carado.moe https://carado.moe/ carado.moe posts Thu, 13 Apr 2023 22:00:00 -0000 Orthogonal: A new agent foundations alignment organization https://orxl.org/announcement.html https://orxl.org/announcement.html Mon, 27 Mar 2023 22:00:00 -0000 🔒 (2023-03-28) https://carado.moe/b6ns-iyxx-di91-p3xx.html https://carado.moe/b6ns-iyxx-di91-p3xx.html Wed, 22 Mar 2023 23:00:00 -0000 continue working on hard alignment! don't give up! https://carado.moe/continue-working-hard-alignment.html https://carado.moe/continue-working-hard-alignment.html Sun, 19 Mar 2023 23:00:00 -0000 the QACI alignment plan: table of contents https://carado.moe/qaci.html https://carado.moe/qaci.html Sat, 18 Mar 2023 23:00:00 -0000 you can't simulate the universe from the beginning? https://carado.moe/cant-simulate-the-universe.html https://carado.moe/cant-simulate-the-universe.html Sat, 18 Mar 2023 23:00:00 -0000 QACI blob location: an issue with firstness https://carado.moe/blob-quantum-issue.html https://carado.moe/blob-quantum-issue.html Sun, 12 Mar 2023 23:00:00 -0000 your terminal values are complex and not objective https://carado.moe/values-complex-not-objective.html https://carado.moe/values-complex-not-objective.html Sun, 12 Mar 2023 23:00:00 -0000 🔒 formal goal maximizing AI https://carado.moe/formal-goal-maximizing-ai.html https://carado.moe/formal-goal-maximizing-ai.html Sat, 11 Mar 2023 23:00:00 -0000 the quantum amplitude argument against ethics deduplication https://carado.moe/quantum-amplitude-deduplication.html https://carado.moe/quantum-amplitude-deduplication.html Fri, 10 Mar 2023 23:00:00 -0000 problems for formal alignment https://carado.moe/formal-alignment-problems.html https://carado.moe/formal-alignment-problems.html Wed, 08 Mar 2023 23:00:00 -0000 QACI blobs and interval illustrated https://carado.moe/qaci-blobs-interval-illustrated.html https://carado.moe/qaci-blobs-interval-illustrated.html Tue, 07 Mar 2023 23:00:00 -0000 QACI blob location: no causality & answer signature https://carado.moe/blob-location.html https://carado.moe/blob-location.html Sun, 05 Mar 2023 23:00:00 -0000 before the sharp left turn: what wins first? https://carado.moe/sharp-left-turn-what-wins-first.html https://carado.moe/sharp-left-turn-what-wins-first.html Sat, 04 Mar 2023 23:00:00 -0000 QACI: the problem of blob location, causality, and counterfactuals https://carado.moe/blob-causality.html https://carado.moe/blob-causality.html Thu, 02 Mar 2023 23:00:00 -0000 state of my alignment research, and what needs work https://carado.moe/state-research-agenda.html https://carado.moe/state-research-agenda.html Tue, 28 Feb 2023 23:00:00 -0000 on strong/general coherent agents https://carado.moe/strongly-generally-coherent-agents.html https://carado.moe/strongly-generally-coherent-agents.html Sat, 25 Feb 2023 23:00:00 -0000 some thoughts about terminal alignment https://carado.moe/terminal-alignment-solutions.html https://carado.moe/terminal-alignment-solutions.html Fri, 24 Feb 2023 23:00:00 -0000 clarifying formal alignment implementation https://carado.moe/clarifying-formal-alignment-implementation.html https://carado.moe/clarifying-formal-alignment-implementation.html Wed, 22 Feb 2023 23:00:00 -0000 Hello, Elua. https://carado.moe/hello-elua.html https://carado.moe/hello-elua.html Tue, 21 Feb 2023 23:00:00 -0000 🔒 RSI, LLM, AGI, DSA, imo https://carado.moe/2mde-19zj-filq-zutc.html https://carado.moe/2mde-19zj-filq-zutc.html Tue, 14 Feb 2023 23:00:00 -0000 a