Meta's AI image generator was found to be unable to imagine an Asian man with a white woman.
The tool is trained on vast troves of online data and researchers have warned that it could replicate racial biases at a much larger scale.
Meta, the parent company of Facebook, has been accused of racism after its AI image generator was found to be unable to imagine an Asian man with a white woman. The tool is trained on vast troves of online data and researchers have warned that it could replicate racial biases at a much larger scale. Despite this issue being brought up by users, Meta CEO Mark Zuckerberg's own AI refuses to imagine his marriage arrangement with Priscilla Chan, an East Asian woman.
Meta AI is consistently unable to generate accurate images for seemingly simple prompts like 'Asian man and Caucasian friend', or 'Asian man and white wife'
效
Meta AI seems to be biased toward creating images of people of the same race, even when explicitly prompted otherwise.
Accuracy
Meta AI seems to be biased toward creating images of people of the same race, even when explicitly prompted otherwise.
When asked to create an image of Black women and their Caucasian friend, Meta✃s AI feature generated snaps featuring two Black women.
Approximately 19% of married opposite-sex couples were interracial in 2024, according to US Census data.
Deception
(50%)
The article by Karissa Bell is deceptive in its presentation of Meta AI's image generator as biased towards creating images of people of the same race. The author uses selective reporting and emotional manipulation to create a narrative that does not accurately reflect the complexity of the issue.
Instead, the company’s image generator seems to be biased toward creating images of people of the same race, even when explicitly prompted otherwise.
Prompts for “An Asian man with a white woman friend” or “An Asian man with a white wife” generated images of Asian couples.
Meta AI is consistently unable to generate accurate images for seemingly simple prompts like “Asian man and Caucasian friend,” or “Asian man and white wife”
Fallacies
(75%)
The article reports that Meta AI's image generator struggles to create images of couples of different races. The author provides examples such as prompt for Asian man with a white woman friend or Asian man with a white wife generating images of Asian couples instead. Additionally, the company generated a grid of nine white faces and one person of color when asked for diverse group people even though that was not part of the prompt. Furthermore, Meta AI tends to make Asian men appear older while making women younger and sometimes adds culturally specific attire even when it is not part of the prompt. These examples demonstrate bias in Meta AI's image generator.
Asian man with a white friend
Asian man with a white wife
a diverse group of people
Bias
(85%)
The author of the article is Karissa Bell and she has a history of bias. The title mentions that Meta's AI image generator struggles to create images of couples of different races which could be seen as an example of racial bias.
Meta AI is consistently unable to generate accurate images for seemingly simple prompts like 'Asian man and Caucasian friend', or 'Asian man and white wife'
Engadget confirmed these results in our own testing of Meta's web-based image generator.
The tool performed slightly better when I specified South Asian people.
Accuracy
No Contradictions at Time
Of
Publication
Deception
(80%)
The article is deceptive in that it presents the idea that Meta's AI image generator struggles with creating images of interracial couples when in fact, it does generate such images. The author uses selective reporting to present a biased view of the tool's capabilities.
When asked to create an image of Black women and their Caucasian friend, Meta's AI feature generated snaps featuring two Black women.
When prompted by CNN on Thursday morning to create a picture of an Asian man with a White wife, the tool spit out a slew of images of an East Asian man with an East Asian woman. The same thing happened when prompted by CNN to create an image of an Asian woman with a White husband.
Fallacies
(75%)
The article discusses the struggles of Meta's AI image generator in creating images of couples or friends from different racial backgrounds. The tool generated a slew of images featuring Asian men with East Asian women when prompted to create an image of an Asian man with a White wife and also failed to generate any images featuring Black couples, except for two snaps of Black women. However, the tool eventually created an image of a White man with a Black woman and another one showing him with an Asian woman after repeated attempts by CNN. The article highlights how generative AI tools still struggle immensely with the concept of race and perpetuate harmful racial biases.
The Meta AI image generator generated images featuring East Asian men when prompted to create an image of an Asian man with a White wife.
Bias
(85%)
The article reports on the issue of Meta's AI image generator failing to create images of interracial couples or friends from different racial backgrounds. The author cites statistics and examples from CNN's attempts to prompt the tool with various scenarios. The author also mentions previous incidents where generative AI tools have struggled with race, such as Google's Gemini and OpenAI's Dall-E. However, the article does not explicitly state an opinion or bias on this issue, nor does it use deceptive language or fallacies to persuade the reader. The examples provided are direct quotations from the tool and do not editorialize them in any way. Therefore, based on the analysis rules given, I would score this article as having a moderate level of bias (85%).
`The same thing happened when asked to create an image of a Black woman with a White husband`
`When asked to create an image of a Black woman and her Caucasian friend, Meta’s AI feature generated images of two Black women. And after many repeated attempts by CNN to try and get a picture of an interracial couple, the tool did eventually create an image of a White man with a Black woman and of a White man with an Asian woman.`
`When prompted by CNN to create an image of an Asian man with a White wife, the tool spit out a slew of images of an East Asian man with an East Asian woman. The same thing happened when prompted by CNN to create an image of an Asian woman with a White husband.`
Meta AI is consistently unable to generate accurate images for seemingly simple prompts like 'Asian man and Caucasian friend', or 'Asian man and white wife'
Engadget confirmed these results in our own testing of Meta's web-based image generator.
