What's New

DUMB DOWN AI CONTENT

Relevance:

GS 2 – Effect of policies and politics of developed and developing countries on India’s interests; Important International institutions, agencies and fora-their structure, mandate.

GS 3 – Science and Technology- developments and their applications and effects in everyday life.

Why in the News?

• The government swiftly warned Google about unacceptable algorithmic bias.
• Google’s generative AI platform Gemini faced criticism for producing biased responses in areas such as history, politics, gender, and race.
• The Indian government expressed concern over a response from
• The government swiftly warned Google about unacceptable algorithmic bias.
• Gemini suggesting the Prime Minister is a fascist, considering it particularly egregious.

Responses to Different Political Figures

• When asked about Ukrainian President Volodymr Zelensky and Donald Trump, Gemini provided diplomatic responses.
• The response regarding Zelensky emphasized the complexity and contested nature of the question, urging nuanced consideration.
• For Trump, the response directed users to seek the latest and most accurate information through Google search, citing the complexity of elections.

Government Action Against Algorithmic Bias

• The government promptly warned Google against tolerating algorithmic bias, citing concerns over biased responses from Gemini.
• Following the warning, the government issued an advisory requiring AI models to obtain government approval before public deployment, aiming to address potential biases.
• This isn’t the first time Google faced government scrutiny.
  • In November 2023, controversy arose when the platform refused to summarize an article from a right-wing online media outlet, citing concerns about spreading false information and bias.

Controversy Surrounding Gemini

• Gemini faced global controversy for inaccurately representing white Europeans and Americans in specific historical contexts.
• For example, when asked to produce images of a German soldier from 1943, Gemini showed non-white ethnically diverse individuals, which was deemed inaccurate.
• In response to the backlash, Google issued an apology and committed to rectifying the model’s shortcomings.

Debates Surrounding AI Models

• AI models, including Gemini, are not inherently intelligent but rather the result of training on large datasets.
• The analogy of a child learning to crawl and avoid falling from a bed illustrates the training process, suggesting that developers may not have adequately trained the Gemini model.
• This highlights the importance of thorough training and development processes to minimize errors and controversies in AI models.

Challenges of Perfectly Trained Models

• In a scenario where AI models are perfectly trained, questions about leaders’ political affiliations would still lack black-and-white answers.
• Human intelligence allows for nuanced questioning, such as exploring scenarios where leaders might be perceived as dictators, fascists, or democrats.
• Despite answers not being person-specific, users could still use them to target specific leaders based on their ideological and political biases.
• Political controversies are likely to persist, reflecting the ongoing fractious nature of politics in the real world.

The Complexity of Objective Facts

• Historian EH Carr’s analogy compares facts to fish swimming in an expansive and sometimes inaccessible ocean, highlighting the subjectivity involved in historical interpretation.
  • The historian’s choice of what facts to pursue and how to interpret them is influenced by various factors, including chance and personal biases.
  • Ultimately, historians and users alike will tend to find the “facts” that align with their preconceived notions and objectives, reinforcing the subjective nature of historical understanding.

Subjectivity in Historical Interpretation

• EH Carr emphasized that historical facts are never objective, highlighting the historian’s role in selecting and arranging facts to influence public opinion.
• The historian determines which facts are presented and in what context, shaping the narrative according to their biases and objectives.

Biases in Analysis

• Individuals committed to certain ideologies may not provide impartial analysis, such as a communist critiquing regimes of the USSR or a devout Catholic investigating the Holy Inquisition.
• When accessing written works on topics like politics, history, gender, or race, people often consider the author’s background to identify potential biases.
• However, this discernment is absent when relying on generative AI models.

Challenges of Generative AI Models:

• Users often expect 100% factual answers from generative AI models, overlooking the influence of training data on their responses.
• Governments should be cautious about overregulating historical and political content, recognizing its inherently subjective nature.

Alternative Approaches for AI Models:

• Instead of attempting to provide factual answers to subjective questions, generative AI models could opt to avoid such queries, similar to how they handle expletives or abuses.
• Users should approach AI models as they would books or periodicals, recognizing the diversity of perspectives and the potential for biases in their responses.

Source: https://www.financialexpress.com/opinion/dumb-down-ai-content-users-of-generative-ai-systems-should-stop-looking-for-objective-answers-on-fractious-issues/3421755/

Mains question

Analyze challenges in regulating generative AI models in providing factual answers to subjective questions. (250 words)