AI Media

“I expect AI to be capable of superhuman persuasion well before it is superhuman at general intelligence, which may lead to some very strange outcomes.”

-Sam Altman

June 11, 2025 — Packetized media, the dynamic, granular media we consume and interact with on social networks, is transforming how we think and, by extension, how we organize society. As unsettling as this transformation has been so far, it’s far from over now that AI is being used to generate media packets. To understand what will happen, let’s break it down and put all of that AI innovation (LLMs to generative images, voices, and video) we’re seeing into context.

Packetized Sensemaking

Before we dig into what AI will be capable of doing, let’s dig into how packetized media works. It’ll give us a clue as to what AI-generated media will focus on.

• In response to the packetization of media, we’re experiencing a shift from long-form literary thinking to network pattern matching — from analysis (logic) to intuition (emotion). Patterns help us cope with high volumes of information, serving as a mental router that lets us categorize new information quickly.
• Packetized media is rocket fuel for intuitive perception because it is incredibly effective at conveying emotion in digestible chunks (below the level we’ve developed cognitive filters for), and emotion is at the heart of intuition. It provides you with the “feel” that lets you instantly understand something or the “sense” that something is off.
• Packetized media clusters around thematic patterns, from the social/cultural theme (trad family) to the political theme (MAGA fascism or woke communism) to the purient (viral videos). All of these themes contain an emotional subcontext that animates them. At the macro level, thematic patterns that resonate (likely due to some commonality in the emotional subcontext) can spontaneously combine into a greater whole; the red/blue networked tribes, for example.

This groundwork gives us a clear focus for AI media: it will generate themed, packetized media that triggers emotional reactions and empathy, enabling it to discover, build, and amplify thematic patterns. Human beings are already doing this— the last three elections were rife with it, and we see commercialized versions on TikTok and X — but AIs (at first with human participation but eventually autonomously) will take it to another level;

• Increase the rate and scale of packet production by several orders of magnitude. The flood will become a tsunami.
• Radically improve the emotive impact of individual packets by quickly stitching together generative video, voices, pictures, and text to maximize the emotional impact.
• This generative capability enables AIs to rapidly discover, explore, saturate, and then amplify thematic spaces.

Let’s flesh out this framework a bit with some detail.

AI Media Pumps

Here are some examples of the types of AI media pumps we’ll see (I’m already seeing them):

• The Trigger pump. AIs that generate emotive traps, from attention to thirst to wholesome, to generate emotions from desire to outrage.
• The Rabbit Hole pump. Endless packet streams that cater to niche patterns, from the political to the purient (think: Reddit stories, from betrayal to conspiracy, being repurposed and thematically varied).
• The FUD pump. Immediately creating packets that troll, misdirect, and distract.

Continued Below…

50 Major Companies That Will Likely Fail to Survive Trump’s MAGA Economy Many have 5-star ratings. Most are “buys” per Wall Street. But they’re dead companies. Click here to see the full list.

How These AI Pumps Work

How pumps will produce packets (all possible now, time will only improve the quality, speed of production, and the autonomy of the AI producing it);

• Text → LLM fleshes it out → Refine/edit → LLM generates variants → variants used to generate a packet cloud to test against live users.
• Increasing emotional impact → Use the successful text to generate visuals/pictures/videos. Add voiceovers (test to see which voices work the best) and/or lip-synched speech (if talking head) for packetized videos.
• Packet assembled (stitched together by AI) and posted to social networks. AI-enabled analysis generates refinement or reinforcement.

An AI-generated packet containing text, pictures, and a video with a voiceover currently costs $ 2.50 per post.

Financial Feedback Loops Spur Growth

Early AI-generated media will get an early boost from the following funding sources;

• Network pays for success (TikTok, X, etc., pays accounts with high traffic). Already seeing this.
• Funded (NGO to individuals to PAC). Due to its potential for political and social impact.
• Solo. Extreme leverage at low cost makes it possible for individuals to access this.

Sources of Media

Sources for AI packetized media pumps;

• Long-form media (podcasts, movies, TV shows, books, articles, etc.), chopped up into parts (packets).
• Successful human-generated examples on TikTok, X, Reddit, Facebook, Instagram, etc., that generative AI can quickly identify and easily exploit autonomously. ‘This works, let’s copy it, ’ from an area of focus to specific posts.
• Dynamic discovery. Try something novel (either based on human creative insight or AI-generated variants not currently explored), see if it works in the wild, discard failure, and reinforce success.

Interactive AI Packets

AI media pumps won’t only generate posts, they will deliver packetized interaction by mimicking human interaction.

• Simple bots (limited to basic tasks). Pre-formatted responses. Bots are already common.
• Agent chat (LLM-based, human low-rez mimic). Contextually nuanced replies, expressions of support, or trolls.
• Social AI (LLM + Social AI layer, full human mimicry). Full personality. Full range of emotion. Potential to serve as a trusted hub for a thematic pattern, much like many big accounts on X do today. It will likely become impossible to know if the account is a real person or not.

