Last Wednesday (29) Google responded to criticism by denying that its new text generation tool, bardwas trained on data from ChatGPTlanguage model of artificial intelligence developed by OpenAI.
Google’s rejection was in response to an article published by The Information that claimed Bard was trained on a modified version of ChatGPT called ShareGPTwhich was created to allow third party companies to train their own language models.
News
Microsoft CEO Bing says Google Bard is “way behind…
Mikhail Parakhin says he has learned to never underestimate Google
In a statement sent to The Verge, a Google spokesperson said that Bard was trained on a varied set of text data and that ShareGPT was not used in training. Additionally, Google stated that Bard was developed using a “from the ground up” approach based on years of research into language models and new learning methods.
Bard was launched by Google earlier this month as an advanced text generation tool capable of creating content in a variety of styles and formats. Since its launch, Bard has been the subject of intense discussion in the AI community, with many experts wondering how the tool was trained and what its implications are for the future of text generation.
Despite Google’s rebuttal, the Bard controversy is likely to continue, with many AI experts arguing that it’s hard to check whether a language model has been trained on specific data or not. Meanwhile, Bard remains an important tool for creating lyrics and should be used with care and responsibility.
Google ChatGPT recognized the article as plagiarism
The Google Bard chatbot has entered beta and promises to answer questions quickly and accurately. However, a chatbot may rely on specialized sources of information without properly referencing them. An investigation by technology portal Tom’s Hardware revealed that the chatbot borrowed part of an article published on the site.
The question arose which processor is faster: Intel Core i9-13900K or AMD Ryzen 9 7950X3D. Bard’s answer was copied directly from the Tom’s Hardware article, but the chatbot didn’t credit the source and referred to the data as “in our tests”, implying that Google ran the tests.
When asked where the tests came from, Bard admitted that the results came from Tom’s Hardware, and when asked if he plagiarized, he answered yes. A chatbot sentence is a retelling of a specific sentence from the original article.
Continuation after commercial
Mundo Conectado Deal Center: Selection of Discounts and Lowest Prices
Best deals on electronics, cell phones, TVs, soundbars, drones and more
Source: Information, The Verge
…..