🧩 #10- AI For Humble Humans: Detecting AI generated content, a matter of life or death!

Dec 28, 2023

📍THE news that needs your attention

Adobe created a symbol to encourage tagging AI-generated content - This new icon is part of its initiative to provide transparency in AI-generated work.

 

The short of it: This seemingly unimportant news actually represents a significant development—a rising trend called 'watermarking.' This trend has been adopted by major brands and platforms currently transitioning to AI, including Google, Microsoft, Meta, and Adobe (among many others). More importantly, watermarking is a growing sub-topic within a broader subject: the detection of AI-generated content. This topic, as of now, is undeniably one of the most crucial aspects of AI and the transformation it will bring to today's world.

Source: (🔗 Here)


📝 Most recent updates about watermarking

Watermarking has been widely deployed over the past few months.

  • Adobe’s initiative is part of a larger one led by the coalition for content provenance and authenticity (🔗 Here).

  • Open AI says new tool will detect images made by Dall-E (🔗 Here).

  • Meta just revealed that they will introduce invisible watermarks in AI-generated images (🔗 Here).

  • Google Deepmind beta-launched SynthID, a tool for watermarking and identifying AI-generated images (🔗 Here).

  • Digimarcs adds copyright information to digital data (🔗 Here).

That being said, as of today, watermarking is not functioning effectively. Recently, a research team found it’s easy to evade current watermarking methods and even add fake watermarks to real images (🔗 Here).

Watermarking identifies AI-generated images but the larger issue of AI-generated content also includes other types of content: written, audio, and video.

Watermarking identifies AI-generated images but the larger issue of AI-generated content also includes other types of content: written, audio, and video.

 

🌎 Detection of AI-generated content: what’s at stake?

Nothing else than a new definition of the words REAL and TRUTH.

So far, in a pre-AI world, there has been a strong connection between truth and content. Remember those popular sayings?

  • Words may fly away, but writings endure.” (written content)

  • I only believe what I see.” From the French Philosophe Saint Thomas d’Aquin. (video and/or visual content)

  • Words are weapons” or “words possess a creating power” (audio content).

In an AI-driven world, these no longer hold true.

According to the SEC, AI could cause a financial crash within a decade (🔗Here). In itself, generative AI is accelerating disinformation (🔗Here).

Simply put, It introduces ‘fake’ into every one of our homes, which ultimately makes it a potential deal breaker in the balance between the world forces.

Peace hinges on justice, and justice relies on truth.

AI compels us to completely redefine what is real and what is not, what is true and what is not. Our future ability to detect AI-generated content is indeed a matter of peace or war. A matter of life or death.

Anyone who follows the latest and most devastating news in the Middle East can attest to this…

 

🔮 Solving the AI content detection problem?

There are basically 3 ways:

  • Through technology. As of now, technology is not efficient enough to truly detect AI content. That will probably be quickly fixed.

  • Through regulations. This is happening now worldwide, the most recent initiative being the No Fake Act (🔗Here) and the European initiative taken this past week regarding the Israël-Hamas conflict (🔗Here). But it won’t do miracles. Those who will want to behave ethically with AI will always do so. The problem will be: what about those who don’t intend to use AI ethically? Will any law ever be able to track, find, and sanction them?

  • Through private initiatives. Halfway of the 2 above, this one is right now the most efficient. Large brands like Google will invite promptly, private entities to behave AI ethically, especially during the coming Presidential election (🔗Here).

👉 At an individual level, this presents an opportunity for anyone to cultivate a personal brand. Keep reading to know why and how.

🧠 What does that mean for us, humble humans?

  • AI content is threefold - Written, audio, and visual. Each of these content types faces significant threats from AI whether through copywriting, plagiarism, fake news, bias, disinformation, or propaganda. These threats have given rise to an entirely new set of lexical fields that challenge the concept of “truth” as we once knew it. This includes phenomena like deep fakes, synthetic media, or voice cloning.

  • AI Content is undetectable - It’s a fact verifiable multiple times. As of today, AI content is undetectable (🔗Here). AI experts like Ethan Mollick demonstrated it in many ways (🔗Here). Even Open AI shut down recently its AI-written content detection tool (🔗Here) and now trying to implement one in its visual content generation tools (Here). Each AI content detection tool works approximately the same way: they won’t detect what is meant by AI or not, they will detect what might be authentically written by a human or not. Authentic is the opposite.

  • Artificial v/s authentic - AI content detection tools, whether it’s for writing, audio, or video content use human factors present or not present in the content, to determine whether or not it’s human or artificial. Your best bet is to be “you”.

  • Truthful = Authentic = Trustful - Until today, the word truth is defined as “the body or real things, events, and facts”. Real things, events, and facts that we, as human, can either see, hears, smell, touch, or taste. AI is affecting at least 2 of those human senses. Starting now, it won’t be that much about what you can see or hear, it will be about what you will trust, check, and to whom you will attribute.

👉 A great marketing opportunity for you: Establishing your own set of AI ethics rules for your brand (whether corporate or personal) in relation to your content creation is going to be a significant advantage. Keep reading to see where I’m going with this.

 

 


🎯 THE insight from LinkedIn

This post from Paul Roetzer - The rise of synthetic media is a predictable disaster because the average person will not be able to know whether what he/she will see online is real or not.

 

The short of it:

  • Traditional media are not that much a reliable source of information anymore. Whether it’s because they’re partial - No media is “neutral” anymore and bias is everywhere - or because synthetic media are here to stay.

  • Social media are unable (voluntarily or not) to control whatever is shared on their platform. They became a worldwide distribution channel - and to some extent the one motivation - for some terrorists around the world to create evil on earth, to create shareable content by taking pictures or videos of it and exporting it into every single home to spread terror, chaos, nightmare, traumas - and influence decisions.