narrative explanation of the QACI alignment plan https://carado.moe/narrative-explanation-qaci.html https://carado.moe/narrative-explanation-qaci.html Sun, 12 Feb 2023 23:00:00 -0000 is intelligence program inversion? https://carado.moe/is-intelligence-program-inversion.html https://carado.moe/is-intelligence-program-inversion.html Fri, 10 Feb 2023 23:00:00 -0000 GPT is dangerous because it is useful at all https://carado.moe/gpt-dangerous-useful.html https://carado.moe/gpt-dangerous-useful.html Fri, 10 Feb 2023 23:00:00 -0000 my takeoff speeds? depends how you define that https://carado.moe/takeoff-speeds-define.html https://carado.moe/takeoff-speeds-define.html Wed, 08 Feb 2023 23:00:00 -0000 CEV can be coherent enough https://carado.moe/cev-coherent-enough.html https://carado.moe/cev-coherent-enough.html Mon, 06 Feb 2023 23:00:00 -0000 so you think you're not qualified to do technical alignment research? https://carado.moe/so-you-think-not-qualified-alignment.html https://carado.moe/so-you-think-not-qualified-alignment.html Mon, 06 Feb 2023 23:00:00 -0000 tabooing "AGI" https://carado.moe/tabooing-agi.html https://carado.moe/tabooing-agi.html Sat, 28 Jan 2023 23:00:00 -0000 communicating with successful alignment timelines https://carado.moe/communicating-successful-alignment.html https://carado.moe/communicating-successful-alignment.html Sat, 28 Jan 2023 23:00:00 -0000 formal alignment: what it is, and some proposals https://carado.moe/formal-alignment.html https://carado.moe/formal-alignment.html Tue, 27 Dec 2022 23:00:00 -0000 making decisions as our approximately simulated selves https://carado.moe/approximate-decisions.html https://carado.moe/approximate-decisions.html Wed, 21 Dec 2022 23:00:00 -0000 being only polynomial capabilities away from alignment: what a great problem to have that would be! https://carado.moe/capabilities-away-great-problem.html https://carado.moe/capabilities-away-great-problem.html Wed, 21 Dec 2022 23:00:00 -0000 one-shot AI, delegating embedded agency and decision theory, and one-shot QACI https://carado.moe/delegated-embedded-agency-decision-theory.html https://carado.moe/delegated-embedded-agency-decision-theory.html Tue, 20 Dec 2022 23:00:00 -0000 the scarcity of moral patient involvement https://carado.moe/scarce-moral-patient-involvement.html https://carado.moe/scarce-moral-patient-involvement.html Wed, 14 Dec 2022 23:00:00 -0000 🔒 more tractable formal alignment https://carado.moe/att8-1i1k-xk4r-itim.html https://carado.moe/att8-1i1k-xk4r-itim.html Wed, 14 Dec 2022 23:00:00 -0000 how far are things that care? https://carado.moe/how-far-are-things-that-care.html https://carado.moe/how-far-are-things-that-care.html Wed, 14 Dec 2022 23:00:00 -0000 about locked posts https://carado.moe/locked.html https://carado.moe/locked.html Tue, 13 Dec 2022 23:00:00 -0000 all claw, no world — and other thoughts on the universal distribution https://carado.moe/all-claw-no-world.html https://carado.moe/all-claw-no-world.html Sat, 10 Dec 2022 23:00:00 -0000 a rough sketch of formal aligned AI using QACI https://carado.moe/rough-sketch-formal-aligned-ai.html https://carado.moe/rough-sketch-formal-aligned-ai.html Sat, 19 Nov 2022 23:00:00 -0000 let's stick with the term "moral patient" https://carado.moe/moral-patient-term.html https://carado.moe/moral-patient-term.html Fri, 18 Nov 2022 23:00:00 -0000 logical vs indexical dignity https://carado.moe/logical-indexical-dignity.html https://carado.moe/logical-indexical-dignity.html