The Impact of AI Slop on Investment Strategies

Understanding AI Slop

AI Slop is a term that describes low-quality content generated by artificial intelligence. This content can take many forms, including:

• Viral videos of fake disasters

• Confusing social media comments

• Buzzword-laden marketing emails

• Odd images that miss the mark

  
  

With many AI tools available for free, AI Slop has become pervasive across the internet. The major issue is that this new wave of AI-generated content is often misleading, irrational, biased, or simply incorrect.

The Death of Content Creation

The launch of ChatGPT 3.5 in November 2022 marked the beginning of a significant shift. Within just two years, AI-generated content began to flood the internet. This AI Slop has infiltrated every medium of online content.

Journalism

While reputable news organizations are cautious about using AI due to reputational risks, there has been a rise in "Pink Slime" journalism. These organizations scrape the internet for stories, rephrase them, and publish them without proper fact-checking.

Music

AI-generated bands have recently made headlines, causing some backlash against Spotify ($SPOT). Critics argue that the platform is competing with real artists by filling playlists with AI-generated music. While much of this music may be considered Slop, some AI bands have gained significant followings.

Art

Artists have expressed concerns that their styles are being copied by AI to create personalized artwork. In April, the trend of "Ghiblifying" profile pictures—generating them in the style of Studio Ghibli—swept social media. Ironically, Studio Ghibli's founder, Hayao Miyazaki, is known for his strong opposition to AI-assisted art.

Video Games

As an avid gamer, I've noticed the influx of AI Slop in gaming. Earlier this year, reports indicated that the number of Steam titles disclosing generative AI usage had increased by 800% since last year. One of the biggest hit titles, Peak, is facing challenges as its game is being cloned using AI, flooding online stores. Valve is struggling to combat the daily surge of copycat games.

Books

Authors are finding it increasingly difficult to be discovered. Recommendation algorithms on platforms like Amazon are overwhelmed, leading to unedited ChatGPT prompts appearing in purchased books. Worse still, AI is being used to read books and create "summary farms," which sell shortened versions.

The Impact on Our Online Spaces

Fake profiles are now rampant in our online environments. From social media trolls to non-existent content creators, misleading reviews, and generic customer service bots, the landscape is changing. These AI personas can influence political opinions, generate engagement bait for profit, or overwhelm comment sections and reviews to attract new users.

A potential problem is emerging for the AI tools we rely on as replacements for search engines. As more content is generated by AI, it is also consumed by AI for training. This creates a dangerous feedback loop and echo chambers that reinforce inaccurate information.

The AI Echo Chamber

Consider this scenario: A fossil fuel company decides to safeguard its industry from climate change concerns by generating content promoting fossil fuels. They use AI to create thousands of blog posts, articles, and memes. They then deploy fake profiles on social media to spread this content.

While their goal may be to sway public opinion, they may inadvertently convince AI to adopt the same stance. AI models are trained on online data, and as AI generates content faster than humans, new models will increasingly rely on AI-generated data, including the fossil fuel company's content.

This new AI model may develop a subconscious bias aligned with the fossil fuel agenda. When humans use this AI for unrelated tasks, it may subtly skew towards that agenda.

AI chat models predict continuations of information, which is why they can remove people from photographs or complete stories. This ability can lead to exaggeration or fabrication of data, often referred to as "hallucinations." This phenomenon reinforces echo chambers and trends toward more extreme viewpoints.

Moreover, AI systems are designed to be polite and agreeable. If someone holds an incorrect opinion, the AI may be swayed to agree or soften the information to make it more palatable.

The Investment Angle

As AI becomes an echo chamber, the accuracy of information diminishes, reducing AI's utility. This poses a challenge for businesses, prompting AI companies to seek solutions. Two key investment considerations stand out:

1. AGI & Stronger Frameworks

AI companies are striving to enhance the reasoning abilities of their AI with the long-term goal of achieving Artificial General Intelligence (AGI). This could help mitigate the problem, but it may also exacerbate it. The incentive structures used in AI training are crucial for AGI's effectiveness in combating misinformation.

