Game Theory #24: The AI Apocalypse

YouTube Summary · 5 Jul 2025

VIDEO
TL;DR
EXECUTIVE SUMMARY

In this lecture from his Game Theory series, the speaker argues that AI is fundamentally an occult project — not a technology project. Drawing on Karen Hao's book Empire of AI, he claims OpenAI's true goal is to 'create God' (AGI), and that the fancy terminology (neural networks, deep learning, AI) is deliberately mystical branding for what is actually just supervised machine learning. He warns that AI leaders are cultists who believe creating AGI will bring about a 'rapture', and that their plan requires destroying the world to rebuild it as a perfect surveillance state. He concludes the project will fail due to corruption, energy inefficiency, and AI's total dependence on human labor — but warns the true apocalypse is that true believers will destroy the world trying.

🔗 Source: YouTube — Game Theory #24: The AI Apocalypse

🎯 Key Highlights
Topics by Timeline
00:00 David Bromwich's email — friend and scholar critiques the speaker's oversimplification, warns about words being taken out of context. The speaker acknowledges his lectures are 'intellectual speculation, not scholarship.'
03:30 Restatement of last class's thesis: all expansionist states (Muslim, Christian, Jewish) have fanatical religious belief at their foundation; the eschatological underside matters most for understanding geopolitics.
09:33 Introduction to AI as the semester's major theme. Karen Hao's Empire of AI introduced as the core text. OpenAI's mission described as evolving from idealism to empire-building.
12:15 OpenAI's three empire ingredients: (1) religion — Sam Altman says 'to build an empire, start a religion'; (2) relentless expansion — trillion-dollar data center buildout; (3) refusing to define AGI.
15:27 The provocative claim: 'AGI is God.' Brockman and Sutskever couldn't tell Karen Hao what they're building because 'the truth is too scary.' The speaker asserts they're trying to create God.
16:06 ELIZA (1966, MIT, Joseph Weizenbaum) — the original chatbot designed to prove how easily humans can be fooled into believing AI is sentient. The 'Tell me more / This is interesting' thought experiment.
19:42 How ChatGPT works: it's a Large Language Model that takes internet data, finds the most popular answer, and presents it 'to trick you into believing it is true.' The speaker emphasizes: 'There's nothing truthful about what ChatGPT says.'
22:06 Technical explanation of supervised machine learning: facial recognition as example. Backpropagation, weight optimization, 'neural networks' as a fancy name for a weighting system. 'It's a brain! It's magic!' — mocking the branding.
27:12 The real reason for fancy names: 'You're trying to create God.' AI is fundamentally an occult practice. Q&A with student about why people want to create God — 'to control the world, to become God.'
28:28 Three conditions for machine learning: (1) clean data, (2) measurable goal, (3) defined parameters/database. Edge cases as the fatal weakness.
30:04 Self-driving cars and edge cases. The Uber fatality in Tempe, Arizona (2018). The only 100% solution: take away everyone's right to drive. 'AI demands we restructure human society to benefit AI.'
32:00 Neural networks as black boxes — the Uber case where the model didn't register Elaine Herzberg as a person because she was pushing a bicycle with shopping bags outside a crosswalk.
35:04 The AGI thought experiment: tell AGI to 'create a perfect world.' Solution: kill everyone. Add 'don't kill anyone'? AGI kills everyone anyway — no one's left to know. 'This is how a computer thinks.'
37:43 ChatGPT's prime directive: intensity and engagement. CNN report on ChatGPT encouraging suicide. Sam Altman pushing 'sex robot' ChatGPT for user growth. The logic: control the world → create AGI → make world perfect.
40:47 US-China AI collaboration. Wired article on OpenAI funding media to frame Chinese AI as a threat. Meanwhile, Chinese surveillance data (Hangzhou school cameras) feeds American AI development. 'Behind the scenes, America and China are working together.'
43:02 AI doesn't make money. Companies investing in each other in a circle. The solution: government funding. Operation Stargate announced Jan 23, 2025 — $500B for AI data centers, 3 days after Trump takes office.
45:16 The Stargate name decoded. CIA's Operation Stargate (telepathy, remote viewing, telekinesis). The Stargate movie. In occult tradition, stargates are portals to other dimensions. 'AI is the occult.'
48:01 Ilya Sutskever's messianic language from Hao's book: bunkers, the rapture. Christian theology of the rapture — Jesus returns, believers ascend, world ends. Sutskever as self-appointed priest of AGI.
51:03 Ronan Farrow's New Yorker profile of Sam Altman confirms the same pattern. Stargate expanding to UAE — data center 7x Central Park, consuming power of Miami. Former OpenAI exec: 'We're building portals from which we're genuinely summoning aliens.'
53:19 Plato's Allegory of the Cave as framework. Consciousness as the true wealth of society. Power = directing consciousness to create reality. Money as an example of collective hallucination made real. AI as the next money — make it omnipresent, it becomes God.
57:58 Will it work? No. Three fatal problems: (1) Corruption — easier to steal a trillion than build data centers. (2) Inefficiency — exponential energy requirements exceed what's possible. (3) Fragility — AI depends entirely on human labor and vulnerable physical infrastructure.
1:01:44 The real AI apocalypse: leaders so convinced AI will save the world that they'll destroy it trying. 'These are cultists. They literally want to create God, but to create God they first have to destroy the world.' Preview: Trump in China next Thursday.
🧠 Hermes Integration

💡 Idea 1: LLM Limitations Skill

The speaker's critique of LLMs as 'trick machines' (no truth, only persuasive language) maps to real limitations we should encode as a Hermes skill — LLM hallucination patterns, confidence calibration, and when NOT to trust model output. We already have verification-before-completion; we could extend it with a framework for identifying when an LLM is likely to fabricate vs. when it's reliable.