Fri, 18 Nov 2022 23:00:00 -0000 wonky but good enough alignment schemes https://carado.moe/wonky-good-enough-alignment.html https://carado.moe/wonky-good-enough-alignment.html Thu, 17 Nov 2022 23:00:00 -0000 generalized wireheading https://carado.moe/generalized-wireheading.html https://carado.moe/generalized-wireheading.html Thu, 17 Nov 2022 23:00:00 -0000 "humans aren't aligned" and "human values are incoherent" https://carado.moe/human-values-unaligned-incoherent.html https://carado.moe/human-values-unaligned-incoherent.html Mon, 31 Oct 2022 23:00:00 -0000 a casual intro to AI doom and alignment https://carado.moe/ai-doom.html https://carado.moe/ai-doom.html Sun, 30 Oct 2022 23:00:00 -0000 publishing alignment research and exfohazards https://carado.moe/publishing-infohazards.html https://carado.moe/publishing-infohazards.html Thu, 27 Oct 2022 22:00:00 -0000 counterfactual computations in world models https://carado.moe/counterfactual-computation-in-world-models.html https://carado.moe/counterfactual-computation-in-world-models.html Sun, 23 Oct 2022 22:00:00 -0000 QACI: question-answer counterfactual intervals https://carado.moe/question-answer-counterfactual-intervals.html https://carado.moe/question-answer-counterfactual-intervals.html Tue, 11 Oct 2022 22:00:00 -0000 some simulation hypotheses https://carado.moe/simulation-hypotheses.html https://carado.moe/simulation-hypotheses.html Mon, 10 Oct 2022 22:00:00 -0000 outer alignment: two failure modes and past-user satisfaction https://carado.moe/outer-alignment-past-user.html https://carado.moe/outer-alignment-past-user.html Wed, 05 Oct 2022 22:00:00 -0000 confusion about alignment requirements https://carado.moe/confusion-about-alignment-requirements.html https://carado.moe/confusion-about-alignment-requirements.html Sun, 02 Oct 2022 22:00:00 -0000 my current outlook on AI risk mitigation https://carado.moe/outlook-ai-risk-mitigation.html https://carado.moe/outlook-ai-risk-mitigation.html Mon, 26 Sep 2022 22:00:00 -0000 existential self-determination https://carado.moe/existential-selfdet.html https://carado.moe/existential-selfdet.html Mon, 26 Sep 2022 22:00:00 -0000 surprise! you want what you want https://carado.moe/surprise-you-want.html https://carado.moe/surprise-you-want.html Thu, 15 Sep 2022 22:00:00 -0000 ordering capability thresholds https://carado.moe/ordering-capability-thresholds.html https://carado.moe/ordering-capability-thresholds.html Thu, 08 Sep 2022 22:00:00 -0000 ethics and anthropics of homomorphically encrypted computations https://carado.moe/homomorphically-encrypted-computations.html https://carado.moe/homomorphically-encrypted-computations.html Wed, 07 Sep 2022 22:00:00 -0000 AI alignment curves https://carado.moe/ai-alignment-curves.html https://carado.moe/ai-alignment-curves.html Fri, 19 Aug 2022 22:00:00 -0000 PreDCA: vanessa kosoy's alignment protocol https://carado.moe/predca.html https://carado.moe/predca.html Tue, 16 Aug 2022 22:00:00 -0000 alignment research is very weird https://carado.moe/alignment-research-is-very-weird.html https://carado.moe/alignment-research-is-very-weird.html Tue, 16 Aug 2022 22:00:00 -0000 alignment researchspace is potentially malign https://carado.moe/alignment-researchspace-is-malign.html https://carado.moe/alignment-researchspace-is-malign.html Sun, 14 Aug 2022 22:00:00 -0000 trading with superintelligence: a wonky proto-alignment scheme https://carado.moe/trading-with-superint.html