Bitzu

AIF-C01 ยท GenAI strengths, limits and AWS services

GenAI: strengths and limits

Adaptability, conversational ability and content generation versus hallucinations, nondeterminism, inaccuracy and limited interpretability.

Key points

  • Strengths: adapts to many tasks without retraining, converses in natural language, generates text, code and images fast.
  • Limits: hallucinations (confident but false output), nondeterminism (same prompt can give different answers), inaccuracy, and limited interpretability (hard to explain why a model produced an answer).
  • Treat output as a draft that needs validation.

Source: https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html

Practice questions

Single choice

What is a hallucination in a generative AI model?

  1. Plausible-sounding output that is factually incorrect or made up
  2. A slow response caused by an overloaded inference endpoint
  3. A different answer returned each time the same prompt is sent
  4. A refusal to answer because a safety filter was triggered
Show answer

Answer: A. Plausible-sounding output that is factually incorrect or made up

A hallucination is fluent, confident output that is factually wrong or invented. Nondeterminism is variation between runs, not falsehood.

Select all that apply

Which statements describe advantages of generative AI?

  1. One model can be adapted to many different tasks through prompting
  2. It interacts with users in natural conversational language
  3. It always returns exactly the same output for the same prompt
  4. It can fully explain the reasoning behind each output token
  5. It guarantees factual accuracy of every generated statement
Show answer

Answer: A. One model can be adapted to many different tasks through prompting, B. It interacts with users in natural conversational language

Adaptability across tasks and conversational interaction are strengths. Guaranteed identical output and full explainability are not properties of GenAI; accuracy is not guaranteed either.

True or false

True or false: if a model's answer sounds confident and well written, it is likely to be factually correct.

Show answer

Answer: False

False. Fluency says nothing about accuracy; hallucinations are usually delivered in a confident tone.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More AIF-C01 topics in this module