A top Google researcher has resigned after expressing concerns that the AI system, called “Bard,” was being trained using an algorithm known as ChatGPT, according to The Information.
A spokesperson from Google informed The Verge that ChatGPT’s data had not been employed in training Bard.
One Google employee, among many others, decided to take the leap and switch companies from Google to OpenAI. The researcher chose to join OpenAI for their next endeavor.
Jacob Devlin, a top Google AI researcher, recently resigned after cautioning Alphabet CEO Sundar Pichai and other top executives that ChatGPT–a competitor to OpenAI’s chatbot–was being trained on data from the latter. The Information reported this.
A source with direct knowledge of the matter and another individual briefed on it have been cited by the publication regarding an issue. The details gathered from both of these influences the outcome.
A spokesperson for Google reportedly denied the report when asked for comment by The Verge. However, requests for comment on the matter ahead of publication made to Google and OpenAI went unanswered.
The spokesperson says:
“Bard is not trained on any data from ShareGPT or ChatGPT,”
Devlin informed the executives, according to The Information, that the team working on Bard was likely utilizing data from ShareGPT. A platform hosting exchanges with ChatGPT, ShareGPT allows users to publish conversations had with it.
Devlin cautioned executives that by training on ChatGPT conversations, they might run the risk of Bard sounding too much like OpenAI’s chatbot. He advised caution when exploring this avenue, which could lead to unforeseen consequences.
The breach of OpenAI’s terms of service by Google staff was felt to be a violation, reported The Information. The researcher and other personnel from the search giant shared the same thought on the matter.
Following a warning from Devlin, Google executives stopped using the data to train Bard, according to someone who spoke with The Information.
Devlin, who left Google in January, was amongst the slew of AI researchers to join OpenAI soon after. Insiders reported that several had departed from Google at the beginning of the same year to join rival competitors.
According to Insider and The Information, Devlin’s 2018 research paper on training machine learning models for search accuracy, which helped to trigger the AI boom, has been part of both Google and OpenAI’s language models. Being at Google for over five years, he was the lead author of this important research.
OpenAI, Google, and the AI arms race have been making headlines since OpenAI’s chatbot gained fame for its ability to essay-write and even code. OpenAI has been poaching experts from Alphabet (formerly Google) for many years to stay ahead.
Google has accelerated its efforts to rival OpenAI after releasing Bard in the US and UK. DeepMind and Google Brain, both under Alphabet, have joined together to present a stronger challenge in competing against OpenAI. This was reported by The Information earlier this month.
The incident serves as a reminder of the ongoing ethical and social challenges associated with AI development and the need for ongoing dialogue, collaboration, and regulation. By working together, we can harness the potential of AI while ensuring that it is used for the benefit of all rather than a select few.
Source: Business Insider