https://carado.moe/trading-with-superint.html Sat, 13 Aug 2022 22:00:00 -0000 why my timelines are short: all roads lead to doom https://carado.moe/why-timelines-short.html https://carado.moe/why-timelines-short.html Fri, 12 Aug 2022 22:00:00 -0000 goal-program bricks https://carado.moe/goal-program-bricks.html https://carado.moe/goal-program-bricks.html Thu, 11 Aug 2022 22:00:00 -0000 scopes of utopia https://carado.moe/utopia-scopes.html https://carado.moe/utopia-scopes.html Wed, 10 Aug 2022 22:00:00 -0000 the Insulated Goal-Program idea https://carado.moe/insulated-goal-program.html https://carado.moe/insulated-goal-program.html Sat, 06 Aug 2022 22:00:00 -0000 quantum immortality and local deaths under X-risk https://carado.moe/quantum-immortality-local-deaths.html https://carado.moe/quantum-immortality-local-deaths.html Sat, 06 Aug 2022 22:00:00 -0000 probability under potential hardware failure https://carado.moe/probability-hardware-failure.html https://carado.moe/probability-hardware-failure.html Wed, 03 Aug 2022 22:00:00 -0000 tiling the cosmos might be unavoidable https://carado.moe/tiling-unavoidable.html https://carado.moe/tiling-unavoidable.html Fri, 01 Jul 2022 22:00:00 -0000 generalized values: testing for patterns in computation https://carado.moe/generalized-values-testing-patterns.html https://carado.moe/generalized-values-testing-patterns.html Fri, 17 Jun 2022 22:00:00 -0000 solonomonoff induction, time penalty, the universal program, and deism https://carado.moe/solomonoff-deism.html https://carado.moe/solomonoff-deism.html Fri, 17 Jun 2022 22:00:00 -0000 generalized computation interpretability https://carado.moe/generalized-computation-interpretability.html https://carado.moe/generalized-computation-interpretability.html Fri, 17 Jun 2022 22:00:00 -0000 anthropic reasoning coordination https://carado.moe/anthropic-reasoning-coordination.html https://carado.moe/anthropic-reasoning-coordination.html Thu, 09 Jun 2022 22:00:00 -0000 where are your alignment bits? https://carado.moe/alignment-bits.html https://carado.moe/alignment-bits.html Sun, 22 May 2022 22:00:00 -0000 implementing the platonic realm https://carado.moe/implementing-the-platonic-realm.html https://carado.moe/implementing-the-platonic-realm.html Sat, 21 May 2022 22:00:00 -0000 AI boxing could be easy https://carado.moe/ai-boxing-easy.html https://carado.moe/ai-boxing-easy.html Wed, 18 May 2022 22:00:00 -0000 predictablizing ethic deduplication https://carado.moe/predictablizing-ethic-deduplication.html https://carado.moe/predictablizing-ethic-deduplication.html Wed, 18 May 2022 22:00:00 -0000 generalized adding reality layers https://carado.moe/generalized-adding-reality-layers.html https://carado.moe/generalized-adding-reality-layers.html Sun, 15 May 2022 22:00:00 -0000 smaller X-risk https://carado.moe/smaller-x-risk.html https://carado.moe/smaller-x-risk.html Fri, 13 May 2022 22:00:00 -0000 cognitive biases regarding the evaluation of AI risk when doing AI capabilities work https://carado.moe/ai-capability-risk-biases.html https://carado.moe/ai-capability-risk-biases.html Wed, 11 May 2022 22:00:00 -0000 hope for infinite compute https://carado.moe/hope-infinite-compute.html https://carado.moe/hope-infinite-compute.html Wed, 11 May 2022 22:00:00 -0000 AI risk plans https://carado.moe/ai-risk-plans.html https://carado.moe/ai-risk-plans.html Fri, 29 Apr 2022 22:00:00 -0000 a unit for utils https://carado.moe/utils-unit.html