BU121 Final Exam Review
, Ubiquitous, Undetectable, Transformative, Super
sophisticated at predicting text, Similar to working with a
Identify characteristics of Gen AI human who has been trained by humans, Good at some
models human tasks IF you give them good instructions and
depending on what tools they have available
Ubiquitous integrated into many parts of our daily lives (being everywhere)
Undetectable invisible/unnoticeable to users
Transformative has fundamentally changed industries, workflows, daily life
Trained on vast amounts of content
Super sophisticated at
predicting text Designed to create human-like text, not truthful text
can provide incorrect output that
Similar to working with a sounds plausible can provide
human who has been trained different answers to identical
by humans prompts
BUT they are NOT human (cannot think, analyze, reason,
make ethical judgements, feel empathy or understand
cultural sensitivity)
Good at some human tasks IF you give them good instructions and depending on what tools they
have available
- Role and Goal/Identity, Audience and Purpose
Identify the features of
effective Gen AI prompts - Give very clear instructions (but DO NOT copy/paste IN
copyrighted work)
- Give examples and steps
- Where you are an expert and can assess quickly whether AI is
good or bad
Ways to use AI - Where using AI will keep you moving forward
- Where AI is better than the next best available human
- Where you need a specific perspective to improve the
work OR when you need a second opinion
- When you need to learn and synthesize new ideas or information.
Ways not to use AI - When very high accuracy is required
- When the effort is the point. (productive struggle)
Incorrect or misleading results that AI models generate.
Hallucinations These errors can be caused by insufficient training data,
incorrect assumptions made by the model, or biases in
the data used to train the model.
It performs well at some jobs, and fails in other
Opacity
circumstances in ways that are difficult to predict in
advance.
trying to make sure the behaviour of AI systems matches
Alignment
what we want and what we expect
Top-down alignment Designers explicitly specify the values and ethical principles for AI
to follow
, Ubiquitous, Undetectable, Transformative, Super
sophisticated at predicting text, Similar to working with a
Identify characteristics of Gen AI human who has been trained by humans, Good at some
models human tasks IF you give them good instructions and
depending on what tools they have available
Ubiquitous integrated into many parts of our daily lives (being everywhere)
Undetectable invisible/unnoticeable to users
Transformative has fundamentally changed industries, workflows, daily life
Trained on vast amounts of content
Super sophisticated at
predicting text Designed to create human-like text, not truthful text
can provide incorrect output that
Similar to working with a sounds plausible can provide
human who has been trained different answers to identical
by humans prompts
BUT they are NOT human (cannot think, analyze, reason,
make ethical judgements, feel empathy or understand
cultural sensitivity)
Good at some human tasks IF you give them good instructions and depending on what tools they
have available
- Role and Goal/Identity, Audience and Purpose
Identify the features of
effective Gen AI prompts - Give very clear instructions (but DO NOT copy/paste IN
copyrighted work)
- Give examples and steps
- Where you are an expert and can assess quickly whether AI is
good or bad
Ways to use AI - Where using AI will keep you moving forward
- Where AI is better than the next best available human
- Where you need a specific perspective to improve the
work OR when you need a second opinion
- When you need to learn and synthesize new ideas or information.
Ways not to use AI - When very high accuracy is required
- When the effort is the point. (productive struggle)
Incorrect or misleading results that AI models generate.
Hallucinations These errors can be caused by insufficient training data,
incorrect assumptions made by the model, or biases in
the data used to train the model.
It performs well at some jobs, and fails in other
Opacity
circumstances in ways that are difficult to predict in
advance.
trying to make sure the behaviour of AI systems matches
Alignment
what we want and what we expect
Top-down alignment Designers explicitly specify the values and ethical principles for AI
to follow
