Podcast Summary: "Pokémon Go to The Military Industrial Complex"
The 404 Media Podcast
Release Date: November 27, 2024

Introduction

In this episode of The 404 Media Podcast, hosts Joseph, Emmanuel Mayberg, and Jason Kebler delve into two interrelated topics that highlight the intricate ways technology intersects with our daily lives and broader societal structures. The episode, titled "Pokémon Go to The Military Industrial Complex," primarily explores Niantic’s recent advancements in AI technology through their popular game Pokémon Go and transitions into a discussion on the burgeoning AI-generated influencer industry on platforms like Instagram.

Niantic’s Large Geospatial Model (LGM) and Pokémon Go

[00:00 – 03:30]

The conversation begins with an exploration of Niantic’s announcement of a Large Geospatial Model (LGM), which Emmanuel Mayberg explains as an AI model analogous to Large Language Models (LLMs) but focused on mapping and navigating the physical world. Emmanuel draws an analogy to help visualize the concept:

Emmanuel Mayberg [02:44]: "It's like predicting what an area is going to be. Imagine you're at a restaurant you've never been to before and need to find the bathroom. You've been to enough restaurants to know it's likely in a corner or along the edge."

Joseph clarifies Pokémon Go's role in this data collection:

Joseph [04:36]: "What data is being collected there? [...] the phone collecting information about surroundings or something."

Emmanuel details how data from Pokémon Go, such as location scans and user interactions, contribute to Niantic’s LGM:

Emmanuel Mayberg [07:32]: "Niantic collects a ton of data from Pokémon Go and other games. This includes pictures of real places attached to location data, feeding into the LGM."

Jason adds that Niantic incentivizes data collection through gameplay rewards:

Jason Kebler [09:48]: "They incentivize players to scan locations by offering rewards like Pokéballs, gamifying the data collection process."

Ethical Implications and Player Awareness

[12:59 – 19:07]

The hosts discuss whether Pokémon Go players were aware they were contributing to a massive data collection effort. Emmanuel categorizes players into two groups:

1. Casual Players:

   Emmanuel [12:59]: "They think it's just a game. They have no idea they're generating data for the company to use."

2. Tech-Savvy Players:

   Emmanuel [13:30]: "These players might suspect data collection but couldn't have predicted its use for developing an LGM."

Jason emphasizes the general lack of awareness among users regarding terms of service:

Jason Kebler [15:00]: "Studies show 99.8% of people don’t read terms of service, and most don’t understand them even if they do."

Emmanuel draws parallels to other platforms, highlighting unforeseen uses of data:

Emmanuel [17:08]: "AI’s current applications are new and unexpected, much like how YouTube processes drone footage for AI without creators' knowledge."

Jason adds that despite Niantic’s revenue from in-app purchases, the use of data extends beyond traditional models:

Jason Kebler [17:59]: "Niantic recontextualizes location data from targeted advertising to building a comprehensive mapping platform for AI."

Military Applications and Ethical Concerns

[19:07 – 25:55]

Transitioning to the ethical implications, the discussion centers on a presentation by Brian McClendon, Niantic’s Senior Vice President of Engineering, at a Bellingcat conference. Brian addressed concerns regarding the potential militarization of Niantic’s LGM:

Emmanuel Mayberg [21:01]: "At the conference, Brian stated that if a military uses the LGM in ways consumers use it, it's okay, but adding amplitude to war would be an issue."

Joseph probes the implications:

Joseph [22:50]: "What do we make of this? Could this AI model be used for military purposes or more innocuous applications?"

Emmanuel explains the potential for LGM to aid autonomous systems in complex environments:

Emmanuel [25:25]: "With centimeter-level precision mapping, robots could navigate warzones more effectively, contrasting with current autonomous systems that struggle with real-time parsing."

Joseph adds a practical example related to robotic navigation:

Joseph [25:55]: "Robots could predict environmental features like curb heights or textures based on extensive LGM data."

The hosts acknowledge the significant value of the collected data and the ambiguous future of its applications, balancing potential benefits against ethical dilemmas.

