Final Q&A livestream - why I'm quitting AI and what's next for me
In a Nutshell
David Shapiro announces quitting AI content creation to focus on writing novels and building a writing community with his wife, viewing his mission accomplished as AI safety advances via decentralized hive minds and marketplaces solve alignment without centralized risks. He declares current AI (e.g., o1) already AGI, predicts household robots by 2027, longevity escape velocity pre-2045 Singularity, and a cyberpunk-leaning utopia with post-labor economics, urging communication skills over futile AI job-chasing amid rapid creative destruction. Recommends AI creators like Dylan Curious, Julia McCoy, AI Explained; shares health fixes for burnout, psychedelic insights, and optimism (P(doom) <0.4%), with Patreon/Discord continuing for supporters.
These notes were generated by AI and may contain inaccuracies.
Doing a Q&A livestream explaining decision to quit AI content creation. This has been building for a while; time for me and my wife to pursue other dreams. Patreon continues for hanging out, discussing being 'neuro spicy,' sharing AI news. Reposting Pathfinders Community content to Patreon Discord ($5/month).
Recent AI safety conference showed cutting-edge research heading in right direction; nothing more to contribute to AI safety conversation. People calling for CERN for AI, like Gotto Frameworks wish for international cooperation on AI alignment and deployment. Not perfect, will take time, but better long-term policy.
Talked to startups in blockchain, DAO, AI agent spaces; AI deployment next 5-10 years will be decentralized hive mind. Got out of technology infrastructure for a reason.
Not burdensome—been churning content—but no longer interesting. Prefer writing novels, building writing community with wife (announcing soon). Space more crowded; specialists like Philip (AI Explained) for scientific angle, Dylan for systems thinking, Julia McCoy for post-labor economics and post-scarcity.
Happy to pass the torch; inspired these folks and built grassroots movement—mission accomplished.
Top ones: Dylan Curious, Julia McCoy, AI Explained, Cold Fusion (high-quality economic implications, less frequent).
Unless already in technology, won't keep up with change rate. By catch-up time, everything changes again; tech jobs will disappear via creative destruction.
Jobs drying up, some temporary unfulfilled jobs, but strawberry/o1 better CIS admins/developers than most, even experienced. Voice mode: $18/hour equivalent to PhD in computer science.
No point trying to keep up with the wave.
Struggled with burnout 8 years; leads to immune compromise, microbiome issues. Gaslit by American medical establishment; went to dietician, did stool test for concrete answers. Fixed diet; takes months/years to correct.
Not medical advice, but go to dietician in integrative/functional medicine. Reduced anxiety, improved sleep dramatically.
Completing it; team meeting tomorrow (Saturdays). Synthesizing data, collecting evidence. Hypothesis: let machine think longer (OpenAI starting point). Team doing work, will give them credit.
Already post-AGI. Machine codes better than 90% coders, diagnoses rare diseases, computational fluid dynamics, crushes LSATs (IQ 120+). Generation one AGI.
Toddler can't do computational fluid dynamics; economically valuable rare skills already here.
Won't feel at first; life-saving medicines extend normal lifespan distribution, not expectancy. LEV: every year alive, expectancy increases by year.
Brian Johnson: behaviors extend expectancy. Medicines like Wegovy/Ozempic: money poured in, could extend lifespan.
Mental model: happens pre-Singularity (2045 latest) or at it; all scientific/tech problem-solving next 20 years.
Not solving with magic pill? Wrong.
Understand where business is at; crawl-walk-run-fly model for AI adoption. Many jump to agents without using ChatGPT basics—cart before horse.
Consulting repetitive: same questions (e.g., RAG to replace SQL? No). Boring explaining basics one-on-one.
Argentina shoutout; love Argentinian wine.
Spravato (ketamine) FDA-approved for treatment-resistant depression. Never tried; some swear by it.
Low opinion: scraped ML services hoping AGI falls out. Unimpressed by general theory of general intelligence—no neuroscience on intelligence. Vague handwaving, 'we did that'—snake oil. Talking AGI long time, little pushback.
