|
> The shift did not take place immediately. Within six months, traffic at smaller, less regulated sites had grown by 55%, and at larger sites by 10%, with point estimates implying that the traffic was entirely diverted to competing firms. This suggests that regulating only the largest platforms may push traffic to fringe sites and less controlled spaces. This rings true to me, especially in the recent context of AI adopters looking for uncensored alternatives. This frame of thinking can be applied not only to models, i.e. many move away from OpenAI/ChatGPT in search of less restricted models, as well as being applied to sites providing AI resources. Just the other day, CivitAI (the current leader for distributing custom checkpoints, LoRAs for image-centric models) announced it was taking a much more heavy-handed approach to moderation due to pressure from Mastercard/Visa. Its users are simply outraged, and many I think will be leaving in search of a safe haven for their models/gens going forward. |