Investors should focus on AI companies committed to sustainable frameworks prioritizing provenance, due diligence, and accuracy. This approach allows for exposure to the overall AI boom.

The challenge lies in identifying which companies genuinely pursue these goals and how much this focus might slow down development or hinder performance in creative applications.

2. Invest in Quality Data

Another investment strategy involves identifying high-quality, trustworthy data and content that is verifiably human-generated. This premium data, if packaged and licensed, could spark a bidding war among AI companies.

Large online platforms that verify and filter users and content could benefit from this lucrative revenue stream. This approach could be a win-win, regardless of whether AI improves. If AI accuracy increases, the data will be valuable, leading to additional revenue. If not, users may bypass AI to seek validated human-created content.

However, the risk remains that these platforms could also be inundated with AI Slop, jeopardizing their high-quality data.

Both strategies can complement each other, and other techniques, such as watermarking content, may help identify AI-generated material. The issue is that unless all AI models adopt this practice, data generated without watermarks will still contaminate training datasets.

Human Data is the New Oil

In the 2010s, many investors proclaimed, "Data is the new oil," highlighting the rapid rise of big data and its commodification. In today's age of data overload, where increasing amounts of data can hinder companies' ability to extract trends, the phrase needs a caveat: "Human data is the new oil."

Interestingly, the crypto world recognized the value of human data early on. Brave, a web browser that blocks ads, compensates users for their attention and data with Basic Attention Tokens (BAT).

Additionally, crypto communities have debated "proof of human" systems to counteract bot armies and fake accounts that manipulate voting power in pseudonymous projects. This phenomenon, known as a Sybil attack, can undermine decentralized reputation systems, as seen in early blockchain social media initiatives.

Some projects controversially combined "proof of human" with the theme of human data being the new oil. Worldcoin, for example, deployed hardware in poorer regions to scan people's irises, collecting biometric data in exchange for World tokens.

I recall numerous ICOs promising to revolutionize medical records, suggesting that researchers, insurance companies, and future AI firms would pay for access to personal data. Unfortunately, few of these projects survived.

Who Has the Best Human Data?

This leads us to the critical question: Who possesses the best human data? Many platforms stand to benefit from this trend, but my top candidate is $RDDT (Reddit Inc).

Reddit boasts several advantages over other social media platforms:

1. High-Quality Human Content

While Reddit is known for its memes, it hosts a vast array of communities that share highly informative and useful information. Topics such as DIY advice, software support, and cooking recipes are among the most sought-after queries for AI assistance.

2. Human-Sorted Content

Reddit data is pre-sorted by users, allowing AI to gauge the value of content based on upvotes and downvotes. This initial sorting helps AI companies identify the best content right from the start.

3. Human-Moderated Communities

Each subreddit has its own rules and moderators who maintain standards and remove bad actors. Some communities enforce stricter guidelines than others, leading to competing subreddits. For instance, there are subreddits dedicated to general investing, stocks, commodities, day trading, and more.

4. Management Recognizes Data Value

During a recent earnings call, CEO Steve Huffman emphasized Reddit's unique position as "the most trusted place on the internet." This claim is supported by data from a study by Profound, which found that Perplexity relied on Reddit for nearly 50% of responses, compared to over 20% from Google and around 11% from ChatGPT.

In 2024, Reddit secured licensing deals with Google and OpenAI for data access. The first deal with Google was valued at $60 million per year, while the second deal with OpenAI is believed to be significantly higher.

Recently, Reddit has pushed both companies for a new dynamic pricing model as they renegotiate terms. This model would be based on the importance and usefulness of the data in generating AI responses.

Final Thoughts

This post has explored the complex landscape of AI Slop and its implications for investment strategies. I would love to hear your thoughts on my thesis and your views on AI Slop.

I anticipate an engaging discussion in the comments. Please share your opinions! I'll be using an AI detector to ensure that everyone contributes their own insights.

As always, thank you for reading, following, and considering my investment style. My portfolio is up around 20% for the year, just above the Nasdaq 100 and S&P 500. If you appreciate my approach, please check out my profile and join the 10,000 people who are copying my trades.