There’s no evidence that their generative tools are more ethical. Even if you believe everything they say, they are lying by omission. For example, for their t

There’s no evidence that their generative tools are more ethical. Even if you believe everything they say, they are lying by omission. For example, for their text to image technology, they never specify what their text language model is trained on - it’s almost certainly CLIP or T5, which is trained on plenty of not-expressly-licensed data. If they trained such a model from scratch - they don’t have enough image bureau data to make their own CLIP, even at 400m images, CLIP only performs well at the 4-7b image-caption pair scale - where’s the paper? It’s smoke and mirrors dude. There’s a certain personality type that is getting co-opted on social media like Hacker News to “mook” for Adobe. Something on the intersection of a certain obsessive personality and Dunning Kruger.