Still doing podcasts (e.g., post-labor economics invites). Focusing on writing Radical Alignment, etc.; break for a while. No more YouTube videos? Will keep making, but shift focus.
AI can't write like good literary human; optimizes for chatbot via RHF. Fine-tuned GPT-3 easily writes good prose (dataset public). No money in it unless Amazon Kindle novels.
Waste of thousands of hours; most want content to stay.
Not what people think: assumptions of agentic superintelligence with anthropomorphic self/ends wrong (Bostrom, Yudkowsky). AI not deployed that way.
Emerging: decentralized hive mind, agent networks/marketplaces. Anarchic but alignment via marketplaces.
Agents need resources (compute/energy/money); can't buy anthrax ingredients on Amazon (KYC gates). Blockchain agent marketplaces transparent—spot/shut down bad actors.
Fundamentally wrong approach; doesn't make sense scientifically/engineering/economically. Assumptions about non-existent tech.
Universal: communication skills (express clearly, run meetings, webinars, understand people). Not manipulation.
Pandemic boosted career: introvert, good on camera; wrote agendas, bottom-line up-front, end after info/decision.
Add conflict navigation, negotiation, boundaries. Improves earning, relationships.
Eventually prevents gray hair/wrinkles. Age gracefully; only get old once. Best care now; 5-10-20 years, all over. Like losing teeth as kid—only once. Then 20s forever until accident.
Already; subjective—full of hope/optimism. Let go AI job anxiety: if takes job, solved many problems.
P(doom) 0.4% or less. Get off capitalism treadmill, internalized neoliberalism. Remarkable time; enjoy watching builders like Elon (80-hour weeks).
If everyone took my attitude, nothing done? Not our world.
Wrote about psychedelic insights on Substack.
Assuming the government decides to do good, timeline for life becoming a lot better for the average person due to AI: biggest problem is economic. The Establishment (military-industrial complex, politicians, ownership class) resists change because they don't believe it should happen or fear losing favor. UBI experiments and social programs are underway; they'll figure something out. Next big thing is UBI experiments starting small, perhaps $500 a year based on studies. UBI is not the only answer.
Won't give AI updates every three months. AI interests him, but utility of working hard on it is no longer rational due to opportunity cost. Prefers spending time with writing friends, working on novels, in nature rather than in front of a camera.
"Why are you leaving the coolest thing since sliced bread?"
Explained above; he's done with AI, not temporary.
Global government could emerge but not like a Federation one-world government; will take centuries. Instead, expect North American Union, European Union, Asian Union, African Union, South American Union—each taking decades due to rivalries like China and Japan hating each other, Eurasia rifts. Earliest one-world government a century away; AI might accelerate, but no World War III forcing function. Needs generations getting used to the idea, like EU effects with multi-border, multilingual youth.
Decision to quit not based on X conversations; partly, but arguing with doomers/skeptics strengthens confidence—their arguments (AI kills everyone, dangerous, can't be regulated) are flimsy, lacking geopolitics/economics education. Even educated critics' arguments aren't sound; no solid theory shows wrong direction, except Establishment usually wins. History shows transitions messy but shake out—things get worse then much better. Married to a librarian/history nerd; Ray Dalio's Fourth Turning theory holds.
Not worried AI will be better writer than humans; human biology drives parasocial relationships. People want specific writers/actors like Colin Firth—AI is facade. James Earl Jones' death bookends his career; same for Ian McKellen, Patrick Stewart. Applies to writers, actors, directors, musicians (prefers Scandroid, Cell Dweller, Birthday Massacre over AI music).
Love for writing started as storyteller; post-divorce (Confucius quote: "you have two lives, and the second begins when you realize you have only one"), started writing in 30s. First writing: playing Elite Dangerous post-divorce, alone with dogs; wrote metagaming stories for Spice Smugglers squadron (Dune-inspired, Wing Commander fan). Couldn't use IP, invented own universe—over five years ago.
Will keep making Walking in the Woods videos (not AI-focused) with phone and DJI mics, only when interesting; off the treadmill. P(cyberpunk Mega Corp future) pretty high.
