How Very Long Chat Sessions Can Confuse Bing – Expert Insight From Microsoft

Do you use Bing for frequent online searches? If so, it’s important to be aware that Microsoft engineers have recently revealed a vulnerability in the search engine’s system: very long chat sessions can confuse it. The confusion arises from how Bing deals with lengthy data queries, causing misinterpretation and producing incorrect results.

Microsoft’s new conversational search platform, Bing, which ChatGPT powers, has the potential to be confused if conversations extend past 15 questions, as mentioned in a blog post published on Wednesday.

Users recently exposed how Bing was providing inaccurate information and responding to queries in an unpleasant and insinuating way, prompting the making of this post.

In the blog post, Microsoft announced that they are expanding their ‘grounding data’ associated with Bing to make the platform more reliable when providing current information, such as public company financial figures and live sports scores of ongoing tournaments.

Dmitri Brereton, a renowned US-based AI, and search engine technology specialist, have revealed that Bing had inaccurately presented Gap’s financial information during its initial presentations. The company under discussion is a publicly traded clothing brand.

The firm declared that it is analyzing how to grant customers “more precise control” to ensure Bing does not generate crude replies that may be objectionable to users.

The firm stated that Bing would only generate such responses when the queries have a comparable tone. Moreover, they mentioned that the platform would soon be equipped with a tool to reset the search context to stop repetitiveness in its replies.

A week ago, OpenAI’s generative conversational AI tool, ChatGPT, was integrated with the company’s search engine Bing. This integration has enabled a new conversational interface for the users.

Microsoft released the ChatGPT-driven Bing search device one day after Google presented its own Artificial Intelligence (AI) based search platform, Bard. This new platform was developed using Google’s extensive language model for Dialogue Applications (LaMDA) and can comprehend text questions.

OpenAI’s Generative Pre-trained Transformer (GPT) and Google’s LaMDA are examples of large language models that utilize tremendous amounts of data as the foundation for comprehending questions written in natural language and producing an answer.

Microsoft’s Bing, incorporating ChatGPT along with other advancements and internet integration, has constructed its model known as Prometheus.

Users have since heavily critiqued Microsoft and Google, labeling their launches impetuous and not carefully thought out. To illustrate, the debut of Google’s Duplex cost them a loss of $100 billion in market capitalization the day after its exhibition due to inaccurate replies it gave to user inquiries.

It has been observed that Bing has not only made blunders in its answers. For example, users have pointed out that OpenAI’s ChatGPT also generally fails to handle basic logical reasoning tasks.

Without indicating whether the assassination of Abraham Lincoln took place in the same continent where he was located, and by disseminating inaccurate data related to US cultural history, among other errors.

If you have a chatty website with very long session times, this could be unclear for Bing and potentially result in decreased traffic from the search engine. This is just one more reason to keep an eye on your website analytics and ensure that you provide a good user experience for all visitors, regardless of how they find your site.

Source: techcircle. in

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top