Google co-founder Sergey Brin acknowledged the mishaps surrounding the troubled launch of the company’s artificial intelligence (AI) model, Gemini.
In a rare public statement, Google co-founder Sergey Brin acknowledged the mishaps surrounding the troubled launch of the company’s artificial intelligence (AI) model, Gemini.
The model’s image generation tool sparked controversy by depicting historical figures in inappropriate contexts. Despite Google’s intention to mitigate bias in AI outputs, Gemini’s performance fell short, drawing criticism from various quarters, including CEO Sundar Pichai.
The genesis of AI model Gemini’s issues lies in its design, which pairs a text-generating “large language model” (LLM) with an image generator. Users inadvertently uncovered instructions guiding the model’s behavior, revealing attempts to ensure diverse representation but ultimately resulting in problematic outcomes.
However, the root cause of Gemini’s failures remains elusive. Brin admitted confusion, stating that the model’s left-leaning tendencies were unintentional. Google’s head of search, Prabhakar Raghavan, outlined two key flaws: inadequate tuning to discern appropriate cases for diversity and an overly cautious approach leading to erroneous responses.
Experts speculate that Google rushed Gemini’s release to compete with OpenAI’s successful models, neglecting thorough testing. Dame Wendy Hall highlights the importance of prioritizing immediate challenges like combating deepfakes over existential concerns.
Andrew Rogoyski emphasizes the unrealistic expectations placed on generative AI models, which must balance creativity with factual accuracy and societal norms. With generative AI still in its infancy, such missteps are to be expected.
The fallout from Gemini’s blunders has raised questions about Sundar Pichai’s leadership and Google’s AI investments. While Pichai’s position may not be immediately threatened, investor patience is wearing thin in the competitive AI landscape.
Hall warns that further setbacks with generative AI models are likely as the technology continues to evolve. Despite the embarrassment for companies like Google, these experiences are essential for learning and improving AI development and usage.
American businessman Sergey Mikhailovich Brin is well known for co-founding Google with Larry Page. Prior to leaving the position on December 3, 2019, Brin served as president of Alphabet Inc., the business that owns Google.