Most Search engines these days are AI-powered and provide an overview of the results. Google, the most used search engine on the internet, is no different. But something important has come to light recently.
Google’s search overview only has a 90% accuracy rating. This may be higher than it sounds, and it is when compared with its accuracy results last year from October, which was 85%. But the problem still stands that the search engine is working with around 10% error margin.
Trillions of searches are performed every year on the Google search engine, and if we consider the 10% of results, it is still returning almost over 50 million wrong search results. According to The New York Times, every one in ten search results on Google could be wrong.
Google claims the research has used flawed benchmark tools
Oumi, which is an open source AI platform popularized for developers to evaluate their AI models, released a research analysis based on around 4,326 simple QA Google searches. The results proved that Google’s Gemini was gradually improving in accuracy.
But a large portion of the results did not match up with the linked sources, according to Oumi. It is a complicated matter because the results may have been factually correct, but the links provided were not proper.
Google actually came out to refute the claims made in the original article, stating that Oumi is a flawed tool for benchmarking, and it does not reflect real-time results. But the internet is a place that is not satisfied with the words from a spokesperson.
Users online express mixed reactions to the research
User online came out to discuss the NY Times research, Gerrit De Vynck of The Washington Post, made comparisons with Chat GPT and how an error of this height would never have been tolerated in the pre-AI dependence era.
Others came out to criticize acceptance over the 10% error margin, citing real-world analogies. One would not accept if their car kept breaking down every 10% of the time, a user said. But it’s also surprising to see more than a dozen users coming out of the woodwork to let this slide for the moment.
It has been proven time and time again that AI is here to stay and will integrate into our future. Whether people accept the results or not, Google will keep working on improving the accuracy of its Gemini overviews. It’s up to the common users to decide what to make of this new research and put their trust in the results or not.
