The case of NSFW AI calls for careful evaluation because as we move further into the onset phase of artificial intelligence, transparency is an issue that cannot be neglected. Interpretabilty in AI Systems: A Criterion that can be quantified by how well the Model documents, explains and therefore serves as a reasoner for its decision making Process The AI Now Institute report said nearly 20% of all the websites had a less-than-ideal level of transparency by surface, but examined in more detail how user friendly and transparent they were.
This is often subsumed under the AI industry term black boxes, referring to systems that make decisions but are not designed to show or explain how and why they arrived at those decision. This opacity is even more worrying for NSFW AI, where the generated content can be far from innocuous. If, say, a NSFW AI creates content that breaks community guidelines (or even the law), without transparency it can become difficult to find out how and why this happned.
The earlier historical examples are further proof of why we need transparency in AI. Even in 2018, Facebook was discovered promoting biased ads due to its AI algorithms; yet it talked about removing bias from their systems. The fact that when applied to areas such as NSFW content the lack of transparency catalyzed huge societal consequences, happens in few incidents.
When we ask the question whether NSFW AI transparency issues matter, looking at ethical questions like this provide some stark clarity. The IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems underscores the importance of accountability in AI through transparency to help allay trust (and privacy) concerns. One downside of the opaqueness is that there is no way to verify whether NSFW AI stays in ethical line or respects user rights.
Furthermore, concerns about transparency are not mere theoretical problems; they have real-world ramifications. This could lead to serious risks of legal and regulatory complications, like what happened (in part) with the General Data Protection Regulation (GDPR) in Europe — which gives people a right-to-explanation when it exposes how AI is using their data. If indeed in breach of these regulations, NSFW AI systems that are not completely transparent could incur significant fines; perhaps as high as 4% of the total worldwide annual turnover for a company.
NSFW AI transparency challenges are closely tied to trust, ethics and legality -- for when developers or users cheat in their ability (or willingness) to develop more transparent models this issue as a technical challenge goes but it would be addressed by everyone developing NSFW AIs. See you all in future for more exploration, if you are interested into this please visit nsfw ai.