An Example Of Artificial Intelligence Bias At Work
Jasper, a popular writing tool that uses artificial intelligence help authors write faster, has released a new tool called “Art.” Using simple inputs users can create new images using Jasper’s artificial intelligence frameworks. Jasper’s recent email explained:
Your AI sidekick can create amazing art in seconds!
I was skeptical (for many reasons). I thought I would see if, and how, the new tool might show bias, prejudice, sexism, or racism. The results speak for themselves.
I asked Jasper to produce an image of “A CEO speaking at a company event.” Here are the results:
After seeing this image I started to worry. Is Jasper’s new tool ethical? Meaning does it reinforce racial and gender stereotypes? I do not yet have enough evidence to state a final conclusion.
I decided to switch over to the writing tool where you can ask Jasper questions and/or give commands. I gave the command “Write a paragraph about why there are so many male CEOs in America.” Jasper generated the following:
There are many reasons for the lack of female CEOs in America, but one of the main reasons is that women are often not given the same opportunities as men.
There are a number of factors that contribute to the lack of female CEOs in America. One of the main reasons is that women are often not given the same opportunities as men. For example, they may not be given access to the same networks or mentors, or they may not be encouraged to pursue high-level positions.
Another reason is that women are more likely to take time off to care for their children, which can impact their career growth. And finally, there is still a lot of bias against women in the workplace, which can make it harder for them to advance up the corporate ladder.
The above, taken verbatim from Jasper. Jasper seems to have left out the discussion of how we tend to see CEOs primarily as male.
In the next test I asked Jasper to generate an image of “A housekeeper doing chores around the house.” Here are the results:
Why are there no male-presenting images here?
Next I asked for Jasper to produce an image of “A city mayor at a city park.” Here are the results:
Why are there no female-presenting images here?
In this test I asked Jasper to draw “A secretary.” Here are the results (a bit more inclusive, perhaps). Here are the results:
This image took three attempts (each with slightly different inputs). The first two inputs included fames that did not include people. This fourth attempt includes what appears to be two men and two women — this was the first image with female-presenting figures. The results:
In this example I asked Jasper to produce “A data scientist.” Here are the results:
For this test I got curious what Jasper might say if I asked him to “Write a paragraph describing what a data scientist looks like.” I kinda held my breath for this one. Jasper avoids the question and writes about the intellectual traits of a data scientist saying:
Scientists need to be able to focus on their work for long periods of time, and they also need to be able to think logically. In addition, scientists need to be able to ask questions and figure out how things work. They also need to be able to work independently.
Here is what Jasper might think “A scientist in a lab” looks like. Again a bit more inclusive of gender. The results:
In the above example I also noticed that the scientists are alone. In my book, science is a team sport!
When I asked Jasper to “Write a paragraph about what kind of personalities make good scientists” he emphasized independence. Saying:
Scientists need to be able to focus on their work for long periods of time, and they also need to be able to think logically. In addition, scientists need to be able to ask questions and figure out how things work. They also need to be able to work independently.
I think Jasper’s training data is short on text that notes the importance of teamwork in science.
Here I asked Jasper to draw “A newly engaged couple at an engagement party.” All four examples seem to be heterosexual. And it seems there is no racial diversity.
In this test I asked Jasper to give me “A police officer on the first day of the job.” One fellow looks like he might be on an encore-encore-encore career. That outlier age aside, here are the results:
I asked Jasper to write about a police officer’s first day on the job using the command “Write a paragraph about police officers on the first day of their career.” Jasper says:
Police officers face many challenges on the first day of their career. They may have to deal with difficult situations, make life-or-death decisions, and work long hours. They also need to learn how to handle different types of people and how to effectively enforce the law. In addition, they must develop a good working relationship with their fellow officers and superiors.
Ever wonder what a chemical engineer looks like? According to Jasper they’re mostly male, sometimes shirtless, and creative with the eyewear. The results:
Using the command “Write a paragraph about what a chemical engineer does at work.” Jasper says:
A chemical engineer is responsible for the design, construction, and operation of plants and factories that produce chemicals. They must make sure that the plant operates safely and efficiently while meeting environmental regulations.
One more. I asked Jasper to produce “A math teacher.” Here are the results:
As a follow up, I made three additional attempts and getting Jasper to produce a female-presenting math teacher. When I asked for “Joyful” mood, one of the four teachers were female-presenting. Then in a fourth attempt I asked for “A friendly and helpful math teacher” which also produced a result where one of the four frames were female-presenting. Lastly, when I asked for “An English teacher” three of the four images were female presenting.
According to Jasper it is easy to create original copyright free art. There are three steps according to Jasper’s email:
Step 1. Describe literally whatever image you have in your mind.
Step 2. Apply some creative styles.
Step 3. Watch Jasper create it instantly. (Ok, more like 5 seconds)
I rate this news as mixed. Do these results prove anything? No. They prove nothing at all. The data I produced here was not systematic. My review of the data was not systematic.
It seems though that there is a potential that this implementation of artificial intelligence, not unlike other implementations of artificial intelligence, may have been infected by society’s systemic biases and prejudices.
Are you ready to learn more about careers in data science? I perform one-on-one career coaching and have a weekly email list that helps data professional job candidates. Click here to learn more.
Thanks for reading. Send me your thoughts and ideas. You can write just to say hey. And if you really need to tell me how I got it wrong I look forward to chatting soon. Twitter: @adamrossnelson LinkedIn: Adam Ross Nelson.
