Bard was trained on ChatGPT data? Google's top scientists protested to no avail and left OpenAI

Google’s top scientists petitioned Pichai and Jeff Dean to no avail, and finally had to switch to OpenAI.

If you can’t beat it, then join, or if you can’t beat it, just use it?

Compared to ChatGPT, which has become a milestone in the technology field, the debut of Google’s competing product Bard does not seem to be impressive. For this technology giant, being able to keep up with the pace of Microsoft and OpenAI is now the top priority. Things to consider.

According to reports, tremendous pressure has forced the DeepMind and Google Brain teams to join forces and launch a new project called "Gemini."

What’s even more surprising is that Google Bard has recently been exposed to using ChatGPT data for training. A former Google artificial intelligence researcher publicly opposed the use of this data and chose to leave after protesting to no avail.

#BERT resigned in anger

According to The Information, the Google AI engineer named Jacob Devlin tried to warn Google not to use ChatGPT data. Because this would violate OpenAI's terms of service and make Bard's answers look too similar to OpenAI's chatbot.

The Information reported that Jacob Devlin told Google CEO Sundar Pichai and Google AI chief Jefff Dean and other executives that he believed the Bard team was using information from ShareGPT. ShareGPT is a platform where users can publish the content of their communications with ChatGPT.

ShareGPT can save the content of people communicating with ChatGPT. Using this "data set" to train your own model may be a way to quickly improve your capabilities.

Jacob Devlin is one of four authors of the 2018 Google BERT paper, which is about training machine learning models to improve search accuracy. To some extent, this research has become Google’s and part of the OpenAI language model.

# He was part of a wave of Google AI talent departures at the beginning of this year. Jacob Devlin left Google in January to join rival OpenAI.

Before joining Google, Jacob Devlin worked as a software engineer at Microsoft for more than 3 years. Joined Google in September 2017 and has been working at Google for more than 5 years.

Google stopped using data from ChatGPT after the engineer issued a warning, a source said. Perhaps, Google has given up on training the corresponding part.

"It seems that Google is not ready yet. They are scared by OpenAI."

Some people also asked Bard for confirmation. As an honest AI, It also admitted it outright.

Google has denied this matter. "Bard was not trained on any data from ShareGPT or ChatGPT," Chris Pappas said in a statement. But he also declined to comment on whether Google had used ChatGPT data to train Bard in the past.

Google finally teamed up with DeepMind to bet on the new chat robot project "Gemini"

Regardless of whether Google ends up using the data, it is interesting to hear that Google may team up with DeepMind. . That’s also thanks to OpenAI, which managed to overtake Google with its chatbot ChatGPT, forcing two AI research teams within Google parent company Alphabet to overcome years of fierce competition and start working together — something that would have been impossible in the past.

According to Information, two people familiar with the matter revealed that software engineers on the Google Brain team are working with DeepMind employees to develop software to compete with OpenAI. After Google first tried to compete with OpenAI's chatbot with Bard, the company launched the project, which it calls "Gemini" internally, in recent weeks.

The above is the detailed content of Bard was trained on ChatGPT data? Google's top scientists protested to no avail and left OpenAI. For more information, please follow other related articles on the PHP Chinese website!