Really, would you have done this without AI? Prove it

“This sounds like artificial intelligence.”

It’s a phrase I’m afraid to see as a writer who dabbles in amateur illustration and photography. In a world where generative AI technology is increasingly adept at mimicking the work of humans, it’s natural for people to be skeptical when online platforms… Refuse to classify even obvious AI content.

This leads me to one conclusion: perhaps we should start labeling human-made text, images, audio and video with something closer to the globally recognized fair trade logo. Machines certainly aren’t eager to have their work categorized, but the creators in danger of being replaced certainly are.

Fortunately, I’m not alone in my thinking.

Instagram header Adam Mosseri suggested the same thing in December, saying it would be “more practical to fingerprint real media than fake media” as AI technology improves to the point of creating content that is visually indistinguishable from content made by creative professionals.

No one can say for sure how much of what we find on the internet is generated by artificial intelligence, but there is a widespread perception that news sites, social media platforms and search engine results are full of it, according to a recent study. A recent Reuters Institute poll.

Verifying man-made works was one thing C2PA content credential standard – which Meta platforms already use – was supposed to do that. But so far, its implementation has been woefully ineffective, despite receiving widespread industry support. It turns out that a lot of people who create and curate AI content are motivated to hide its origins because of the clicks, clutter, and money it can generate.

In an effort to help human creators differentiate their work from that of AI generators, a plethora of solutions have emerged in recent years. As with C2PA, they face a number of challenges to widespread adoption.

There are currently too many AI-free classification alternatives to choose from. In total, I counted at least a dozen people, all trying to address the same problem using a variety of eligibility criteria and authentication methods. Some are industry-specific, such as Authors Guild.Human-authored testimony” for books and other written works, and cannot be applied broadly to all forms of creative content.

Other solutions like Proudly Human and Not by AI aim to be broader, covering published text, visual arts, video, and music, but the verification processes used by these services can be as questionable as those used by AI tagging solutions. Some, like Man madeoperates on a pure trust basis, making badges and tags available for anyone to download and apply to their business without actually identifying the source. Others like No AI code They say they visually scan businesses and run them through AI detection services, which can be notoriously unreliable.

Most of the services I examined do it the hard way: by having creators manually submit their work processes to a human proofreader, such as sketches and written drafts. It’s very labor intensive, but without any technological shortcuts, it’s the most reliable way we currently have of determining whether something was made by an actual human.

Another issue is agreement on the meaning of the term “man-made”. With artificial intelligence now being integrated into many creative tools, it is being used Encouraged by creative teachersWhere do you draw the line?

“The problem will be definition and verification. Does chatting with an LLM about an idea before implementing it manually count as using AI? And how can a creator prove that there is no AI?” said Jonathan Stray, a senior scientist at the UC Berkeley Center for Humanized Artificial Intelligence Edge. “Other consumer labels, like ‘organic,’ have regulations and agencies that enforce them.”

We’ve already entered an era of hybrid content that goes against how we define something as authentically made, says Nina Bigos, a lecturer at the School of Information at the University of California, Berkeley.

“Any creative production today can be affected by AI in one way or another without us being able to prove it,” Bigos said. Edge. “Authorship is disintegrating into new directions, becoming more technologically enhanced and more collective. We need to revamp our standards of creativity that were set only for humans.”

The solution offered by a competitor to man-made labels is called Not by artificial intelligence He tries to take this ambiguity into account. It offers a variety of badges that creators can apply to websites, blogs, art, films, articles, books, podcasts, and more, provided that at least 90 percent of the work is created by a real human. But the voluntary approach lacks any veracity check.

Other solutions such as Proof that I did it They rely on blockchain technology to provide a permanent record that anyone can use to point to creators and works verified by the service. By storing verification on the blockchain, creators receive an unforgeable digital certificate proving that a human created their work, which is much more reliable than trying to use software to guess whether a piece of media was created by AI.

Thomas Baer, ​​executive director at UCLA’s Rady School of Management, says Web3 and blockchain technology can provide a powerful solution by shifting the question from, “Does this look like artificial intelligence?” to “Can this account prove its human history?”

“By issuing ‘Made by Human’ tokens to approved creators, the market creates a ‘premium level’ of art where authenticity is mathematically guaranteed,” Baer said. Edge. Other experts such as Bigos echoed similar sentiments regarding the potential increase in value of “human and biological creativity” amid the deluge of synthetic media.

Despite their flaws, established standards like C2PA provide something that AI-free labeling solutions desperately need: standardization. Big names in the technology industry, such as Adobe, Microsoft, and Google, have committed to this standard, and AI providers are working to implement it to satisfy global regulatory bodies. However, when I weigh the different pros and cons between AI-driven labeling efforts and those focused on verifying original human-made content, I feel the latter is more likely to succeed.

Many creative professionals, even those who are not completely opposed to using AI tools, have understandable motivations to differentiate their work from the artificially generated competition that saturates the industry and threatens their livelihoods. And while yes, there are plenty of AI evangelists across social media platforms who are happy to showcase what the technology can achieve, there is a reluctance to disclose its use when money and influence could be lost.

Take a case Porn actors create digital copies Of themselves they will stay hot and young forever, or Artificial intelligence influencers Selling a fantasy life that doesn’t exist. Revealing that they are artificial intelligence may break the illusion of people who think they are getting a real human experience. Scammers using images generated by artificial intelligence To sell products online You certainly don’t want to be exposed either, and the platforms like Etsy that host them She doesn’t look too worried. Likewise, anyone who uses generative AI to sow discord or create harm on social media can only succeed when people believe it is real. It’s no wonder that AI classification using C2PA has failed to catch on.

We know that some AI-focused creators will eschew transparency because this is already happening. A notable example of this is Coral Hart, the author of the romance that narrated New York Times It made a six-figure sum after producing more than 200 AI-generated novels last year. She does not have a label on any of her books revealing that they were written using AI tools, due to concerns that it would “hurt her business because of this work” due to the “strong stigma” surrounding the technology.

We can see this disdain in action with the number of times artificially generated content is described as “sloppy,” even if the works themselves are visually, audibly, or technologically impressive. This raises the question of how man-made or AI-free labeling providers can prevent misuse of their logos by those who profit from deception. Trevor Woods, CEO A proud human beingacknowledges that doing so may not be possible.

“Like certification marks and other company logos, we cannot prevent the Proudly Human certification mark from being fraudulently displayed,” Woods said. “However, we do make it easier for consumers to verify it.” Edge. “If the bad actor we have identified refuses to stop using the designation, we will take legal action against them.”

If the goal is to achieve a globally recognized and implemented solution, the standard must be agreed upon not only by creators and online platforms, but also by global governments and regulatory authorities. As far as I know, these conversations are currently few and far between.

“Proudly Human has at times briefed government and industry associations, but has not participated in formal negotiations regarding a standardized certification of human origin,” Woods said. “The rapid evolution of AI capabilities and AI-generated content will outpace government and regulatory responses.”

There is clearly a demand to make human-made works easier for consumers to recognize, so creators, regulators and authentication agencies need to choose which approach to endorse. If a single standard can rise to the same level as symbols like fair trade and organic – which have their own concerns, but are universally recognized as something that conforms to a certain ethos – perhaps we can return to the days of trusting what we see with our own eyes.