How AI Media Pumps Mature

• Low. A focus on participating and subtly influencing a thematic pattern (amplify, dampen, or financial gain).
• Mid. A focus on filling and dominating thematic patterns and going to war with others trying to do the same.
• High. Complete control over multiple thematic patterns AND the ability to shift, truncate, and redirect them. This control is the outcome that governments and tech companies seek.

Instead of trying to summarize all of the potential impacts, let’s focus on how AI media will impact a foundational social structure: culture.

AI Media’s Impact on Culture?

Here’s a high-level orientation that may prove helpful when thinking about this topic. As we have already seen, packetized media has been destructive of culture — the shared set of beliefs, values, norms, and behaviors that a society uses to stabilize its interactions, both internally and with the external environment (see this report for more).

• AI media will quickly become the universal solvent of culture, as it will shatter any shared understanding provided by it, as an endless stream of AI-generated media exploits every potential divergence. The result will be a complete loss of societal coherence and cohesion (we’re already waist deep in this).
• That leaves two alternatives. An open system that uses AI media to stitch together coherent and cohesive meta patterns that people can choose between. Or a single centralized pattern controlled and enforced by a tyrannical government (China’s solution).
• Which works better?
  • Open patterns (competing with each other) will change, adapt, and innovate faster. A monolithic closed pattern will be more stable.
  • While open patterns generate a divergence and unpredictability that can create instability in the short term, they tend to navigate in the general direction of reality due to feedback from their impact on the external environment. The trick is surviving the near-term chaos.
  • In contrast, closed patterns are very stable in the short and medium term; however, they have the potential to deviate significantly from reality due to an inability to adapt to the challenges of a changing environment. NOTE: We saw a minor example of this when a closed media ecosystem adopted a globalized orientation and shut down any feedback loops generated by the failure of this orientation in the real world (from the border meltdown to a lack of general prosperity to Afghanistan).
• Optimally, the system that works best over the long term allows patterns to develop and diverge, but sets a boundary condition that prevents excess. A simple, minimal rule set that prevents the system from going open loop (running off the rails).

What To Do Personally

Here are some ideas on what you can do to prevent chaos generated by unrestricted AI-generated media.

• At the macro level, oppose closed systems that control pattern alignment (at both the government and corporate levels). The risk of tyranny from a system like that is too significant. At the local, be skeptical of patterns that you find yourself pulled into. Patterns that feel instinctual and make you reactive.
• Start to use AIs as a way to filter our packetized media and detect intrusions (such as fakes, intent, persuasion detection, theme/pattern, etc.). We’re already seeing the start of that with Grok on X. Eventually, we’ll see AI become the equivalent of anti-virus/anti-malware for our minds. At that point, if you can, opt for an AI that you have complete control over (run locally, or in your cloud account). Suboptimally, pick a vendor you can trust (a good sign is that it doesn’t censor you).
• Long term, once social AIs arrive, they will become a constant companion. A tutor/nanny/guard for those unable to direct it themselves (kids, elderly, etc.), a trusted advisor or coworker for you. As a minimum requirement, the social AIs you have control over shouldn’t be pre-aligned, allowing you to set the alignment. Optimally, your interactions with it and what it learns from its interactions with you should be something you own. If you don’t, the downside risk is extreme.

John Robb
Global Guerrillas & Grey Swan

P.S. from Addison: One of our first forecasts for the new Grey Swan Investment Fraternity, made in February 2024, was that President Biden wouldn’t be the Democratic candidate in 2024.

We were, of course, correct, if early. We anticipated chaos in deciding his replacement – an area where our expectations were wrong. Who knew the Democratic Party would appoint a candidate rather than elect one… stalwarts of defending democracy they insist on being.

Ha. Ha. Ha.

One reason we expected some chaos was from the rise of AI media and “deepfake” videos. The trend was well in place before John’s summary today.

While AI media didn’t play as influential a role in the 2024 election as we anticipated, the game is changing fast, “quickening” as the kids like to say.

New tech is being deployed and iterated upon by the day. Tribal networks are quite literally fighting in the streets in LA and in cities across the country. AI is already a potent factor in close races in the 2026 midterms.

Since the publication of his book Brave New War in 2007, John has been a leading voice in online networks and military tech development. He’s a former advisor to the US Joint Chiefs of Staff on “net war” and a regular contributor to The Grey Swan Investment Fraternity.

John will join us for Grey Swan Live! tomorrow morning at 11 a.m. EST to unpack what’s going on with Ukraine’s drone attacks against Russian military targets – deep behind enemy lines. As well as the deployment of AI technology in the Asian theatre and the development of the Golden Dome.  If you’re a paid-up member, please join us with questions.

With the metals space in mind as structural fiscal spending issues remain, our portfolio director, Andrew Packer, will be attending the Rule Investment Symposium in Boca Raton FL, July 7-11, 2025

The Symposium is Rick’s continuation of the Agora Financial Symposium we co-sponsored in Vancouver with him for a decade in the early 2000s.

Now, in Boca Raton, it’s a five-day affair featuring in-depth research from dozens of small-cap resource companies, including gold and silver mining companies, but also copper, uranium, and other critical commodities we’ve explored in-depth in our research over the past year. Click here to view the stellar speaker line up and learn how you can attend yourself.

Your thoughts? Please send them here: addison@greyswanfraternity.com