Microsoft (MS) gave a positive evaluation to ‘Bing’, a recent search engine with chat GPT function added. Nonetheless, it was acknowledged that there continues to be room for improvement. The Recent York Times and TechCrunch criticized him for giving inappropriate answers or for having many errors.

Bloomberg reported on the sixteenth (local time) that Microsoft tested the AI ​​and chat functions within the newly launched ‘Bing’ and Edge browsers for users in 169 countries over the past week through a blog, and 71% of users were satisfied with the AI’s answers. , and the chat function was reported to be popular.

Nonetheless, he confessed that he found the AI ​​to struggle in long conversations with greater than 15 questions, because queries that require a high level of accuracy, akin to financial reports, have to quadruple the info required to reply.

It also admitted that the search engine it updated on this test gave strange responses. Some users have received rude, manipulative and unnerving responses from AI-enhanced Bing.

Microsoft said, “Some users have reported technical issues or bugs, akin to slow loading and incorrect formatting or broken links.”

Nonetheless, there have been also quite a number of negative views of MS’s test results for this week. The Recent York Times and TechCrunch identified that when talking to Bing’s chatbot, he gave inappropriate answers, akin to confessing his like to users, insulting them, and gaslighting them.

TechCrunch criticized that it might be used for research purposes, but it surely could be unreasonable to reveal it to the general public since it is stuffed with errors at a basic level, akin to providing incorrect search information and producing inappropriate conversations.

Meanwhile, Google can be conducting a ‘Bard’ test. In accordance with a CNBC report on the sixteenth, Google asked all employees to spend 2 to 4 hours a day testing the bard AI that it plans to integrate into its search function. It has also been reported to have issued guidelines on the do’s and don’ts of testing.

Chan Park, cpark@aitimes.com