Google’s new AI Overview product has been causing quite a stir on social media lately. Users have been sharing examples of the AI saying bizarre things, such as suggesting putting glue on pizza or eating rocks. This has led to a rush by Google to manually disable AI Overviews for specific searches as memes spread rapidly online. Despite the fact that Google has been testing AI Overviews for a year now and served over a billion queries, the messy rollout has left many questioning the quality of the output.
Google CEO Sundar Pichai has mentioned that the company has managed to reduce the cost of delivering AI answers by 80 percent through various optimizations. However, it seems that this cost-cutting may have come at the expense of quality. One AI founder, speaking anonymously, highlighted that Google’s reputation for cutting-edge technology has now been overshadowed by low-quality output that is becoming the subject of memes.
Google has defended its AI Overview product, stating that it largely outputs “high quality information” to users. A company spokesperson mentioned that many of the questionable examples circulating on social media were either uncommon queries or doctored. The company is taking swift action to remove problematic AI Overviews and is working on broader improvements to the system.
The Challenge of Achieving 100% Accuracy
AI expert Gary Marcus explained that many AI companies promise perfection, going from 80 percent accuracy to 100 percent. However, achieving that final 20 percent is incredibly challenging, as it requires human-like reasoning abilities. Marcus believes that this last 20 percent may be the most difficult aspect of AI development.
Marcus and Meta’s AI chief Yann LeCun both agree that the current large language models powering AI systems will not be the ones to create artificial general intelligence (AGI). The industry is pushing towards more advanced AI capabilities, but it seems that the focus on improving basic functionality is paramount at this stage.
Despite the rocky start, Google has big plans for its AI Overview feature. The company envisions capabilities such as multistep reasoning for complex queries, AI-organized result pages, and video search in Google Lens. While these ambitions are impressive, Google must first address the fundamental issues with the current implementation to restore its reputation.
Google’s AI Overview product has faced significant challenges in its rollout, leading to questionable outputs and memes circulating on social media. The company’s focus on optimizing costs may have inadvertently compromised the quality of the AI output. As Google works to address these issues and improve its AI systems, the industry as a whole continues to push towards more advanced AI capabilities. It is clear that achieving perfection in AI is a daunting task that requires ongoing refinement and innovation.