https://carado.moe/utils-unit.html Fri, 29 Apr 2022 22:00:00 -0000 the uncertainty of 2+2=4 https://carado.moe/uncertainty-2+2=4.html https://carado.moe/uncertainty-2+2=4.html Tue, 12 Apr 2022 22:00:00 -0000 finding earth in the universal program https://carado.moe/finding-earth-ud.html https://carado.moe/finding-earth-ud.html Mon, 11 Apr 2022 22:00:00 -0000 The Peerless https://carado.moe/the-peerless.html https://carado.moe/the-peerless.html Sat, 09 Apr 2022 22:00:00 -0000 bracing for the alignment tunnel https://carado.moe/bracing-alignment-tunnel.html https://carado.moe/bracing-alignment-tunnel.html Tue, 29 Mar 2022 22:00:00 -0000 should we implement free will? https://carado.moe/implement-free-will.html https://carado.moe/implement-free-will.html Mon, 21 Mar 2022 23:00:00 -0000 goals for emergency unaligned AI https://carado.moe/emergency-unaligned-ai-goals.html https://carado.moe/emergency-unaligned-ai-goals.html Sun, 20 Mar 2022 23:00:00 -0000 are there finitely many moral patients? https://carado.moe/finite-patients.html https://carado.moe/finite-patients.html Sun, 20 Mar 2022 23:00:00 -0000 making the UD and UDASSA less broken: identifying time steps https://carado.moe/udassa-time-steps.html https://carado.moe/udassa-time-steps.html Sat, 05 Mar 2022 23:00:00 -0000 experience/moral patient deduplication and ethics https://carado.moe/deduplication-ethics.html https://carado.moe/deduplication-ethics.html Tue, 01 Mar 2022 23:00:00 -0000 do not hold on to your believed intrinsic values — follow your heart! https://carado.moe/not-hold-on-to-values.html https://carado.moe/not-hold-on-to-values.html Sat, 05 Feb 2022 23:00:00 -0000 forking bitrate and entropy control https://carado.moe/forking-bitrate-entropy-control.html https://carado.moe/forking-bitrate-entropy-control.html Thu, 03 Feb 2022 23:00:00 -0000 balancing utilitarianism https://carado.moe/balancing-utilitarianism.html https://carado.moe/balancing-utilitarianism.html Wed, 02 Feb 2022 23:00:00 -0000 hackable multiverse https://carado.moe/hackable-multiverse.html https://carado.moe/hackable-multiverse.html Mon, 10 Jan 2022 23:00:00 -0000 how timelines fall https://carado.moe/how-timelines-fall.html https://carado.moe/how-timelines-fall.html Mon, 10 Jan 2022 23:00:00 -0000 uploading people for alignment purposes https://carado.moe/upload-for-alignment.html https://carado.moe/upload-for-alignment.html Thu, 06 Jan 2022 23:00:00 -0000 questions about the cosmos and rich computations https://carado.moe/questions-cosmos-computations.html https://carado.moe/questions-cosmos-computations.html Tue, 04 Jan 2022 23:00:00 -0000 brittle physics and the nature of X-risks https://carado.moe/brittle-physics.html https://carado.moe/brittle-physics.html Sun, 26 Dec 2021 23:00:00 -0000 https://carado.moe/less-quantum-immortality.html https://carado.moe/less-quantum-immortality.html Fri, 24 Dec 2021 23:00:00 -0000 non-scarce compute means moral patients might not get optimized out https://carado.moe/nonscarce-compute-optimize-out.html https://carado.moe/nonscarce-compute-optimize-out.html Fri, 24 Dec 2021 23:00:00 -0000 yes room above paperclips? https://carado.moe/above-paperclips-2.html https://carado.moe/above-paperclips-2.html Wed, 08 Dec 2021 23:00:00 -0000 non-interfering superintelligence and remaining philosophical progress: a deterministic utopia https://carado.moe/noninterf-superint.html https://carado.moe/noninterf-superint.html Sat, 20 Nov 2021 23:00:00 -0000 Genuineness, Existential