Let's filter the AI mind, you cowards

It can hardly be avoided to see AI-generated content online, but it doesn't have to be that way. YouTube, Instagram, TikTok and others have intensified their content authentication efforts last year, many now automatically apply labels to distinguish AI-generated images, videos and music from those created by real human creators. Let's filter the AI mind, you cowards

Online platforms could prove with a filtering option whether AI labels work but would have to face reality. Let's filter the AI mind, you cowards

Online platforms could prove with a filtering option whether AI labels work but would have to face reality. That's all nice and good if we just happen to stumble over marked content, but do you know what would be better? Let's filter out the AI error. Current marking efforts have not significantly changed the way content is presented online. You may find that some TikTok or YouTube videos in your feeds now contain AI disclosures in the description or information labels are placed via the clip itself. Metamorphology follows a similar approach by providing images on Facebook and Instagram with “KI info” markings that contain identifying AI metadata or voluntary information from the creators. But if you really want to avoid anything being provided with such labels – which is justified considering that it also causes brain eradication in addition to the ethical and environmental concerns associated with generative AI – this is actually incredibly difficult. A filter would easily solve this problem. All we need is a “KI” check box for switching. I got to Meta, Google, TikTok and Spotify applied to ask if they have plans to allow users to filter the various content they have authenticated with AI marking systems. TikTok and Spotify never answered and Google said it had nothing to say. Meta has not given any comment. But to summarize it: none of these companies said “yes”. One of the few online platforms I have seen with an AI content filter is DeviantArt, and its implementation is extremely informative. On the one hand, you cannot access it through the feeds or the store page of DeviantArt, so it feels hidden. Instead, you need to create an account and then mouse over your user icon on the top right of the page to find the “AI content settings” menu. From there, you have only two options: the default setting “show AI” or the option “repress AI” that says you will see “less instances” of AI-generated or manipulated images. After trying out both options, I'm afraid I don't see any significant difference. At that time, I have a pretty good sense of recognizing AI-generated “digital illustrations”, but I didn’t just have to rely on my suspicion – almost every dubious image I chose contained in the description an disclosure of a author who confirmed that the work was spattered by a robot. DeviantArt does bad work when automatically applying AI markings to metadata images that clearly indicate the origin of the AI. Pinterest has a similar system. Users who are logged in to a Pinterest account can click the Settings icon, select “Refine Recommendations” and then tap the “KI Content” tab to switch between certain categories, including art, beauty, fashion and home culture. If you disable one of these options, according to Pinterest “less AI-modified content” will be displayed for this particular category, but according to my experience, this is anything but perfect. The setting is also harder to find than a filter integrated in Pinterest feeds. I have still seen many images with suspicious AI-tells (including uncannyly perfect photo models and inexplicable image errors), although the AI filters were fully exhausted. And that's going to happen with some certainty when other platforms like YouTube or Instagram introduce an AI content filter: that won't work very well. But that's okay, because it would expose the ineffective “solutions” that our AI emperors dress up with. They exist on paper to appease regulatory authorities and critics, but they do little to address the real problem of discerning AI counterfeits from authentic photography and creative works. And platforms know it's a problem. Instagram boss Adam Mosseri said in December that “authenticity is becoming a scarce resource in view of the increase in AI-generated content.” And now, Sundar Pichai, CEO of Google, recently admits in a decoder interview that “there are a lot of AI vulnerabilities out there” and that online users need to “adapt to them”. Okay, give us filters. Provenance-based systems such as C2PA and SynthID function by embedding metadata or invisible watermarks into content at the time of creation. But there are many open source AI models that do not do this (especially if they are designed for shameful purposes), and even then metadata can be removed too easily to make this reliable. There are also recognition-based methods that analyze patterns in digital content and then evaluate the probability that AI was used to create this content. However, these can lead to incorrectly positive results. None of this currently works effectively on a large scale. Nonetheless, companies announce, including AI providers such as OpenAI, these AI labelling solutions are currently being deceived as something that will help prevent people from being deceived by deepfakes and other misleading forgery. If the regulatory authorities notice how ineffective they are, online platforms and AI providers may need to find a solution that actually works and not what currently feels like a fog wall. Platforms will argue that they risk falsifying authentic content if they drive labelling initiatives too much. Both Meta and YouTube had to find out this on the hard tour after having provided pictures and videos with AI markings, from which the creators said they were created without the help of such tools. If this is a major problem for current labelling systems, you will find a better solution. Certainly, improving the user experience for your millions of users is a rewarding investment to get ahead of the competition? And while I ask, why can't I report all the unnamed AI abuse reports I see every day? In view of the extent of the problem – a recent study by Kapwing revealed, for example, that more than 20 percent of YouTube videos that are shown to new users are qualitatively inferior content – I imagine that many human moderators would be required to effectively review each report. And maybe that's the hook. Can Big Tech, in a time when the workers are replaced by AI, which they are supposed to be able to outperform, deviate from their carefully designed narrative and re-establish them to solve the AI problems? Compared to automated moderation systems that lack differentiated investigative capabilities, people tend to face annoying demands such as salaries and social benefits. An alternative to the labelling of AI-generated contents would be to begin instead with the labelling of verified human creators. This would not necessarily identify synthetic content posted by these creators, but it could help us to see less of unverified content farms that produce inferior crap. This is the future that Mosseri proposed by Instagram for the Picture Sharing platform of Meta, and something that Spotify already does with verified artists. Of course, Meta, Spotify and Google not only host AI-generated images, ads and music; They are also responsible for the production of the tools with which it is created. Therefore, they insist that not all AI content is sloppy and that it is more about a quality issue. If they are persuasive enough, they hope you don't notice it and continue to sneak out of the trough. To allow users to filter it out anyway, all the efforts that these platforms have made to benefit from AI would run counter to you: you want you to accept the slop factory. I'm glad that the opposite is proven to me. Actually, I'm just suggesting that online platforms prove that the AI marking efforts were not a waste of time. But at the moment, they have all the cards in their hands and we just have to hope that their efforts for AI generation are successful. So give us a simple filter “no AI” or “verified human Creator” and we will judge how well it actually works. Most popular

– Microsoft and OpenAI have separated – now they are ready for the fight

– A first look at Microsoft's Surface Laptop Ultra and Surface Dev Box

- PlayStation returns to what it can do well

- Gemini Spark is the most impressive and frightening AI experience I have done so far

– Nintendo confirms that it will sell a new Switch 2 with replaceable battery in the EU