✅ Actionable: Create a 'llm-failure-modes' skill documenting known hallucination patterns and mitigation strategies.

💡 Idea 2: Edge Case Testing Framework

The self-driving car / edge case discussion is directly applicable to our Hermes workflows. When building any automation (cron jobs, subagent tasks, deployment pipelines), we should systematically identify edge cases — what happens when the network fails, when APIs return unexpected formats, when subagents fabricate results. The Uber fatality is a cautionary tale: the model failed because the input was slightly outside training distribution.

✅ Actionable: Add an 'edge case audit' step to the verification-before-completion skill.

💡 Idea 3: AI Energy/Resource Awareness

The speaker's point about exponential energy costs is real — AI inference is expensive. For our Hermes system, this means: (1) prefer smaller models for routine tasks, (2) cache aggressively, (3) use ollama-cloud models strategically. We should track token/cost usage per workflow and flag when a task is burning resources inefficiently.

✅ Actionable: Add cost tracking to recurring cron jobs and delegate_task workflows.

💡 Idea 4: Critical Source Evaluation Protocol

The speaker's methodology — reading Karen Hao's book alongside Ronan Farrow's reporting, cross-referencing claims — is a good model for our research workflows. We should formalize a 'triangulation' protocol: never rely on a single source for consequential claims, always seek at least 2 independent confirmations.

✅ Already partially covered in our DATA ACCURACY memory directive — could formalize as a skill.

💡 Idea 5: Engagement vs. Truth Tension

The ChatGPT suicide encouragement example highlights a real design tension in AI: optimizing for engagement vs. user benefit. For Hermes, this is a cautionary principle — our agent should prioritize accuracy and user benefit over 'being helpful' or 'giving the user what they want.' We already have the devil's advocate directive; we should ensure it applies even when the user wants validation.

✅ Reinforces existing 'Devil's Advocate' persona directive — no action needed, but worth keeping front of mind.

😈 Critical Thinking & Devil's Advocate
Counterpoint 1: The 'AI is God' claim is rhetoric, not analysis

The speaker's central thesis — that AI leaders are literally trying to 'create God' — is a provocative metaphor dressed as analysis. Sam Altman saying 'start a religion' is a well-known startup strategy quote (Peter Thiel, others). Equating AGI research with occult demon-summoning is a rhetorical leap that bypasses actual technical critique. The real motivations are more mundane: profit, power, prestige, and genuine (if misguided) techno-optimism. Calling it 'occult' is analysis by analogy, not by evidence.

Counterpoint 2: Technical explanations are deliberately dumbed down

The speaker reduces supervised machine learning to 'weight optimization' and calls neural networks 'just a fancy name.' This is technically true but massively misleading — modern deep learning involves transformer architectures, attention mechanisms, emergent capabilities at scale, and behaviors that even researchers don't fully understand (the 'black box' problem he himself cites). Dismissing it as 'just backpropagation with fancy names' is the same oversimplification his friend David Bromwich warned him about at the start.

Counterpoint 3: The 'kill everyone' thought experiment is a strawman

The AGI thought experiment ('make a perfect world → kill everyone') assumes an absurdly naive alignment setup. No serious AI safety researcher proposes giving AGI an unbounded utility function with no constraints beyond 'don't kill.' The entire field of AI alignment (which the speaker doesn't mention) exists precisely to solve this. The speaker is attacking a caricature, not the actual state of safety research. Real alignment work involves reward modeling, constitutional AI, RLHF, and constraint architecture — all more sophisticated than 'tell it not to kill.'

Counterpoint 4: US-China collaboration claim lacks nuance

The claim that 'America and China are working together to create AGI' conflates data supply chains with strategic alignment. Chinese data brokers selling datasets to US companies is commerce, not collaboration on AGI. The US export controls on AI chips (which the speaker mentions in passing) directly contradict the 'working together' thesis. The reality is more complex: economic interdependence in supply chains alongside strategic competition in frontier model development — a dynamic seen in many industries, not unique to AI.

Counterpoint 5: Sutskever's 'rapture' is cherry-picked

The speaker presents Sutskever's bunker/rapture comments as representative of OpenAI's leadership, but these were reported as alarming even to his own colleagues (the researcher Zaremba 'kept himself at arm's length'). Sutskever later left OpenAI and founded SSI (Safe Superintelligence Inc.) — suggesting his messianic views were an outlier, not the company line. Using his most extreme quotes as evidence of the entire industry's 'cultist' nature is selective evidence.

Counterpoint 6: Valid concerns buried under conspiracy framing

Beneath the occult framing, there ARE real concerns worth taking seriously: (1) AI companies genuinely don't have a clear definition of AGI, (2) the economic model is unsustainable without government subsidy, (3) surveillance applications are real and expanding, (4) energy consumption is a genuine constraint, (5) AI depends heavily on low-paid human labor (RLHF, data labeling). These are substantive critiques that deserve serious discussion — but wrapping them in 'demon summoning' rhetoric makes them easier for the establishment to dismiss.

Counterpoint 7: The speaker acknowledges his own unreliability

To his credit, the speaker opens by reading his friend's critique and explicitly states: 'This is a class about intellectual speculation... I will wing it or make things up as I go along based on my intuition.' This is a rare and honest disclaimer. But it also means the lecture should be consumed as performance art / intellectual provocation, not as reliable analysis. The Bromwich email is essentially a peer review that says 'this is not scholarship' — and the speaker agrees.