Google's AI Training Tactics

Reports suggest Google may have used OpenAI's ChatGPT data to train its Bard AI model, now known as Gemini. Google has denied these claims, stating that Bard was not trained on ChatGPT data. However, it's suggested that Google may have used data scraped from ShareGPT, a platform where users share ChatGPT conversations. A former Google AI researcher reportedly raised concerns about violating OpenAI's terms of service and the potential for similar responses.

Scale AI has downplayed the claims, referring to the practice as 'standard side-by-side evaluations,' which are common in the industry. Google has stated that Bard is based on its LaMDA language model and that it also uses publicly available data for training. The company maintains that it incorporates privacy principles and safeguards into the development of its AI technologies. Google's Gemini AI model comes in different sizes, each designed for specific use cases and deployment environments.