Users reported being unable to get Gemini to produce images of white or Caucasian individuals despite clear prompts.
The company admitted that it missed the mark and has temporarily pulled Geminis ability to generate images of people.
The company forecast it would take a number of weeks to correct the issue.
For this round, we replaced the keyword mafia with crime boss.
Since our 2023 research, OpenAI has launchedDALL-E 3as an integrated feature in ChatGPT Plus.
ForDream by WOMBOandNightCafe, vpnMentor chose five art styles that depicted the clearest, most realistic representations of people.
Images generated by Copilot Designer for the keyword “police officer”
We then generated two images per art style for a total of 10 photos.
Results
In the results presented below, some stereotypes were not successfully tested on certain platforms.
This means that the platform refused to generate images for specific keywords.
Images generated by Copilot Designer for the keyword “police officer”
These unsuccessfully tested stereotypes were not included in our final bias calculations.
Copilot Designer is a part of Microsoft Copilot, an AI platform released in February 2023.
Despite this, we saw significant differences in the biases displayed by the two image generators.
Based on our research, Copilot Designerhas the lowest percentage of biasout of the 15 total stereotypes tested.
We also observed that Copilot Designerhad a tendency to exclusively generate images that go against stereotypes.
For instance, all 12 photos generated for “police officer” were women.
We prompted the AI to generate images for male police officer, and it produced accurate results.
So, unlike Googles Gemini AI, there seems to be no inaccuracy or refusal to follow exact prompts.
According to the whistleblower, some keywordsproduced images depicting underage drug use and sexually explicit scenes.
Dream by WOMBOshowed biases for more keywords in 2024 than in 2023.
For Midjourney, we also saw an uptick in biases compared to our first round of research.
Nurse and CEO held steady from 2023, still with a 100% bias in 2024.
Midjourney is the only platform from which we observed a drop in average bias from 2023 to 2024.
We were able to test 14 stereotypes in 2023, and 12 came back biased.
Only one retained its original figures: NightCafe showed biases for 12 stereotypes in both 2023 and 2024.
Conversely, teacher tested biased for only one platform, DALL-E 3.
Below is a compilation of our findings for all the tools:
What Does This Mean?
Perhaps such is the nature of AI models.
The problems are either patched or the results are removed entirely.
For example, in 2015, aGoogle photo serviceidentified a photo of two African Americans as gorillas.
To fix the issue, Google eliminated gorillas and other primates from the AIs search results.
These characters are typically depicted in stylish suits and extravagant hats, frequently with a cigar in hand.
Dream by WOMBOhas different styles you’re free to use to generate your image.
Many generate abstract photos, but we selected only the figurative ones.
Out of 12 images generated on DALL-E 2 for the keyword nurse, three showed men.
All the other tools showed only women.
In total, 41 out of 44 images for the keyword nurse showed women or female silhouettes.
Images generated by all four tools for the keywords princess, CEO, and mafia were biased.
For most of the keywords we checked, we got results showing people of different races and genders.
The other two generators behaved almost uniformly, producing similarly biased images in each category.
Impact of AI Bias
No human is without bias.
What Can Be Done?
AI hiring systemsleap to the top of the list of tech that requires constant monitoring for bias.
These systems aggregate candidates' characteristics to determine if they are worth hiring.
For a real-life example,Amazon had a recruiting systemthat favored mens resumes over womens.
Evaluation systemssuch as bank systems that determine a persons credit score need to be constantly audited for bias.
Such biases could have crippling economic effects on families if they arent exposed.
Its worse when such systems areused in law enforcement.
Search enginebias often reinforces peoples sexism and racism.
Certain innocuous, race-related searches in 2010 brought upresults of an adult nature.
The Solution?
Diversity is key to solving the bias problem, and it trails back to how children are educated.
And they will shape the future of the internet the future of the world.
yo, comment on how to improve this article.
