What does the watermark really give you?

It gives a false sense that you can tell what’s AI and what’s not. Especially when anything created malicously is likely going to remove that watermark anyway. Pandoras box is already open on those abilities and there’s no putting the lid back.

And, even in the case of non-maliciously generated work, if you suspect that something is AI, but it doesnt have a watermark, do you start investigations into how a video/image/story(text) was created? Doesn’t that mean that any artist or author is now going to need to prove their innocence just because someone suspects that their work had some form of AI involved in the process at some point?

It’s bad enough that they have to worry about those accusations from average people to begin with, but now you’re just giving ammunition for anyone (or any corporation) to drag them through the legal system based on what “appears” to be AI generated.

Edit: typo

Looks like it’s open source: https://github.com/google-ai-edge/gallery

Exactly this. We need a shift in how teaching/education is handled. It’s similar to when calculators became common. You can use a calculator to get the right answer, but relying on it too much, or using it incorrectly will hamper your progress.

Teachers might have to start pointing out best practices for using LLMs. Ex: Don’t ask the LLM to write the essay for you, instead, write as much as you can, then use LLMs to evaluate what you have written. Ask it to poke holes in your arguments, or make suggestions on how it can be improved.

At universities you should have access to something like a “writing center” where tutors can help you with exactly that. Although those tutors would be ideal (and better than an LLM), sometimes there’s a wait to get access to a tutor, or they’re only open for certain times.

An LLM can definitely be a useful tool in the writing process, but the students of today need to learn how to use them properly as well as how to evaluate whether an LLM response is useful or even if it should be trusted.

Teachers will need to rely on in-class only work/tests/activities to help them actually evaluate a student’s progress instead of relying on homework to do that.

Ikidd updated their comment, it’s only a 7B model.

Looks like they’re finally cleaning up a bunch of junk.

In July 2024, Google announced it would raise the minimum quality requirements for apps, which may have impacted the number of available Play Store app listings.

Instead of only banning broken apps that crashed, wouldn’t install, or run properly, the company said it would begin banning apps that demonstrated “limited functionality and content.” That included static apps without app-specific features, such as text-only apps or PDF file apps. It also included apps that provided little content, like those that only offered a single wallpaper. Additionally, Google banned apps that were designed to do nothing or have no function, which may have been tests or other abandoned developer efforts.