Google expands boundaries of AI training using public data

Ömer

July 4, 2023

Google has decided to push the boundaries of artificial intelligence training. According to the latest change to the company’s privacy policy, publicly available data will now be fully used to train language models. Tools such as Google Translate, Bard and Cloud AI will be trained using user data.

Google to use public data in AI training

Google has signalled that it will begin to expand its privacy policies. Public data training, which was previously applied to features developed for testing purposes, will now be applied to all language models. This means that all Google products, from Bard to Translate, will be powered by user data.

The move has raised eyebrows as it comes after growing concerns about privacy. OpenAI was recently investigated for using user data, including private information, to train GPT models.

23423342

Many sources are used in productive artificial intelligence training, from tweets to articles on websites. The new privacy policy shows that data from Google-owned platforms will be used directly to train language models.

In other words, a translation you make using Translate will be collected to train Google’s productive artificial intelligence. Previously, Twitter and Reddit took some steps to prevent this. Both platforms started offering APIs with user data for a fee.

Although Google’s move attracts attention, it does not bother to explicitly state this in its privacy policy. How the data will be processed remains unclear.