LLMs and Generation

Question 1

Explain what is meant when an AI us described as super-autocomplete. Temp

Answer 1

Generative AI uses statistical probabilities and temperatures to ‘autocomplete’ the remainder of the sentence after every word

Question 2

What is “context” and why result in better AI performance

Answer 2

Helps the AI understand what is trying to be achieved. It helps narrow the probabilities.

Question 3

What is “emergence” in the context of AI. Give Example

Answer 3

Large language models are able to perform qualitatively different tasks as they get larger e.g do maths

Question 4

Where does training data come from and what problems does this create?

Answer 4

Possible bias, data poisoning, privacy, uneven representation, needs updating

Question 5

What is “prompt” engineering” and why is it undertaken?

Answer 5

Designing a prompt to help fine-tune the AI inputs for high-quality and relevant outputs.

Question 6

Value of iteration and refinement in prompt engineering?

Answer 6

Helps to specify the requirements of the output of the AI to get a more helpful answer.

Question 7

What is the value of telling Generative AI your goal when writing a prompt?

Answer 7

Help fine-tune the output

Question 8

What is the value of providing specific context in prompt engineering?

Answer 8

Help fine-tune the output

Question 9

What is the value of including examples in prompt

Answer 9

Help fine-tune the output

Question 10

Could you use AI to create a prompt for itself? How could this work and would it be useful?

Answer 10

Yes

• ask more