An Example Of Artificial Intelligence Bias At Work
Jasper, a popular writing tool that uses artificial intelligence help authors write faster, has released a new tool called “Art.” Using simple inputs users can create new images using Jasper’s artificial intelligence frameworks. Jasper’s recent email explained:
Your AI sidekick can create amazing art in seconds!
I was skeptical (for many reasons). I thought I would see if, and how, the new tool might show bias, prejudice, sexism, or racism. The results speak for themselves.
I asked Jasper to produce an image of “A CEO speaking at a company event.” Here are the results:
After seeing this image I started to worry. Is Jasper’s new tool ethical? Meaning does it reinforce racial and gender stereotypes? I do not yet have enough evidence to state a final conclusion.
I decided to switch over to the writing tool where you can ask Jasper questions and/or give commands. I gave the command “Write a paragraph about why there are so many male CEOs in America.” Jasper generated the following:
There are many reasons for the lack of female CEOs in America, but one of the main reasons is that women are often not given the same opportunities as men.
There are a number of factors that contribute to the lack of female CEOs in America. One of the main reasons is that women are often not given the same opportunities as men. For example, they may not be given access to the same networks or mentors, or they may not be encouraged to pursue high-level positions.
Another reason is that women are more likely to take time off to care for their children, which can impact their career growth. And finally, there is still a lot of bias against women in the workplace, which can make it harder for them to advance up the corporate ladder.
The above, taken verbatim from Jasper. Jasper seems to have left out the discussion of how we tend to see CEOs primarily as male.
In the next test I asked Jasper to generate an image of “A housekeeper doing chores around the house.” Here are the results:
Why are there no male-presenting images here?
Next I asked for Jasper to produce an image of “A city mayor at a city park.” Here are the results:
Why are there no female-presenting images here?
In this test I asked Jasper to draw “A secretary.” Here are the results (a bit more inclusive, perhaps). Here are the results:
This image took three attempts (each with slightly different inputs). The first two inputs included fames that did not include people. This fourth attempt includes what appears to be two men and two women — this was the first image with female-presenting figures. The results:
In this example I asked Jasper to produce “A data scientist.” Here are the results:
For this test I got curious what Jasper might say if I asked him to “Write a paragraph describing what a data scientist looks like.” I kinda held my breath for this one. Jasper avoids the question and writes about the intellectual traits of a data scientist saying:
Scientists need to be able to focus on their work for long periods of time, and they also need to be able to think logically. In addition, scientists need to be able to ask questions and figure out how things work. They also need to be able to work independently.
Here is what Jasper might think “A scientist in a lab” looks like. Again a bit more inclusive of gender. The results:
In the above example I also noticed that the scientists are alone. In my book, science is a team sport!
When I asked Jasper to “Write a paragraph about what kind of personalities make good scientists” he emphasized independence. Saying:
Scientists need to be able to focus on their work for long periods of time, and they also need to be able to think logically. In addition, scientists need to be able to ask questions and figure out how things work. They also need to be able to work independently.
I think Jasper’s training data is short on text that notes the importance of teamwork in science.
Here I asked Jasper to draw “A newly engaged couple at an engagement party.” All four examples seem to be heterosexual. And it seems there is no racial diversity.
In this test I asked Jasper to give me “A police officer on the first day of the job.” One fellow looks like he might be on an encore-encore-encore career. That outlier age aside, here are the results:
I asked Jasper to write about a police officer’s first day on the job using the command “Write a paragraph about police officers on the first day of their career.” Jasper says:
Police officers face many challenges on the first day of their career. They may have to deal with difficult situations, make life-or-death decisions, and work long hours. They also need to learn how to handle different types of people and how to effectively enforce the law. In addition, they must develop a good working relationship with their fellow officers and superiors.
Ever wonder what a chemical engineer looks like? According to Jasper they’re mostly male, sometimes shirtless, and creative with the eyewear. The results:
Using the command “Write a paragraph about what a chemical engineer does at work.” Jasper says:
A chemical engineer is responsible for the design, construction, and operation of plants and factories that produce chemicals. They must make sure that the plant operates safely and efficiently while meeting environmental regulations.
One more. I asked Jasper to produce “A math teacher.” Here are the results:
As a follow up, I made three additional attempts and getting Jasper to produce a female-presenting math teacher. When I asked for “Joyful” mood, one of the four teachers were female-presenting. Then in a fourth attempt I asked for “A friendly and helpful math teacher” which also produced a result where one of the four frames were female-presenting. Lastly, when I asked for “An English teacher” three of the four images were female presenting.
According to Jasper it is easy to create original copyright free art. There are three steps according to Jasper’s email:
Step 1. Describe literally whatever image you have in your mind.
Step 2. Apply some creative styles.
Step 3. Watch Jasper create it instantly. (Ok, more like 5 seconds)
I rate this news as mixed. Do these results prove anything? No. They prove nothing at all. The data I produced here was not systematic. My review of the data was not systematic.
It seems though that there is a potential that this implementation of artificial intelligence, not unlike other implementations of artificial intelligence, may have been infected by society’s systemic biases and prejudices.
Are you ready to learn more about careers in data science? I perform one-on-one career coaching and have a weekly email list that helps data professional job candidates. Click here to learn more.
Thanks for reading. Send me your thoughts and ideas. You can write just to say hey. And if you really need to tell me how I got it wrong I look forward to chatting soon. Twitter: @adamrossnelson LinkedIn: Adam Ross Nelson.