5 ms·
Unless the watermark is a giant “This is fake” diagonally across the center of the image, everything will get shared virally in the 3 second attention span of d
by mysterydip 3mo ago
Unless the watermark is a giant “This is fake” diagonally across the center of the image, everything will get shared virally in the 3 second attention span of doomscrolling without a second look.
- zb3 3mo agoWe already have open source image editing models, so how is this Google fault? This is UGC.. If I type something nefarious in Google search bar and then share that screenshot, is Google to blame for not even refusing rendering that text inside the input field?
- varun_ch 3mo agoWhat is an AI image generator inside Google Earth useful for if not generating believable but fake images that look like they’re from Google Earth?
- zb3 3mo agoThat's a separate issue - of course it's only so that someone at google can tick a box "I've shipped an AI feature" to get promoted. But we didn't gain any new ability to produce fake images, fake images can already be produced easily, even those that pretend they're from Google Earth. This was already possible, so I don't understand the concern here.. Are tech companies now responsible for everything the user does? That would end with total enslavement, don't go in that direction.
- varun_ch 3mo agotech companies are responsible for the products they ship, aren’t they? a company shouldn’t release something that could only be used for Bad Things
- zb3 3mo agoIt could be used for "good things" too, as you could see in their promo post. If I make a text editor, am I now responsible if someone uses it to write Mein Kampf? Back then people didn't complain that a notepad doesn't censor user input, but if we continue with your logic, we'll be surveiled constantly, I don't want that dystopian future to become reality.
- voakbasda 3mo agoSo you want to ban guns? Good and bad quickly become subjective, depending on how you view their use in different scenarios.
- boston_clone 3mo agoinstead of making potentially weak comparisons, could we consider why this feature was implemented in the first place? image editing models exist for good reason - I struggle to see a clear “why” for this product other than to cram generative content in yet another existing product.