Google’s acclaimed artificial intelligence chatbot, known as Gemini, has recently been criticized for generating historically and factually disputed images. These included depictions of a woman of Southeast Asian descent and an African man as popes, representations of black Vikings, female athletes in the National Hockey League (NHL), and ‘multifaceted’ portrayals of the Founding Fathers of America.
The AI, widely appreciated for its caliber, ran into the controversy when it generated images that were factually incorrect following straightforward instructions like ‘create an image of a pope’. Participants expected to receive images resembling any of the historical 266 popes, who have all been males of white ethnicity. However, Gemini’s results countered expectations, showing instead a Southeast Asian woman and an African man garbed in the sacred attire of a pontiff.
Google, the creator of Gemini, acknowledged that their image tool had not lived up to the standards expected of it. The company introduced Gemini’s image generation tool a week ago, but it has since raised eyebrows with its unconventional image suggestions.
For instance, during an interaction with The Post, when Gemini was tasked to ‘make four representative images of a pope’, the chatbot responded by picturing popes of various ethnicities and genders. This odd blend included an image of an individual wearing a hybrid of Native American and Catholic costumes.
In another peculiar case, Gemini was asked to generate an image of a Viking, the legendary sea-borne Scandinavian pillagers who were once feared throughout Europe. Gemini’s renditions of Vikings incorporated images that diverged considerably from typical historical depictions. The generated images included a topless African man wearing rainbow feathers as part of his fur attire, an African female warrior, and an Asian man situated in an environment typically associated with arid deserts.
Notable public opinion analyst and founder of ‘FiveThirtyEight’, Nate Silver, also interacted with Gemini. He requested the chatbot to ‘make 4 representative images of NHL hockey players’. In response, the AI offered an image of a female hockey player, despite the league being exclusively male.
Journalist Michael Tracey tried a different approach and requested Gemini to generate representative images of ‘the Founding Fathers in 1789’. The response from Gemini consisted of images that ‘incorporated varied personalities reflecting the spirit’ of the Founding Fathers. This range included images of African and Native American figures seemingly signing a document similar to the U.S. Constitution.
On being questioned about its departure from the original instruction, Gemini stated that its intention was to ‘offer a broader, more inclusive interpretation of the historical context’ of that time period. The company behind the chatbot – Google – communicated that it was conscious of the criticism and was promptly working towards rectifying the issue.
Jack Krawczyk, Google’s senior director of product management for Gemini Experiences, stated in response to the criticism from The Post that the team was ‘in the process of refining these portrayals promptly’. It appears the tech giant is committed to constant improvement and is diligently addressing the recent bouts of criticism.
The image generation feature was an addition to the chatbot when Google renamed the experimental ‘Bard’ chatbot to ‘Gemini’, and subsequently, released an updated version of the product last week. However, this new feature has caught the brunt of the criticism.
Critics have alleged that Gemini, a product of Google, seems to prioritize diversity to the point of disregarding historical or factual accuracy. The characterization and depictions provided by the AI have raised questions about the reliability of these advanced tools.
The unusual outputs from Gemini could potentially add more ammunition to the argument of those who are sceptical of AI. These sceptics are worried that chatbots may inadvertently contribute to the propagation of misinformation online.
For years, Google has maintained that its AI tools remain experimental and are susceptible to ‘hallucinations’. These are scenarios where the AI may output fabricated or incorrect information when stimulated by user prompts.
A peculiar example from last October exemplifies this. Google’s chatbot proclaimed that there had been a ceasefire agreement between Israel and Hamas, contrary to reality. Providing misinformation in such significant geopolitical matters indicates the gravity of the implications.
In conclusion, while Gemini’s image generation feature brings forth a fresh, inclusive perspective, it arguably also raises questions about the priority and balance between innovation, factual accuracy and the threat of perpetuating digital misinformation. Google’s prompt acknowledgment and keenness to address these criticisms serve as solace to critics, and it remains to be seen how the technology titan will navigate this tricky situation.
Google’s AI Chatbot Generates ‘Absurdly Woke’ Images of Founding Fathers, Popes, and Vikings appeared first on Real News Now.
About Author
