TL;DR
Google may be offering users new settings to control Gemini’s reactions. This “content filter” will presumably allow Gemini itself to control the extent to which it is censored. The Gemini API already has a “hazard category” setting that developers can access.
Over the past year, artificial intelligence has been introduced into virtually everything. When it comes to AI and Google, Gemini is the king. The company has been hard at work improving the model, announcing Gemini 2.0 earlier this month. Because of the powerful capabilities that AI provides, companies like Google spend a lot of time considering what protections they need to put in place for their AI systems. We’re currently exploring new ways to give users more control over the type of content Gemini produces.
APK teardowns help predict features that may be delivered to a service in the future based on the code in progress. However, such predicted features may not be publicly released.
Today we’re going to check out version 15.51.24.sa.arm64 beta of the Google app for Android. This isn’t public yet, and it’s not really public, but you can enable a new option in Gemini’s settings, and it looks like we’re preparing the ability to filter Gemini’s output by content. .
Assemble debugging / Android permissions
Selecting this option does nothing at this time. Simply navigate to a URL on Google’s Gemini site that is not yet live. So it’s a little hard to say with certainty how this setup will work, but we can make some inferences based on what we already know about Gemini.
Google already has some formal policies in place about what it’s allowed to do with Gemini, and by extension, some things about what kind of content it will try to get Gemini to avoid. We have set goals for In particular, developers now have access to fairly granular Gemini safety settings, allowing them to control many “hazard categories” such as hate speech, harassment, and sexually explicit content.
Given all the features currently available, it might be possible for Google to extend some of the same levels of control to end users of Gemini on Android and the web. That said, Google could go in a number of different directions with its implementation here, from the broad developer-level filters the API currently has to perhaps even more restrictive filters, all aimed at preventing Gemini. there is. Don’t share anything especially offensive.
At this time, it is difficult to say when this new option will take effect, but we will keep an eye on further developments.
comment