The Rise of the AI-Generated Influencer Industry

[30:12 – 46:18]

Shifting focus, the podcast delves into the AI pimping industry, exploring the emergence of AI-generated influencers on Instagram. Emmanuel and Jason discuss how these virtual personas create and monetize content that seamlessly mimics human influencers.

Characteristics of AI Influencers:

• Appearance: Predominantly female, often depicted in aspirational settings like beaches or luxury environments.
• Content Types: Includes both static images (grid posts) and dynamic content (reels).
• Authenticity: Highly convincing, making it difficult for regular users to distinguish from genuine accounts.

Monetization Strategies:

Jason outlines various methods through which AI influencers are monetized:

Jason Kebler [35:58]: "AI influencers generate content inspired by real creators, then monetize through platforms like Fanview or OnlyFans competitors, often involving stolen content from adult performers."

The hosts highlight the ethical concerns surrounding content generation and theft:

Emmanuel Mayberg [35:11]: "AI-generated faces are consistent across posts, giving the illusion of a real person, but they steal videos from real women to create viral accounts."

Communities and Instruction:

Jason describes communities like Digital Divas, where AI avatars teach others to create and monetize their own AI-generated influencers:

Jason Kebler [37:06]: "These communities offer guides and coaching on generating AI personas, claiming to be ethical by remixing rather than directly stealing content."

Impact on Real Content Creators and Instagram’s Response

[41:48 – 46:18]

The discussion turns to the repercussions for real content creators, especially within the adult industry. Emmanuel shares insights from speaking with Elena St. James, an adult content creator affected by AI-generated accounts:

Emmanuel Mayberg [42:26]: "Elena faces competition from AI accounts that can produce content endlessly without the effort required of human creators, making it harder for her to attract and retain followers."

Instagram’s Role:

The hosts critique Instagram’s response to the proliferation of AI-generated content:

Jason Kebler [44:39]: "During a quarterly earnings call, Zuckerberg indicated AI-generated content boosts engagement, hinting at the platform’s tacit approval by considering dedicated feeds for such content."

Emmanuel highlights the challenges in enforcing anti-impersonation policies:

Emmanuel Mayberg [46:16]: "Despite flagging numerous AI accounts to Instagram, only a few are taken down. The burden remains on the original content creators to report infringements."

Jason adds that Instagram's inconsistent enforcement disproportionately affects adult content creators:

Jason Kebler [46:18]: "Instagram's policies are vague, making it difficult for creators like Elena to protect their content without facing additional scrutiny on their legitimate accounts."

Conclusion

The episode intricately weaves together the complexities of data collection in popular gaming and the unintended (or perhaps intended) consequences of such practices in broader technological and societal contexts. From Niantic’s use of Pokémon Go data to develop advanced AI models with potential military applications, to the rise of AI-generated influencers disrupting traditional content creation on platforms like Instagram, The 404 Media Podcast sheds light on the hidden dynamics shaping our digital landscape. The discussions underscore the need for increased transparency, ethical considerations, and robust policies to navigate the evolving intersection of technology, privacy, and societal impact.

Notable Quotes

• Emmanuel Mayberg [02:44]: "It's like predicting what an area is going to be. [...] It's like training an AI model to do the same thing."

• Emmanuel Mayberg [12:59]: "There are two types of people who played the game... really ridiculous to assume that they would know that they are generating data for this company to use."

• Jason Kebler [17:59]: "Niantic recontextualizes location data from targeted advertising to building a comprehensive mapping platform for AI."

• Emmanuel Mayberg [21:01]: "Brian stated that if a military uses the LGM in ways consumers use it, it's okay, but adding amplitude to war would be an issue."

• Jason Kebler [35:58]: "AI influencers generate content inspired by real creators, then monetize through platforms like Fanview or OnlyFans competitors, often involving stolen content from adult performers."

• Joseph [25:55]: "Robots could predict environmental features like curb heights or textures based on extensive LGM data."

For more in-depth discussions and access to bonus content, subscribe to 404 Media at 404media.co.