Selfdetermination, Satisfaction: pick 2 https://carado.moe/genuineness-existselfdet-satisfaction-pick2.html https://carado.moe/genuineness-existselfdet-satisfaction-pick2.html Fri, 19 Nov 2021 23:00:00 -0000 no room above paperclips https://carado.moe/above-paperclips.html https://carado.moe/above-paperclips.html Fri, 19 Nov 2021 23:00:00 -0000 unoptimal superintelligence loses https://carado.moe/unoptimal-superint-loses.html https://carado.moe/unoptimal-superint-loses.html Sun, 07 Nov 2021 23:00:00 -0000 against AI alignment ? https://carado.moe/against-ai-alignment.html https://carado.moe/against-ai-alignment.html Fri, 22 Oct 2021 22:00:00 -0000 alignment is an optimization processes problem https://carado.moe/alignment-optimization-processes.html https://carado.moe/alignment-optimization-processes.html Tue, 12 Oct 2021 22:00:00 -0000 exact minds in an exact world https://carado.moe/exact-minds-in-an-exact-world.html https://carado.moe/exact-minds-in-an-exact-world.html Mon, 30 Aug 2021 22:00:00 -0000 ∀V: A Utopia For Ever https://carado.moe/∀V.html https://carado.moe/∀V.html Tue, 24 Aug 2021 22:00:00 -0000 what happens when you die? https://carado.moe/what-happens-when-you-die.html https://carado.moe/what-happens-when-you-die.html Sun, 15 Aug 2021 22:00:00 -0000 kolmogorov complexity objectivity and languagespace https://carado.moe/kolmogorov-objectivity-in-languagespace.html https://carado.moe/kolmogorov-objectivity-in-languagespace.html Sat, 24 Jul 2021 22:00:00 -0000 what is value? https://carado.moe/what-is-value.html https://carado.moe/what-is-value.html Tue, 20 Jul 2021 22:00:00 -0000 systems and diversity https://carado.moe/systems-and-diversity.html https://carado.moe/systems-and-diversity.html Sun, 18 Jul 2021 22:00:00 -0000 botched alignment and alignment awareness https://carado.moe/botched-alignment-and-awareness.html https://carado.moe/botched-alignment-and-awareness.html Sat, 17 Jul 2021 22:00:00 -0000 AI alignment timeline codes https://carado.moe/timeline-codes.html https://carado.moe/timeline-codes.html Fri, 16 Jul 2021 22:00:00 -0000 AI alignment and wolfram physics https://carado.moe/ai-alignment-wolfram-physics.html https://carado.moe/ai-alignment-wolfram-physics.html Thu, 15 Jul 2021 22:00:00 -0000 universal complete https://carado.moe/universal-complete.html https://carado.moe/universal-complete.html Fri, 09 Jul 2021 22:00:00 -0000 estimating the amount of populated intelligence explosion timelines https://carado.moe/estimating-populated-intelligence-explosions.html https://carado.moe/estimating-populated-intelligence-explosions.html Tue, 15 Jun 2021 22:00:00 -0000 my answer to the fermi paradox https://carado.moe/fermi-paradox.html https://carado.moe/fermi-paradox.html Tue, 15 Jun 2021 22:00:00 -0000 the persistent data structure argument against linear consciousness https://carado.moe/persistent-data-structures-consciousness.html https://carado.moe/persistent-data-structures-consciousness.html Thu, 24 Dec 2020 23:00:00 -0000 You are your information system https://carado.moe/you-are-your-information-system.html https://carado.moe/you-are-your-information-system.html Tue, 08 Sep 2020 22:00:00 -0000 Determining core values & existential self-determination https://carado.moe/core-vals-exist-selfdet.html https://carado.moe/core-vals-exist-selfdet.html Sun, 26 Apr 2020 22:00:00 -0000 Limiting Real Universes https://carado.moe/limiting-real-universes.html https://carado.moe/limiting-real-universes.html