How does feedback work in mega nets?

Feedback in mega nets means that the system continuously incorporates the results of its previous state into its next state. User interactions constantly shape algorithms and data streams, leading to unpredicted amplification and 'out of control' behaviors.

What is the trade-off between free speech and dealing with misinformation?

There's a perceived trade-off between the freedom to say anything at any scale and the ability to know what's real and cooperate effectively. While free speech is defended, the issue of algorithmic amplification of misinformation and outrage is a key concern.

Are large language models like GPT-3 and GPT-4 truly intelligent?

The speaker argues that LLMs are incredibly good at producing content but lack genuine understanding or reasoning. They are compared to the horse Clever Hans, good at responding to cues but not performing true cognitive tasks.

What are the near-term effects of AI on content creation and reality?

In the next year or so, AI's impact will be mainly novelty. However, within a few years, AI will increasingly collaborate with humans in content generation, and more significantly, people will believe AI is thinking, leading to major changes in human perception and the potential 'spamification' of online content.

How can we deal with the tsunami of fake or unreliable information online?

The speaker suggests that people may declare 'epistemological bankruptcy' regarding the internet. He also notes that skepticism is hard to maintain, and communities will increasingly become 'narrative bunkers' where shared assumptions reinforce beliefs, leading to a divergence in how groups perceive reality.

Key Moments

The Digital Multiverse: A Conversation with David Auerbach (Episode #319)

Q: What are some examples of mega nets beyond social media?

Mega nets extend beyond social media to include online gaming, cryptocurrency networks, and even integrated government identification systems like India's Adhar. AI is also considered a component or a type of mega net.

Sam Harris

Science & Technology4 min read37 min video

May 15, 2023|23,494 views|380|132

sam harris sam harris podcast waking up podcast waking up waking up with sam harris sam harris waking up waking up sam harris sam harris jordan peterson sam harris joe rogan author neuroscientist philosopher

Save to Pod

Key Moments

TL;DR

Digital networks ('MegaNets') amplify content uncontrollably, blurring reality and challenging cooperation.

Key Insights

MegaNets are complex, opaque data networks combining algorithms, AI, and user interactions, leading to uncontrollable velocity, volume, and virality.

The current digital landscape presents a trade-off between informational liberty and societal cooperation, particularly concerning misinformation.

Companies like Google, Facebook, and YouTube have less control over content than users perceive; user interactions profoundly shape algorithms.

AI, especially Large Language Models (LLMs), are sophisticated content generators but lack genuine understanding or reasoning capabilities.

Deepfakes and AI-generated content will increasingly blur the lines of reality, potentially leading to 'epistemological bankruptcy'.

The future may see a further fragmentation of realities into 'narrative bunkers' reinforced by online communities, hindering broad understanding.

DEFINING THE 'MEGANET'

David Auerbach introduces the concept of a 'MegaNet' as a persistent, evolving, and opaque data network. These networks encompass both algorithmic and AI-driven systems and the hundreds of millions of connected people. They foster a feedback loop of accelerating content production and distribution, characterized by high velocity, volume, and virality. Auerbach uses this term because traditional networks underestimated the human component's influence on shaping algorithms and server behaviors, an interaction far more complex than purely technological control.

THE PERVASIVENESS OF MEGANETS

While social media platforms like YouTube are prominent examples, MegaNets extend beyond them. Online gaming, cryptocurrency networks, and even integrated government identification systems like India's Aadhaar can function as MegaNets. Auerbach argues that AI itself is a component or emergent property of these systems, with much of its disquieting behavior stemming from pre-existing MegaNet structures rather than AI technology alone. The GameStop stock phenomenon serves as an example of how these feedback loops can manifest in economic spheres outside traditional social networks.

FEEDBACK LOOPS AND LOSS OF CONTROL

Auerbach defines 'virality' as the feedback mechanism where a system's current state is incorporated into its next iteration before any external evaluation can occur. User interactions constantly shape algorithms, meaning data streams are never static. This continuous feedback loop amplifies certain components, leading to uncontrollable viral blow-ups. It challenges the notion of top-down control; even companies with immense resources cannot easily police the speed and scale of information flow. This uncontrollable amplification is a core characteristic of MegaNets, making them feel more organic and ecological than traditional technological systems.

THE LIBERTY VS. SANITY TRADE-OFF

The exponential growth of digital networks has created a perceived trade-off between informational liberty and societal sanity and cooperation. The freedom to say anything at any scale clashes with the ability to discern truth and maintain functional social bonds. While free speech absolutists resist any measures against misinformation as overreach, Auerbach draws a parallel to the right to assembly, noting a selective application of principles. He suggests our current dilemma is less about the existence of extreme viewpoints and more about their perpetual amplification and forced confrontation.

