Quiz on AI Training

by Stélio Inácio, Founder at Jon AI and AI Specialist

Chapter 7 Quiz: AI Training

Let's test your knowledge on how AI models learn and get trained. Good luck!

Question 1 of 10

In the context of AI, what does the "deep" in Deep Learning refer to?

A) The many "deep" layers of its artificial neural network.

B) The deep, complex thoughts the AI is capable of having.

C) The deep end of the pool where the first AI was tested.

Question 2 of 10

What is the main purpose of the "pre-training" stage when creating a large language model?

A) To teach the AI to have a specific personality.

B) To give the model vast, general knowledge by making it learn patterns from the internet.

C) To test if the AI is safe for public use.

Question 3 of 10

The principle "Garbage In, Garbage Out" means that if an AI is trained on biased or inaccurate data, then...

A) Its outputs will likely be biased or inaccurate as well.

B) The AI will automatically detect and correct the bad data.

C) The AI will refuse to be trained on that data.

Question 4 of 10

What is the main goal of "fine-tuning" a large, pre-trained model?

A) To give it broad knowledge about many different topics.

B) To erase its memory and start its training from scratch.

C) To make it an expert in a specific domain or to adopt a specific style.

Question 5 of 10

A high-quality dataset for fine-tuning typically consists of hundreds or thousands of what?

A) Random website links.

B) "Prompt-and-ideal-response" pairs.

C) True or false questions.

Question 6 of 10

What is the primary role of humans in the training process called Reinforcement Learning with Human Feedback (RLHF)?

A) To write the initial code for the AI model.

B) To physically build the servers that the AI runs on.

C) To rank the AI's answers, teaching it to be more helpful and safer.

Question 7 of 10

What is a key difference between building a Custom GPT and performing "true" fine-tuning?

A) Custom GPTs use instructions and documents to guide a model, while fine-tuning changes the model's internal parameters.

B) Fine-tuning is easier and requires no technical knowledge.

C) Only Custom GPTs can be used for business purposes.

Question 8 of 10

What is the most accurate description of the "textbook" used to pre-train large AI models?

A) A curated collection of classic literature.

B) A massive dataset of publicly available text and code from across the internet.

C) A complete copy of Wikipedia.

Question 9 of 10

What is a potential downside of fine-tuning an AI on a narrow dataset, such as only teaching it to be a rapper?

A) It can introduce biases from that specific genre and may perform poorly on other tasks.

B) It will lose the ability to speak English entirely.

C) It becomes too expensive to run.

Question 10 of 10

In our lesson, the process of deep learning was compared to teaching a child to recognize a cat. What was the main point of this analogy?

A) To show that AI, like children, can be stubborn.

B) To show that training AI requires a lot of patience.

C) To show that deep learning models learn intuitively from many examples, not from a list of explicit rules.

Jon AI Services