AI & the Future
Every idea in Doomscholar on ai & the future. Open any one to read it in layers, with its sources.
- AGI Ruin: A List of LethalitiesEliezer YudkowskyAligning superintelligent AI is unsolved and likely fatal by default; current efforts are inadequate.
- AlphaFold and the End of the Protein Folding ProblemJohn Jumper et al. (DeepMind)For fifty years, going from a protein's amino-acid sequence to its three-dimensional shape took months or years of experiment per protein. In 2020 a neural network did it in minutes at near-experimental accuracy, and two years later the predicted structure of nearly every catalogued protein was free to download.
- Apple, SaaS, and the 18-Month WindowAttributed to Naval Ravikant (via secondary commentary)A secondary commentary claims Naval Ravikant argued on a podcast that pure software has become uninvestable as AI commoditizes the interface layer — and that SaaS founders have roughly 18 months to build a defensible moat (distribution, network effects, data flywheels, or hardware) before their valuations compress.
- Atmospheric Anomalies and Technosignatures on TRAPPIST-1eScience TimeSpectral anomalies detected on the tidally locked exoplanet TRAPPIST-1e defy standard atmospheric models, raising the remote but real possibility of non-natural chemistry indicative of a Kardashev Type I civilization.
- Attention Is All You NeedVaswani et al. (Google)The Transformer architecture, based solely on attention, outperforms recurrence/convolution for sequence modeling.
- Big IdeasARK Invest (Cathie Wood)Disruptive innovation platforms (AI, robotics, genomics, crypto, energy storage) will compound explosively.
- Block Goes Intelligence-NativeJack DorseyA much smaller team paired with AI can do more and do it better, so Block cut ~40% of staff to rebuild 'intelligence-native.'
- Cosmological Cannibalism and Baby UniversesScience TimeThe rapid expansion of the universe — traditionally explained by cosmic inflation and dark energy — can alternatively be modeled as our universe systematically colliding with and absorbing smaller "baby universes" within a larger multiverse framework.
- Dark Forest TheoryLiu CixinThe universe is full of life, but everyone is hiding in absolute silence because revealing your location guarantees your destruction.
- Deep Residual Learning for Image Recognition (ResNet)He, Zhang, Ren, Sun (Microsoft)Residual connections let networks train to unprecedented depths, driving a leap in vision performance.
- DeepSeek-R1: Incentivizing Reasoning Capability via RLDeepSeekReinforcement learning can elicit strong reasoning in LLMs at a fraction of frontier training cost.
- Efficient Estimation of Word Representations (word2vec)Mikolov et al. (Google)Words can be embedded in vector space where semantic relationships become linear arithmetic.
- ImageNet Classification with Deep Convolutional Neural Networks (AlexNet)Krizhevsky, Sutskever, HintonDeep CNNs trained on GPUs shatter the ImageNet benchmark, igniting the modern deep-learning era.
- Internet Trends ReportMary MeekerAn annual data-dense synthesis of global internet adoption, usage, and monetization trends.
- Language Models are Few-Shot Learners (GPT-3)Brown et al. (OpenAI)Scaling language models to 175B parameters yields emergent few-shot task performance without fine-tuning.
- Machines of Loving GraceDario AmodeiPowerful AI could compress a century of scientific progress into a decade, curing disease and transforming society.
- Meditations on MolochScott AlexanderMultipolar traps and coordination failures ('Moloch') drive races to the bottom across civilization.
- Mobile Is Eating the WorldBenedict EvansMobile is the largest computing platform ever and is subsuming every other technology and industry.
- Moore's Law for EverythingSam AltmanAI will drive the cost of goods and labor toward zero, requiring a new social contract to distribute the wealth.
- OpenCloud ReportBattery VenturesOpen-source and cloud infrastructure are reshaping enterprise software economics and buying.
- Prime Editing and the Reprogrammable Human GenomeScience TimeCRISPR 2.0 (Prime Editing) elevates genetic modification from a crude cutting tool to a precise word processor, unlocking the capability to cure most genetic diseases and propelling humanity into an era of directed, non-Darwinian evolution.
- Scale-Invariant Unparticles and Cosmic ExpansionScience TimeThe accelerating expansion of the universe, traditionally attributed to the Cosmological Constant, may be driven by "unparticles" — a theoretical form of scale-invariant matter that behaves as a macroscopic fluid defying standard particle physics.