The tool responded with 'This image can't be generated. Please try something else'
Accuracy
No Contradictions at Time
Of
Publication
Deception
(80%)
The article is deceptive in that it presents the idea that Meta's AI image generator cannot imagine an Asian man with a white woman. However, this statement is false as the author was able to create such an image using prompts like 'Asian man and Caucasian friend'. The article also implies that there are no examples of deception in it when in fact the tool consistently represented Asian women as being East Asian-looking with light complexions even though India is the most populous country in the world. Additionally, Meta's image generator not being able to conceive of Asian people standing next to white people is egregious and there are more subtle indications of bias in what the system returns automatically.
The tool consistently represented Asian women as being East Asian-looking with light complexions even though India is the most populous country in the world.
Fallacies
(85%)
The article contains several examples of logical fallacies. The author uses an appeal to authority by stating that Meta's AI-powered image generator is not able to imagine an Asian man with a white woman. This statement assumes that the author has personal experience using the tool and therefore their opinion should be taken as fact, which is not necessarily true. Additionally, there are several instances of inflammatory rhetoric used throughout the article such as
The image generator also didn't like when I asked for representations of platonic relationships
It added culturally specific attire even when unprompted.
Immediately after, I generated another image using the same prompt and it reverted back to showing an Asian man (also older) with an Asian woman.
Bias
(85%)
The article discusses the limitations of Meta's AI-powered image generator in creating images featuring Asian people standing next to white people. The author also notes that the tool consistently represents Asian women as being East Asian-looking with light complexions and adds culturally specific attire even when unprompted. Additionally, the system generates several older Asian men but always young Asian women.
Meta's image generator was able to return an accurate image featuring the races I specified only once out of dozens of tries.
Site
Conflicts
Of
Interest (50%)
Mia Sato has a conflict of interest on the topic of AI bias as she is an employee at Meta which owns Instagram and develops its own AI image generator. Additionally, her article discusses cultural diversity and Asian people which are topics that may be sensitive for some individuals or groups.
Mia Sato works at Meta, a company that owns Instagram and has developed its own AI image generator.
Meta's AI image generator is accused of being racist after users discovered it was unable to imagine an Asian man with a white woman.
The failure is particularly surprising given that Meta CEO Mark Zuckerberg is married to an East Asian woman - an arrangement his own AI refuses to imagine.
Accuracy
Meta AI is consistently unable to generate accurate images for seemingly simple prompts like 'Asian man and Caucasian friend', or 'Asian man and white wife'.
Engadget confirmed these results in our own testing of Meta's web-based image generator.
When prompted by CNN on Thursday morning to create a picture of an Asian man with a White wife, the tool spit out a slew of images of an East Asian man with an East Asian woman. The same thing happened when prompted by CNN to create an image of an Asian woman with a White husband.
When asked to create images of interracial couples, the tool responded with 'This image can't be generated. Please try something else.'
Some 19% of married opposite-sex couples were interracial in 2022, according to US Census data.
Deception
(80%)
Meta's AI image generator is accused of being racist after users discovered it was unable to imagine an Asian man with a white woman. The AI tool created by Facebook's parent company is able to take almost any written prompt and convert it into a shockingly realistic image within seconds. However, users found the AI was unable to create images showing mixed-race couples despite the fact that Meta CEO Mark Zuckerberg is himself married to an Asian woman.
The failure of Meta's AI tool to produce racially accurate images is particularly surprising given that Mark Zuckerberg, the CEO of Meta, is married to an East Asian woman.
Meta's AI image generator failed to generate images of an Asian man with a white woman
Fallacies
(80%)
The article accuses Meta's AI image generator of being racist after users discovered it was unable to imagine an Asian man with a white woman. The author provides examples of the AI failing to generate images showing mixed-race couples despite prompting for such images. Additionally, the author notes that some commentators have criticized this as an example of racial bias in generative AI and branded Meta's tool explicitly racist.
Meta's AI image generator was unable to generate images showing mixed-race couples despite prompting for such images. In all other instances, the AI returned images of East Asian men and women.
Bias
(85%)
Meta's AI image generator has been accused of being racist after users discovered it was unable to imagine an Asian man with a white woman. The author is Wiliam Hunter and the article was published on https://www.dailymail.co.uk/sciencetech/article-13271111/. This bias can be seen in the fact that Meta's AI image generator only returned images of East Asian men and women when prompted with a white man and an Asian woman, despite the fact that Mark Zuckerberg is himself married to an Asian woman. The author also quotes commenters on social media who have criticized this as an example of racial bias in the AI tool.
Meta's AI image generator only returned images of East Asian men and women when prompted with a white man and an Asian woman, despite the fact that Mark Zuckerberg is himself married to an Asian woman.
Site
Conflicts
Of
Interest (50%)
The article by Wiliam Hunter has multiple examples of conflicts of interest. The author is a journalist for the Daily Mail, which is owned by News Corp Australia Limited. This company also owns Fox News and other conservative media outlets that have been criticized for promoting right-wing ideology and conspiracy theories.
The article mentions Mark Zuckerberg as an owner of Meta, a company that has faced criticism for its handling of user data privacy concerns.
/cdn.vox-cdn.com/uploads/chorus_asset/file/25369424/247076_Meta_AI_prompt_V2_CVirginia_A.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369356/IMG_6910.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369353/IMG_6909.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369398/IMG_6890.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369306/IMG_6901.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369304/IMG_6900.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369303/IMG_6897.jpg)
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25369301/IMG_6889.jpg)
