TRYEngineering - AI Adventures - 13

Example: AI Bias
in Video Games
Ethics in AI
Bias in AI: In AI, bias refers to the unfair influence or preference for one side.
Simply put, it's like having a favorite you always choose, even if it's not the
best. For example, imagine someone working out by doing the same exercises,
and one of their friends tells them about better exercises that will make them
healthier. If an individual understands the new exercises are better for them,
but still chooses to perform the old exercises, it is indicative of bias. They prefer
their exercise routine , even if it's not ideal. Bias in AI happens when AI is trained
or created in a way that doesn't treat all possibilities fairly and favors some over
others. There are numerous ways to introduce bias into AI, which include:
* Data bias - When the AI is trained using biased data, the AI will also,
in turn, acquire bias.
* Algorithmic bias - When the AI learns in a biased manner, it favors certain
things over others, potentially leading to the AI becoming biased.
* Measurement bias - When we instruct AI to analyze data in a biased
manner, the AI will adopt this bias and analyze all data accordingly.
* Human bias - When individuals creating the AI are biased, they inadvertently
introduce this bias into the system. This can happen through any of the other
three options.
Privacy Concerns: Apps and websites collect personal information such as
your name and favorite games, a process known as data collection. Companies
use this special information to show you things you might like, but it can be
misused if not properly protected. Tools like cameras or tracking apps can watch
what you do, helping with safety and affecting privacy. To stay safe, always ask a
parent before sharing info online, use privacy settings to control what you share,
and think before you post. Safeguarding your privacy guarantees responsible
use of your information and keeps you safe.
Deepfakes: These are videos, pictures, or sounds intentionally made using
special computer programs to look and sound real, even though they are not.
Imagine you have a video of someone talking. A deepfake can change the
video to make it look like the person is saying something they never actually
said. It can also change their face or voice to look and sound like someone else.
Deepfakes can be intentionally made for fun, like swapping faces with your
favorite movie star in a video. However, people can also intentionally create
deepfakes to deceive or track others, so exercise caution and consult a trusted
adult whenever something appears unusual.
Imagine playing a video game
with numerous NPCs (nonplayable
characters), walking
around and doing different
actions. The creators use
AI to control the characters
and determine their actions
during gameplay, utilizing data
from real people to make it
more believable. However, if
the creators solely focus on
specific groups or cultures,
the characters in the game
may not accurately reflect the
behavior of all people, but only
that of a specific group. This
is an example of the biased
data used to train the AI, which
makes the characters it controls
biased toward that group, too.
Example: AI Bias in Music
Recommendations
Imagine using an app that
suggests new songs based
on what you've liked. If the
AI within the app exhibits
bias, it may only recommend
songs from specific genres or
artists, despite the possibility
that you may also enjoy other
types of music. If the AI's
training data primarily consists
of popular songs from a few
genres, it may lack knowledge
about other genres or your
personal preferences.. The
AI's training on incorrect data
could lead it to consistently
provide unsatisfactory
recommendations, as its
limited data led it to assume
that's the only music you enjoy.
TRYEngineering - AI Adventures

Table of Contents for the Digital Edition of TRYEngineering - AI Adventures
