The official denies that Google's chat AI 'Bard' is trained with ChatGPT data



It was pointed out that the chat AI `` Bard '', which Google announced as a rival of ChatGPT, was trained using data from ChatGPT and ShareGPT , which records the content of user conversations. However, Google denies this allegation.

Alphabet's Google and DeepMind Pause Grudges, Join Forces to Chase OpenAI — The Information
https://www.theinformation.com/articles/alphabets-google-and-deepmind-pause-grudges-join-forces-to-chase-openai



Google denies Bard was trained with ChatGPT data - The Verge
https://www.theverge.com/2023/3/29/23662621/google-bard-chatgpt-sharegpt-training-denies

Google denies using ChatGPT data to train its Bard AI
https://bgr.com/tech/google-denies-using-chatgpt-data-to-train-its-bard-ai/

Google Rejects Allegations of Training Its Bard Chatbot on Data from ChatGPT
https://winbuzzer.com/2023/03/30/google-rejects-allegations-of-training-its-bard-chatbot-on-data-from-chatgpt-xcxwbn/

Bard is a conversational AI released by Google as a rival chat AI for ChatGPT. You can find out what kind of conversation you can actually have by checking the following article.

Google has launched ChatGPT's rival 'Bard' to the public, and what kind of answer will you get when you actually ask? -GIGAZINE



Alphabet, Google's parent company, seems to have two groups, DeepMind, an AI laboratory under its umbrella, and Brain AI, Google's AI development department, jointly developing Bard. The development team seems to have used ChatGPT's conversation data to speed up Bard's development.

One AI researcher who was involved in the development of Bard told The Information, 'Google relies on ShareGPT for data to train Bard.' ShareGPT is a site that records user interactions with ChatGPT.

According to reports, it was Jacob Devlin, who left Google to join OpenAI, who warned Google executives against using ChatGPT data. Devlin not only said that Bard's answer was too similar to ChatGPT, but also pointed out that using ChatGPT conversations for Bard's training may violate OpenAI's terms of service. On the other hand, on the overseas bulletin board Hacker News , there was also a voice pointing out that there is no legal problem in using the data, saying, ``The output of ChatGPT is not copyrighted by OpenAI.''




According to another source, Google seems to have stopped using ChatGPT data after receiving Devlin's warning. In addition, it is reported that the model trained using ChatGPT data has also been deleted.

When The Verge asked Google, ``Does Bard use ChatGPT data for training?'' There is no fact that there is.” However, ``Unfortunately, this is all I can share,'' he said, avoiding answering whether Bard was trained with ChatGPT data in the past.

In addition, it seems that ShareGPT's 'Explore' page contained more than 112,000 conversations, but this page was deleted in the fourth week of March 2023. Therefore, it is pointed out that it was deleted 'for use in Bard'.




Immediately after Google announced Bard, there was a critical voice from within the company that was too sudden to announce.

Google employees are dissatisfied with the too hasty announcement of 'Bard' - GIGAZINE



in Software, Posted by logu_ii