At the Google I/O event, the company announced a new search feature called “AI Overviews.” Since its launch, the feature has become the talk of every conversation due to its odd and misleading responses. While people started to have doubts over Google’s accuracy and AI efforts, the company finally decided to respond to the growing concern and address users of its working and why it generated such questionable responses. Check out what Google said regarding AI Overviews mishappenings.
Google defends AI overviews
In the blog post, Google first highlights how the AI overviews are gaining traction and how people are appreciating its usage and capabilities. Google said, “people have higher satisfaction with their search results, and they’re asking longer, more complex questions that they know Google can now help with.” Later, the blog pointed out the odd overviews and why it is so important to address and clarify why it generates such misleading results. Before getting into the topic, Google briefly explained the workings of AI Overviews.
Also read: Google AI Overview generates misleading responses, tells users to eat rocks, glue
Google clearly mentioned that AI overviews do not generate results based on training data but it is powered by a customised language model which is integrated with core web ranking systems. It means that the feature picks information from relevant or high-quality results. Google said, “AI Overviews generally don’t “hallucinate” or make things up in the ways that other LLM products might.”
On the other hand, it also highlights that AI overviews generated odd responses due to “ misinterpreting queries, misinterpreting a nuance of language on the web, or not having a lot of great information available.”
Also read: Google Search AI overview: How to bypass them and directly see results
AI Overviews odd responses
In the blog, Google addresses the responses that went viral on social media such as “How many rocks should I eat,” or “using glue to get cheese to stick to pizza.” The company said that the feature is not capable of generating responses for “satirical content” due to an information gap. Therefore, the features link to websites which has such answers to queries or provide reliable first-hand knowledge.
Also read: Google password manager now gets family password-sharing feature- What is it and how it works
However, Google also acknowledged that AI Overviews misinterpret language on web pages and that they have integrated improvements for such queries.
AI Overviews improvements
The company mentioned that they have made a dozen technical improvements to avoid such mishappenings:
- Google added an improved detection mechanism for nonsensical queries and restricted satire and humour content.
- Google has limited the use of user-generated content (UGC) in responses that could potentially provide misleading advice.
- It has also added restrictions over queries where AI overviews are not helpful or could generate accurate or informative responses.
- For news and topics related to health, Google will restrict displaying AI Overviews altogether due to the dynamic events and nature.
One more thing! We are now on WhatsApp Channels! Follow us there so you never miss any updates from the world of technology. To follow the HT Tech channel on WhatsApp, click here to join now!
