I Honestly Humbly Silently Remotely Passively & In Remorse -> Slightly Disagree With Observations Made In GETTY IMAGES VS STABILITY AI Case!

 

Take this example & please do correct me if I‘m (mis)interpreting anything wrongly. And before that, please know that I Support OPEN Source, But NOT the Quote -> Open For Thee Not For Me!

Let’s say there’s Subject ABC who steals Raw Materials from WALMART and later builds a new product from it! WALMART files a Complaint about the robbery. Now, Subject ABC contends that, it didn’t STORE any of the Raw materials herein at his place, as it was all bought from beyond the borders; and rather, only used the mediums to Create a New One; thus, the charge of robbery cannot be maintained. Let us say, some part of those raw materials are also perceivable in the newly created End Work, YET, now those have been transcribed into a new product considering it as a whole product, and further, rather not been Stored, and thus the case of stealing/robbery never arise. How far you would support this argument given by Subject ABC, because firstly, there’s no storage of the raw materials, and as the materials which were brought from beyond the border have been converted from one form or another! In any Museum Heist, would you first check where it was stored or taken from, or, just cease those stolen goods?   

This analogy may be applied in the GETTY IMAGES VS STABILITY AI case; wherein, the observations made as what cannot be stored at a place, circumvents the idea as that to be considered as an infringed subject!

And if only the Storing of Data in the Training process of Stable Diffusion what ought to have only constituted the infringement, then that means, no medium would also ever become liable in case and in the act of the transfer of any impugned thing; as the medium itself ne’er possessed the Data, but, was merely transferring it from one end to another! Can this Theory be applicable in case of Money Laundering, Art heists etc. cases also as now, the medium can be anything and anyone that ne’er stored, rather assisted in the passing of the Data, but not passing off! But at the end of the day, whether this medium, or, the Stable Diffusion, they all made some gain! So how can we distinguish one Medium to Another!

 What all this means? The Generative AI companies may be now openly saying that the USER alone would be Liable; Yes YOU, if in case the AI Model creates an Output (which was trained on the infringed copies), and which creates an infringed copy, like what happened in the case of the Getty Images that bore the Watermark of it. As not all DATA bears Watermark, so if in case the USER creates an infringed DATA or Image from the Generative AI, then, the USER would be punished, and somehow, not the ones who created the AI Model by using the grey DATASETs from the Internet!

Let’s say GETTY’s case was weakened for the reason, as later it contended of the Stable Diffusion not storing the Getty’s images, BUT, my above arguments still hold, along with the submissions that some of the images that the Diffusion produced, had the Watermark cum Trademark of GETTY’s! But even if the latter part is removed, then too, what images the Stability AI used to train their models should’ve been inquired upon, at least under confidentiality! Seal the observations and findings as a secret, but at least, at the time of judging the matter, know the secrets of both sides, without disclosing it to the public at large.  

And if the training models are not storing the pixelated data, and rather only been trained on the patterns and outlines, yet, using the images itself for the commercial purposes, and not alone for R&D activities; would make that ones behind the Model, to some extent culpable. And the DATASETs made vide Scraping the Data from the Internet (in one of my earlier blogs, I distinguished between the Private and Public Datasets, which the AI Companies used to Confuse the public at large), how authentic they’re, and is equivalent to the Data which are available on the Dark Web.

Tomorrow, the policy makers would come and say, that, let these generative AI companies keep scrapping your DATA; and now it is your responsibility, as a USER / Non-USER to ask them whether you want to withdraw your name and DATA from their Training Models! Someone should explain at least this Logic to me, if it ever practically implemented.

Am I missing something. Please correct me, so that I can correct my observations! Thanks! 😊

© Pranav Chaturvedi