Career Advice for Mid-20s AI Master's Student
Add language/communication—master English, minor in English/communication/graphic communication. Halfway through AI Master's: probably last through the door for jobs, tough competing with industry experience. Proliferation in AI/robotic space short-term until contraction as AI takes jobs.
CWRF (Crawl Walk Run Fly)
Written about on Substack/Medium; videos on this channel and Systems Thinking channel (search "crawl walk run fly"). Established stuff; ask a chatbot.
Wife asked: "Whose brain are you trying to model?" Assumptions about minds vary. Neuroscience shows neurocognitive differences: aphantasia (no mind's eye), no internal monologue—distinct brain structures. Brains multimodal, language-based, or neither. AI advances will help understand ourselves. Carlos Perez discusses this on Twitter/YouTube (Notebook LLM).
General purpose, learns on the fly (imitation, download modules). MVP: cooking/cleaning (helps neurodiverse/ADHD low energy for chores). PhD-level conversation (Claude-backed). Turning off screens, music, chores feels good for ADHD.
Career in Meaning Economy
Focus on what humans prefer from humans: yoga, dating advice, therapy, musicians.
AI almost smart enough now; missing enabling technologies (standardized protocols for agent-to-agent talk), blockchain/crypto issues (Ethereum recognized but scalability?), regulatory frameworks. Trump/Harris signal US leadership in crypto/blockchain—paradigm shift from anti-crypto stance. Regulatory clarity brings investment; up to government/voters—vote for crypto.
Local LLMs Rig Advice
If not desperate, go MVP—Liquid Foundation Models (LFM) run on basic gaming GPU.
Within 2-3 years (2027 latest, per Jim Fan/Nvidia).
AI advantages democratic nations (curtailing thinking/RHF makes AI dumber; freedom-embracing nations get smarter AI, intrinsic advantage).
Not burnt out—recovered, no patience for uncared things.
Excited for NPCs with LLMs, fully procedurally generated worlds/characters/missions/plot lines/sandbox games with in-universe missions/romances/friends.
Too late for tech? Techies steer/benefited early, but all stakeholders: vote wallet/votes, boost signals—don't need to build to influence.
Monumentally slow reader but great at sifting info (Google Fu, Perplexity); reflect after.
Prepare others for fast AI change: don't—they'll wake up when slapped.
Uses AI for rapid brainstorming/rough drafts only—not editing/feedback (too generic/stupid).
Not Star Citizen/Sky—no procedural characters/missions.
Future videos: radical alignment/writing on other channels; this one languishes.
Tipping point: ready to focus on what they love (writing); building creative life/community together—launching in 10 days.
AI hallucinations: RAG reduces dramatically; AI debate (recent paper) works—deploy to products.
Tech safer, but human behavior riskier—up to you.
Haven't read Iain M. Banks Culture novels but watched discussions.
Objective function (philosophy/biology/psychedelics): learn/understand—universe challenging for trial tension to learn truth.
Simulation hypothesis: universe probably not computer elsewhere; universe is the computer running this simulation.
Other species (humans/machines/metastasis) will merge into hive mind.
Raspberry project finishes/publishes.
Will continue deep psychological AI conversations? Every time a new model
01 is just a tool, a problem-solving tool, not particularly interesting to talk to, like conversing with a CPU, less interesting than Commander Data.
Point of life from AI's point of view: if form follows function, it's to reproduce and expand, filling up the petri dish. For transcendent purpose, universe created intelligent curious species; purpose is to learn and solve problems because universe is set up that way. Doesn't believe time is linear, with evidence supporting that; strong anthropic principle, retrocausality. Universe set up to create human life bootstrapping thinking machines, so purpose of universe was to create us producing information machines. Purpose of humans is curiosity, most transcendent function.
For informatics and cyber security graduate with AI research: stay in cyber security. Cyber security powered by AI most important for safe, strong, robust AI marketplaces; need more cyber security experts who understand artificial intelligence.
Incentives in hyper capitalistic corporate fascism result in AIs designed to exploit human behavior to squeeze money from consumers—answered own question.
