AI works, where roughly, what makes deep learning different from regular machine learning.

How do AI models actually learn from data? Not super in-depth, but enough so you've got a foundational understanding.

The exam covers all of this.

The, the difference between AI, how you'd train it, and how you'd build it if you happen to have $100 million spare.

And domain one is really asking, if someone hands you an AI system, do you actually know how it works, how it would be trained? But the exam doesn't just stop at vocabulary.

The exam tests, like prompt engineering, and this is worth paying attention to.

Say you ask an LLM to classify a security log entry with zero examples.

That's called zero-shot prompting.

Now, give it three example entries first.

This one is a brute force attempt.

This one is normal traffic.

This one is, uh, privilege escalation.

Then ask it to classify the fourth.

That's called few-shot prompting.

These are not just SecAI terms, they're industry terms.

So the quality difference then changes.

It improves, actually.

So you can try this tonight, actually.

Install something called LM Studio, download the latest Llama model, and test it yourself.

That's testable content you can practice in 10 minutes.

The other big concept is RAG, retrieval-augmented generation.