Google has just changed the wording of its privacy policy, and it’s an interesting tweak that’s been applied to include the AI technology the company works with.
As TechSpot reports, there is a section of the privacy policy where Google discusses how it collects information (about you) from publicly available sources, and to clarify this, there is a note that reads: “For example, we may collect information that is publicly available online or from public sources.” Others to help train Google’s AI models and create products and features, such as Google Translate, Bard, and Cloud AI.”
First of all, that paragraph read that publicly available information would be used to train Language Models and only mentioned Google Translate.
Therefore, this section has been expanded to show that training takes place with AI and cool models.
It’s an obvious change, and essentially means that anything you post online publicly can be picked up and used by Google’s Bard AI.
Analysis: So what about privacy, plagiarism, and other concerns?
We already knew that Google’s Bard, and indeed Microsoft’s Bing AI for that matter, are essentially giant data warehouses, extracting and grinding online content from all over the web to refine conclusions on every topic under the sun that might be questioned.
This change in Google’s privacy policy shows that its AI works this way, and seeing it with cold, hard text on the screen might make some people hold back and question this a bit more.
After all, Google has been putting out the Bard for a while now, has it been working this way for some time, and just decided to just update its policy? This in itself looks very malicious.
Don’t you want to use things you’ve posted online where others can see to train Google’s big AI machines? Well, hard. If it exists, it’s fair game, and if you want to spar with Google, good luck with that. Despite the obvious concerns about not just basic privacy issues, but plagiarism (if the AI response uses content written by others, Bard’s training picked it up) – where do any boundaries lie with the latter? Of course, it would be impractical (or indeed impossible) to monitor that anyway.
There are also broader issues around accuracy and misinformation when data is scraped from the web in a high volume fashion too of course.
Moreover, there are concerns recently expressed by platforms such as Reddit and Twitter, that Elon Musk appears to be taking a stance against “scraping people’s public Twitter data to build models for AI” with those frustrating restrictions just introduced (which they could be). Big win for Zuckerberg and Yarn, at last).
This is all a huge minefield, really, but the big tech groups that are making big strides with AI systems that scrap data with LLM (Large Language Model) are simply getting ahead, and all eyes are on their rivals and the race to establish themselves on top, it seems hardly thought of. How to implement some practical side of the equation.
