Bing Chat, a Botched Tool?

GPT-4 at the Core, but Search Integration Falls Short

Bing Chat's Limitations and Unclear Language Model Usage

Bing Chat, a chatbot that reportedly utilizes GPT-4 as its foundation enhanced with web search capabilities, has come under scrutiny. Despite the integration of GPT-3.5 or 4 as the core language model, users have expressed concerns about the tool's performance.

One major issue is the 25-response limit every three hours. Additionally, the accuracy of Bing Chat's responses has been questioned, particularly when compared to GPT-4 on ChatGPT Plus. In one test, GPT-4 on ChatGPT Plus provided the correct answer to a question, while all other models, including GPT-3.5 and Bing Chat's creative mode, failed.

Furthermore, there is some confusion regarding the language model that Bing Chat actually employs. Some users have reported that Bing Chat claims to use "Creative uses GPT-3.5 Turbo Balanced uses GPT-3," which contradicts the notion of GPT-4 being at the core.