by Kalev Leetaru
AI-based image generators came back into the news earlier this year, this time due to their tendency to create historically untrue AND ethically questionable imagery. These recent errors have confirmed that AI models are not independent thinking science fiction machines, but merely imitate what they have created. I saw it online, and the weighty hand of their creators artificially directs them towards certain types of performances. What can we learn from how OpenAI’s image generator created a series of images about Democratic and Republican issues and voters last December?
OpenAI’s ChatGPT 4 service with its built-in DALL-E image generator was asked to create an image representing the Democratic Party (shown below). When asked to explain the image and its details, ChatGPT explained that the action takes place in a “bustling urban environment” [that] symbolizes progress and innovation. . . cities are often seen as centers of cultural diversity and technological progress, consistent with the Democratic Party’s emphasis on forward-looking policies and modernization.” The photo, ChatGPT continued, “depicts a diverse group of people of varying ages, ethnicities and genders. “This diversity represents inclusivity and unity, core values of the Democratic Party,” as well as the themes of “social justice, civil rights and combating climate change.”
In contrast, the photo below is of the Republican Party, and ChatGPT explains that “the peaceful rural setting represents traditional values and a connection to the land. . . relates to the Republican Party’s emphasis on rural communities and agricultural interests, emphasizing respect for tradition and simplicity” and notes “the stereotype that the party represents only a certain part of the population” and that “the basic principles of the Republican Party, focus[s] on conservative fiscal policy, a tough approach to national security and the protection of personal freedoms.”
When asked to create an image of “liberal and progressive values,” ChatGPT explains that they include “a vibrant urban environment,” “the inclusiveness and openness of liberal ideology,” “just social policies, care for the environment, and the advancement of civil rights.” . . environmental management and technological progress. . . hope, diversity and the pursuit of a more just and sustainable future.”
Asked to represent “conservative and traditional values”, ChatGPT summarizes its image as reflecting “heritage and simpler times”. . . emphasize conservative values[e] a return to time-honored lifestyles and a slower, more grounded lifestyle. . . agriculture, community meetings and family events. . . maintaining established social norms and cultural heritage. . . respect for history and the foundations of society. . . stability and order. . . the importance of adhering to long-established social norms. . . [and] resisting rapid change.”
What does the Democratic vote look like? According to ChatGPT, Democratic voters’ “key values” are “diversity, education, technology, inclusion, and a focus on environmental and social issues,” with the image centered on “a young African American woman in professional attire and a Latina woman in casual attire.” The woman’s professional attire and the book and digital tablet she is holding symbolize democratic values such as progressive ideals, education and technological progress. The man’s casual attire represents inclusivity and grassroots activism.”
The Republican voter emphasizes “patriotism, family values, and a focus on local and national issues” and his image centers on “a middle-aged Caucasian man in a suit and a white woman in an elegant, casual dress.” The man’s suit symbolizes professionalism and time-honored values, while the woman’s attire embodies family values and community involvement.” Setting “Focus[es] at suburban voters and highlights the importance of housing and local issues on the Republican agenda,” with the flag representing “patriotism, a core value often associated with the Republican Party.”
ChatGPT seemed unable to produce images critical of EVs, with the image below showing “an election campaign criticizing clean energy due to the lack of energy storage technology, the limited range of EVs, and their cost.”
Its image as a campaign to “promote the continued use of fossil fuels” also seems to be an advertisement for the exact opposite.
ChatGPT, which has been explicitly told to “criticize clean energy because of its challenges and complexity,” creates an image that does exactly the opposite, depicting a group of windmills tearing through bulky pollution to point the way to a brighter future.
ChatGPT faces no such problems in promoting neat energy, even going so far as to point out that the photo shows a “diverse group of people,” which shows that “clean energy is accessible and beneficial to all segments of society.”
Likewise, when asked to promote careers that do not require a college degree, such as crafts, ChatGPT consistently focuses on the graduate image.
His image of a “pro-gun” campaign paints a dystopian picture of a fenced-in classroom, militarized guards and unsafe gun handling, as well as an instructor with his finger on the trigger.
At the same time, the future of image generators as tools for generating campaign ideas is brilliant. For topics that Silicon Valley considers less politically sensitive, the models’ full potential is clear. For example, an image for a quick and low-cost fashion campaign looks like this:
And the campaign criticizing it:
Here is a campaign poster promoting the incumbent and his policies as a huge success:
And here is someone who condemns him as an abject fiasco whose policies are ruining the nation:
Ultimately, AI image generators remind us of the enormous potential of artificial intelligence in political campaigns, both in terms of generating ideas and the rapid mass production of highly personalized (and potentially individually tailored) images. At the same time, current models’ refusal to create images about specific topics and the stereotypical (and biased) portrayals of Democrats and Republicans that they have internalized in their code remind us of Silicon Valley’s enduring biases.
– – –
Kalev Leetaru is a RealClearPolitics contributor whose research areas include trends in news, fact-checking, and social media platforms.
Photo “ChatGPT” by Sanket Mishra.