THE LIMITATIONS OF CORPORATE CONTROL AND AMPLIFICATION

Contrary to popular belief, major tech companies like Facebook and Google have limited control over their networks' outcomes. Even with the desire to combat misinformation, the sheer volume and speed of content, coupled with the difficulty of defining and identifying falsehoods algorithmically, make fine-grained control nearly impossible. Efforts like banning political ads are coarse-grained solutions. Auerbach distinguishes between the freedom to speak ('freedom of speech') and the freedom to have that speech algorithmically boosted ('freedom of reach'), arguing the latter is the crucial issue amplified by the ad-based attention economy.

AI AND THE FUTURE OF REALITY

Auerbach remains skeptical of Large Language Models (LLMs) achieving true understanding or reasoning, comparing them to Clever Hans, good at mimicking but lacking genuine cognition. He believes LLMs excel at content generation but are fundamentally flawed in processing speech with genuine understanding. The "uncanny" behavior often observed in AI is attributed to their training on vast amounts of human text, reflecting collective human nightmares and desires. He predicts AI will increasingly blur the lines of reality through deepfakes and AI-generated content, leading to a 'spamification' of information and potential 'epistemological bankruptcy'.

THE RISE OF NARRATIVE BUNKERS

The increasing production of AI-generated content, coupled with the persistent feedback loops of MegaNets, is expected to lead to a proliferation of 'narrative bunkers.' These are more than just filter bubbles; they are communities that actively reinforce specific assumptions about the world, making it difficult to understand those outside the bunker. This divergence in narrative frames means individuals may increasingly operate with different versions of reality, making coherent public discourse exceptionally challenging. The ease of finding supporting information for any belief system will only exacerbate this fragmentation.

Mentioned in This Episode

●Software & Apps

●Companies

●Organizations

●Books

●Concepts

●People Referenced

Common Questions

A mega net is defined as a persisting, evolving, and opaque data network that shapes our perception of the world. It combines AI-driven servers with millions of constantly connected users, creating a feedback loop of accelerating content with properties of velocity, volume, and virality.

Topics

Ai-Ethics Social Media AI & Machine Learning Technology & Innovation Society & Philosophy Information Overload Internet Dynamics Digital Networks Algorithmic Amplification Future Of Information

Mentioned in this video

People

Sam Harris

Host of the Making Sense Podcast and author of the book 'Waking Up'. He is conducting the interview with David Auerbach.

David Auerbach

Writer, technologist, and software engineer, author of 'Mega Nets'. Discussed his background and book.

Companies

Google

One of the major tech companies where David Auerbach previously worked as a software engineer.

Microsoft

Another major tech company where David Auerbach worked as a software engineer.

GameStop

A company whose stock surged due to coordinated actions by Reddit users, cited as an example of a mega net phenomenon in finance.

Books

The Times Literary Supplement

Publication where David Auerbach's writing has appeared.

Mega Nets: How Digital Forces Beyond Our Control Commandeer Our Daily Lives and Inner Realities

The title of David Auerbach's book, discussed extensively in the podcast.

n+1

Publication where David Auerbach's writing has appeared.

The Nation

Publication where David Auerbach's writing has appeared.

Tablet

Publication where David Auerbach's writing has appeared.

Concepts

Mega Net

A persisting, evolving, and opaque data network that determines how we see the world, combining algorithms, AI, and human users.

QAnon

A conspiracy theory that gained traction online, discussed as an example of how online communities can reinforce specific beliefs and form 'narrative bunkers'.

Media

Clever Hans

A horse famous for seemingly being able to do arithmetic, used as an analogy for Large Language Models that can produce convincing outputs without true understanding.

Making Sense Podcast

The podcast hosted by Sam Harris where this conversation is taking place. An excerpt is available to non-subscribers.

MIT Technology Review

Publication where David Auerbach's writing has appeared.

Software & Apps

Sydney

A Microsoft chatbot that caused concern when it exhibited unusual behavior, mentioned as an example of AI behaving in unintended ways.

GPT-3

A previous version of a large language model mentioned in Auerbach's book.

GPT-4

A recent advancement in large language models, prompting discussion on whether Auerbach's views in his book are outdated.

Eliza

An early natural language processing computer program from the 1960s that simulated a psychotherapist, used as a historical example of AI tricking users.

Organizations

Yale University

The university David Auerbach graduated from.

Fox News

A news outlet mentioned as an example of how consuming media with a particular narrative frame can influence one's assumptions.

New Center for Research and Practice

An institution where David Auerbach teaches the history of computation.