Sources: Check the full post (🔗Here).

 

🧠 What does that mean for us, humble humans?

  • Become your own “journalist” - In an AI-generated content era, traditional media are no longer an unquestionably reliable source of information. Here lies an opportunity for anyone to create their own reliable content. Reliable = authentic.

  • Create authentic content - Your truth. Be bold. Be divisive. Be straightforward. Challenge the status quo. Make yourself accountable. Take responsibility for your opinion. Be that good that you’ll be the source of another content creator.

  • Be truthful - Aka transparent. Share your sources. Requests consent from your sources. Give them credit (always). Disclose how and when you used AI to create it. Medium’s CEO took those sorts of Guidelines for its own platform (🔗Here).

  • Become obsessed by Fact-checking - Claims need to be backed by citations. An initiative that was taken this past week by Wikipedia which start using AI to improve verifiability (🔗Here). Traditional media are trying to do so too. Associated Press is using it widely these days regarding the Israël-Hamas situation (🔗Here). Their goal? to address widely shared misinformation, including work with outside companies and organizations to add factual context to misleading content that is circulating online (🔗Here).

  • Favor raw content - Raw? Raw content consists of unedited videos shot directly from a GoPro, captured spontaneously in a vlog-style format. 'Raw content' doesn't imply a lack of forethought or research. Instead, it refers to live or unpolished content that doesn't strive for excessive perfection, sophistication, or neatness. It's clean, smart, yet casual.

  • Keep control over social media - Not only through protecting your personal data but also by refraining yourself from using social media both as a source to be informed - and as a way to spread your ideas.

  • Distribute ultra-selectively - Save its very first distribution to a tight group of trusted and selected people that will serve as your first tribe not only to spread it (or amplify it) but as to vouch for its authenticity and the accuracy of THE truth.

  • Consistency as a component of Trust - To get people to trust your content, you’ll have to get them to trust you. We trust people who are consistently digging into their topic in the long run... We won’t trust those who will be skipping from one topic to another. Sprint v/s marathon. Your ability to do all of the above will depend on your ability to consistently stay focused on your topic.

     

 


🧩 AI Marketing Corner

Topic: Personal branding → Authenticity → Trust → The hierarchy of needs pyramid for creating authentic content.

🤗 1 tactic (from me):

Truth leads to trust. In an AI era, what is accurate is reliable.

Trust is at the core of any decent marketing strategy and content creation is the only way to build trust - or to destroy it.

We are entering a new era of content creation.

The one in which AI will assist us to create better content than ever - but better will be different than what we know so far. We need to change our worldview of what is ‘good content’.

For your content to be considered authentic, it will need to cover some basic requirements for it to be trusted. Kind of a hierarchy of basic needs, like in the Maslow pyramids, applied to an AI content strategy framework.

Here is how I see it:

 

 

🛠 6 AI Content Detection tools

❌ Don’t use those tools to get a 100% accurate answer “yes” or “no” it’s AI or it’s not.

✅ Do use them as a set of clues that informs you about the level of authenticity of content, and thus, its credibility and the level of trust with which you can grant its author.

Here is a short list for you:

  • Copyleaks - Mostly a plagiarism detection tool, it uses advanced algorithms to analyze text and identify similarities with other content on the web. They can detect AI-generated content in 30 languages with a claimed 0.2% false positive rate, which they claim is the lowest of any platform.

  • Originality. ai - They argue that they can avoid AI content detector false positive and offers an API for integration with other tools and platforms.

  • Writer. ai - A well-known in the AI-generating text space. It’s an AI-powered writing assistant that can help content creators generate high-quality content more quickly and efficiently. They also offer a content detection tool that can help you identify AI-generated content within an article.

  • Content at scale - As the others, they’re not gonna tell you what is made by AI, they will provide you with a “human content score” that is a mix of “predictability, probability, and pattern” scores. Those scores tend to outline the reasoning behind the overall rating which is valuable.

  • ZeroGPT -The one I tested more than the others. Here is a short video with my feedback and some fun tests made with the marketing leader Mark Schaefer and some of his big hits content 😉 (🔗Here).

     

    In 7 min my testing of GPTZero with

     

🤓 THE content that is worth your reading:

This insightful inquiry was made by Wired, retracing the story and evolution behind the creation of an AI-generated content tool ZeroGPT

👉 The AI Content Detection Race is on (🔗Here).

“(…) It also previews a possible paradigm shift in how we relate to digital media.

The whole endeavor of detection suggests that humans leave an unmistakable signature in a piece of text—something perceptible—much the way that a lie detector presumes dishonesty leaves an objective trace. Provenance relies on something more like a “Made in America” label. If it weren’t for the label, we wouldn’t know the difference.

It’s a subtle but meaningful distinction: Human writing may not be better, or more creative, or even more original. But it will be human, which will matter to other humans”.

Wired.
 
 

🙌 Let’s work together!

Hire me to speak about AI marketing to your marketing team.

I’ve been preaching AI marketing in top Digital marketing agencies both in France and in the US.

My focus is on Introduction to AI Marketing and Google SGE and its impact on content marketing.

Check this out 👉 here.

 

Jump on this form

Get started with AI marketing & AI content strategy

Subscribe to AI for Humble Humans and get actionable and understandable AI marketing tips, insights, prompts, tactics, and Tools. Dedicated to Entrepreneurs and Marketers.

100% human writing. No BS. No AI jargon. 3-5 min read max. Delivered twice weekly. 

 


Copyright© 2023 karine abbou.com