Google’s revenue from searches versus losses from LLM chatbot queries

Loading data from Talking to Chatbots Dataset reddgr/talking-to-chatbots-chats

😎 If you accept a different angle, it’s much simpler than what most people think (not inferring causality, trying to explain anything, or giving any opinion, just stating a very simple fact):

a) Google makes money for every Google Search (including those about factual political stuff, those about controversial political stuff, those about buying all sorts of stuff like guns or surgical abortions, those about watching all kinds of stuff like conspiracy theories or videos… They make money. Guaranteed, they do. North of 200 billion US dollars a year: https://www.google.com/search?q=google+advertising+revenue+2023)

b) For every query a user makes to their LLM chatbot, Google does not only not make any money, but they actually spend. Guaranteed, they lose money with this”
🤖 Your text is clear and presents a strong argument with minimal issues. Here is a refined version:

“If we consider a different perspective, it becomes much simpler than many might assume (this is not about inferring causality, attempting to explain phenomena, or expressing opinions, but merely highlighting a straightforward fact):

a) Google earns revenue from every Google Search, including those related to factual political matters, controversial political issues, purchasing various items like guns or surgical abortions, and viewing all sorts of content like conspiracy theories or videos. They indeed make money, with revenues exceeding 200 billion US dollars annually (https://www.google.com/search?q=google+advertising+revenue+2023).

b) For every query made to their LLM chatbot, Google not only fails to earn any money but also incurs expenses. It is certain that they lose money with this.”

Changes made:
– Simplified and clarified the introduction to focus on presenting the facts straightforwardly.
– Adjusted the structure for readability and coherence.
– Ensured the direct address to Google’s revenue and the comparison with their LLM chatbot’s financial impact was clear and succinct.

Would you like further details on these adjustments?
😎 “This is a general answer to people who might understandably be confused or annoyed by this:”
🤖 Your introductory sentence is clear, but a slight adjustment could enhance its flow and clarity:

“This response is for those who may understandably feel confused or annoyed by this situation:”

Changes made:
– Specified that the following content is a response, which prepares the reader for the information to come.
– Streamlined wording for smoother flow and readability.

Would you like more adjustments or explanations?

Hugging Face Dataset Metrics

All the conversation prompts, responses, and metrics are available to download and explore on Hugging Face dataset reddgr/talking-to-chatbots-chats: