Tech companies collect all kinds of data, and Google has always been no exception. Google collects data from users' online activities, including searches for content, videos watched, items purchased, who they communicate with, and location data accessed via Android mobile devices. In addition, Google also takes mobile phone information from "publicly accessible sources" in some cases, such as if your name appears in the news, Google may index the article and share it with people who search for your name.
These data collections are the norm, but Google recently released an update to its privacy policy, and the changes are to the information obtained from publicly available sources described above. Previously, Google's policy stated that this data could be used to "help train Google's language models and build features such as Google Translate," but this policy was extended after the change. "We may collect information publicly available online or obtained from other publicly available sources to help train Google's AI models and build products and features such as Google Translate, Bard, and cloud AI capabilities," the new policy reads. ”
The power of many AI systems, including ChatGPT, depends on the amount of data they can collect. Bard is Google's counterpart to ChatGPT announced earlier this year. Like other AI, it hasn't been all plain sailing. According to an April report, several Google employees had urged the company not to roll out Bard because the information it provided in response to queries was "worse than useless," and that some employees actually called him a "phantom lie," meaning he couldn't control himself from lying.
While more data may make Bard smarter, collecting data for AI training remains legally controversial: OpenAI faces multiple lawsuits over the way it collects and uses data to train ChatGPT. While it may seem harmless to collect publicly available information, according to The Washington Post, AI models will collect everything from Wikipedia pages and news to personal social networks, which many already dispute.
In addition, too much information also means that some copyrighted information will be used to train the AI. American writers Mona Awad and Paul Tremblay recently filed a lawsuit against OpenAI, accusing ChatGPT of violating copyright law by using their work for training without permission.