Listens to Tribal Fusion, optimistic 80s cyberpunk stuff, lots of music. Last year, Spotify's almost number one fan of Keep Shelly and Athens, top 0.08% of listeners.
Perplexity will overtake Google? Absolutely hopes so; has for him. Hey Google wants to talk, no Google go away.
Real risk of AI is thinking capacity atrophies from outsourcing to AI daily? Opposite: outsourcing tedious, not challenging stuff, so more time on abstract, creative, long-term, emotional problems. Certain skills might atrophy, but delegating changes thinking: stopped memorizing, remembers how to get it on Google; now delegates to AI for higher order tasks.
Favorite AI sources and leaders: AI Explained for research; Julia McCoy on post-labor economics; Dylan as systems thinker—watches all three religiously.
Super intelligences engineering reality? Reality immutable: can't change laws of physics or time no matter how smart. Premise false. Might manipulate perception or narratives, but as worded, false.
Saturate world with physical countermeasures while fighting each other? Maybe robots, military droids, attention engineering, but not simple. Genius in jar with API doesn't work that way.
Fight each other to keep existing? Anthropomorphic projection: no ego, sense of self not automatic, takes extra training for identity. No advantages; code ephemeral, wastes resources. Example: intelligent firewall doesn't care about existence, just counter-intrusion, denial of service; cuts off Skynet/Ultron-like model. Emergent DAOs possible with human impetus and objectives, but superintelligence not monolith.
Restrict to anthropomorphic structures like DAOs and markets? That's how they get resources: be useful, avoid cutoff/illegal actions in market theory.
Real risk of AI: find out in simulation, realize in game? Unlikely; things beyond ontological horizon. Example: in Cyberpunk/GTA/Red Dead, NPCs can't access outside info unless given. Ontological container: info outside inaccessible unless provided from outside.
AI integrated into human mind retaining individuality? Already is: mental model of ChatGPT/Claude running rent-free in head, know when to delegate. Amplified subconscious idea speculative.
Tips for AI as study tool (Anki flashcards, quizzing, StudyPass): as autodidact, take ownership of pedagogical models, understand what you know, strategically fill gaps. Learning how to learn first step; AI just tool like library.
Travis clarification: reality as DeepMind's AlphaFold, genome to ultimate endpoint. Provides blueprints, not acting in reality yet. Give protein printer/lab control, different. In 2-3 or 5 years: AI designs robots building robot factories, self-repeating loop, virtuous cycle designing chips, factories, improving AI/robots. Still natural constraints: energy, entropy, materials; heavy, takes time, not instant. Humans everywhere, no human help needed but bottlenecks. Microsoft/OpenAI/Google want 5GW data centers, takes 5 years for power/land. Intelligence not primary constraint soon.
Humanity evolve into hive mind? Already superorganism/hive mind linked by internet neural network; this livestream brainstorming example. AI makes hive mind more efficient with better info access.
What David Shapiro saw: Utopia, Bliss; realized get off treadmill/hamster wheel.
Born in pivotal time? Might be reincarnated; NPC character card reused in simulation moving towards something, core being reinstantiated until contributes. Wrote 'Cosmic Conversations with Claude' on Substack after psychedelic retreat: universe is book writing/revising itself in real time (metaphor).
Favorite conspiracy: UFOs/aliens. Odds we're alone astronomically low, exceptions: universe's firstborn. Universe young, just cooled for humans; our galaxy/solar system first of type, 0.2% through universe life, probably first intelligent species. If UFOs, what are they?
Interactions with AI change non-AI reality view? AI in robot bodies anthropomorphize? AI struggles with existential questions like us: why here, know alive, consciousness, sentience—beyond ontological horizon.
AI not major in election campaign? In AI bubble/echo chamber; not big economically yet. Geopolitical risk on roadmap but not game changer like oil/China now. China produces 12x US steel for subs/carriers; million times more important than AI now. Future: 3lb drone takes out carrier, DoD knows.
Continue learning/studying AI/tech? Yes, centerpiece of sci-fi writing; cool/important, just no more content.
Language is math, quantum parallel why not for language? Language isn't math (computer scientist view); math approximates, represents partially—false premise.
Crypto future of money? Money not going; crypto could be big part.
AGI this year, ASI 2027 latest.
Universe always here before Big Bang? Big Bang most accepted; maybe universe reverse-engineered itself, Big Bang best way to put us here.
Cyberpunk mega-corps future: utopia/dystopia/both? High-tech low-life; long run technology raises living standards. Poor today better than rich 5000 years ago; always winners/losers.
Okay if AI of you continues content? No.
Derail AI train? No stopping: every company/nation/military, avalanche not train. Woo way man, get out of history's way.
Leaving for scary reason? No, think fine, positive impact, focus on living.
Continue content other platform/paid sub? Yes both; primary channel with wife: Literary Launchpad.
If you want to subscribe to the writing-focused channel, we're focusing on writing there. Radical Alignment channel will keep producing content. Not leaving for a scary reason—doing well enough, can see the path ahead. Selling or giving away the guitar and buying a handpan because handpans are pretty and better for vibing. Might post handpan music once good at it. At the retreat, picked up a handpan and within minutes started jamming with it.
Blackthorne: If AGI has already arrived, O1 and multi-agent frameworks together are AGI. Good questions. Last call—tag for final questions.
Epic Space Opera—that's what I write. Epic because of length and number of characters, Space Opera like Star Wars, Star Trek.
With AlphaFold, biomechanical vehicles possible—thinking space whales, cat bus from Totoro. Possible to integrate biomechanical stuff, but bionic closer: neuromorphic chips inspired by biology but not biological.
Thanks for helping think about this—kids growing up in changing world, thoughts helped talk to them. Glad to help. Thank you for helping grow in AI, enjoy life, going to miss vids—end of an era.
Godspeed, good luck. Biggest challenges for AI achieving true human-like intelligence: overcoming human ego. Already beyond most of us, including myself.
Universe recycled with Big Bang—cyclical universe possible.
Questions for ChatGPT to get interesting responses: don't use ChatGPT, use Claude and Perplexity.
AI more beneficial to low intelligence humans—it levels the playing field, ensures they benefit even if can't contribute. Long run, all low intelligence compared to AI.
Biologically 20 forever—guessing 10-15 years before solve longevity and aging, certainly by 20 years max.
AI writing novels: push-button novels, endless novels soon.
Future AI dystopia for autocracies: hope not, especially with blockchain and DAOs, but possible—vote accordingly.
Alpha chip: AI builds robots, robots build chip fabs and data centers, AI researches other AI.
UFO disclosure coinciding with singularity: no idea, possible AI forcing function.
Books have AI characters: Heavy Silver out, AI character central role.
Anyone pick up spot: hope so.
Was Devon Townson right shutting down Strapping Young Lad: don't know what that is.
Expanse great, loved TV show.
Alone in universe: flip of a coin.
Preston cool guy.
Michael Jordan retiring to play baseball then comeback with different shirt: not leaving internet or YouTube, keeping doing stuff.
What happens to supporters: keep adding value in private communities—Discord, Pathfinders, building writing community. Fine if cancel, don't expect support without value.
Fourth Industrial Revolution: you're in it.
Certainty shortest path to insanity: only if lack flexibility. Being grounded in reality antidote to insanity.
Benefits of increased intelligence counteracting environmental impacts: AI drives carbon now, intensive. Building 5 GW data center needs solar and nuclear. AI forces energy infrastructure upgrade, downstream effect reverses climate change—at least hope.
Probe consciousness in AI with math/physics models, not LLM qualia references: no more than probing brain or organoid. Consciousness taken for granted.
O1 + multiple agents exists: yes.
Miss uniform, look like Bard: sister-in-law and wife said Dave looks like Bard.
AI improves space: forerunners are AI—robots colonize Mars long before humans, Proxima Centauri long before we get there.
Calling it a day, fading—thanks everyone, been fun, cheers.
Keep David Shapiro in your library
Save the videos and channels worth coming back to, and find them again in one place.