- Scaling Laws for Neural Language ModelsKaplan et al. (OpenAI)Model performance scales predictably as a power law with compute, data, and parameters.
- Situational Awareness: The Decade AheadLeopold AschenbrennerAGI by ~2027 and superintelligence shortly after will trigger a trillion-dollar compute buildout and a US-China race.
- Software 2.0Andrej KarpathyNeural networks are a new software paradigm where code is learned from data, not written by hand.
- Spacetime as an Emergent PropertyScience TimeSpacetime and gravity may not be fundamental components of the universe, but macroscopic illusions emerging from the quantum entanglement of underlying, non-spatial data points — akin to temperature emerging from molecular motion.
- Superintelligence: Paths, Dangers, StrategiesNick BostromA machine superintelligence could rapidly become uncontrollable and poses an existential risk to humanity.
- The Bitter LessonRich SuttonGeneral methods that leverage computation ultimately beat human-knowledge-engineered approaches in AI.
- The Intelligence Explosion and ASI AlignmentScience TimeThe transition from Artificial General Intelligence (AGI) to Artificial Superintelligence (ASI) will likely occur as an instantaneous, recursive self-improvement loop, making the alignment of machine ethics the ultimate bottleneck for human survival.
- The Kardashev ScaleNikolai KardashevA civilization's technological advancement can be measured purely by the amount of energy it is able to harness and utilize.
- The Most Important CenturyHolden KarnofskyTransformative AI could make this the most pivotal century in human history, warranting outsized attention.
- The Next Big Thing Will Start Out Looking Like a ToyChris DixonDisruptive innovations are dismissed as toys because they underperform on established metrics while excelling on new ones.
- The Pelican Test: Why Every AI Benchmark Eventually DiesSimon WillisonA benchmark that becomes famous enough to matter also becomes a target — and once labs and models start optimizing (deliberately or not) for whatever the benchmark measures, its correlation with real capability quietly decays, even while the benchmark itself keeps getting cited.
- The Scaling HypothesisGwern BranwenIntelligence may emerge from scaling up compute and data. GPT-3 shows the hypothesis holds.
- The Sovereign IndividualJames Dale Davidson & Lord William Rees-MoggThe Information Age will dissolve the nation-state's power and empower mobile, digitally-sovereign individuals.
- The Techno-Optimist ManifestoMarc AndreessenTechnology and markets are the engines of human flourishing; stagnation and de-growth are the real enemies.
- Training Compute-Optimal Large Language Models (Chinchilla)Hoffmann et al. (DeepMind)For a fixed compute budget, models should be smaller and trained on far more data than prior practice.
- Trends – Artificial IntelligenceMary Meeker / BONDAI adoption and capex are scaling faster than any prior technology wave in history.
- Turing Machines & the Halting ProblemAlan TuringTuring proved that no general algorithm can decide in advance whether an arbitrary program will ever halt — some questions are not just hard, they are uncomputable, which sets a hard limit on what automation can ever guarantee.
- Two Ways to Win the Robot Race: Smarter vs. CheaperTrendForceThe US and China are running two structurally different playbooks for humanoid robots — the US is betting that intelligence (AI models trained on real-world data flywheels) will be the deciding moat, while China is betting that supply-chain scale and manufacturing speed will out-iterate any intelligence gap, echoing how China caught up in electric vehicles.
- We Need a New Science of ProgressPatrick Collison & Tyler CowenWe should study the causes of scientific and economic progress as a rigorous discipline to accelerate it.
- What Happened to the Future?Founders Fund (Bruce Gibney)Venture abandoned hard technology for trivial consumer apps; 'we wanted flying cars, instead we got 140 characters.'
- What We Owe the FutureWilliam MacAskillFuture people matter morally; positively influencing the long-term future is a key moral priority (longtermism).
- Why AI Will Save the WorldMarc AndreessenAI will augment human intelligence and improve nearly everything it touches, and fears of civilizational risk are misplaced.
- Why Software Is Eating the WorldMarc AndreessenEvery company is becoming a software company; software firms will disrupt incumbents across every industry.
- Zero to OnePeter Thiel & Blake MastersTrue progress is vertical (0 to 1) via monopoly-creating innovation, not horizontal copying (1 to n).
Keep these, don't just read them
The app brings each idea back just before you'd forget it, with a quiet daily reminder.