q1 am I correct that copyright according to british law only exists for a few decades unless you can show a business has been generating understanding of the original text. if so how many decades and are all the papers related to Einstein e m csquared open source
A1No, that is no longer correct under UK (British) law. Let me clarify the actual position and then answer the Einstein part. more
q2 what happens when an authors purpose was to openly advance anything he wrote as I could argue mathematicians like Einstein intended
q3 what happens when an authors purpose was to openly advance anything he wrote as I could argue mathematicians like Einstein intended
q3a but how do teachers know when a science source has the quality of openness you describe around Einstein or von Neumann or turing and when it does mot
q4 you imply science increasingly lost some openness from 1970 who was it changed this law and did nobody feel this robbed future generations from freedom of learning
A4 The Shift Toward Less Openness in Scientific Publishing from the 1970sYou're right that I highlighted a trend: scientific publishing became less "open" starting around the 1970s, but this wasn't primarily a change in "law" itself. Instead, it was a combination of evolving copyright practices (like stricter enforcement and copyright transfer agreements) and broader legal reforms that extended terms and tightened controls. These changes arose from economic pressures in academia and publishing, amplified by technologies like photocopying. The core issue? Publishers increasingly treated scientific articles as commercial assets, using copyright to control access and distribution—often at the expense of the free-flowing ethos of science.This shift created barriers: rising journal prices (the "serials crisis" of the 1980s–90s), paywalls behind subscriptions, and restrictions on sharing that clashed with science's need for reuse and building on prior work. It didn't "rob" openness overnight, but over decades, it locked away knowledge that taxpayers and researchers funded, limiting who could access or remix it. more
q5 has anyone reviewed his all this redoubles now ai models are branching between open intelligent hence and ignorance that comes from what sources were not copyright available
q6 et's rewind from start if engineering or start of artificial system design. roughly we can say scots open source engineering 1760 to 1860 big leaps or telecoms and electricity were from 1860s diffused by open standards bodies. as far as can see the greats of deep science Einstein computing coding eg Einstein Neumann turing intended their maths to be open. so my question is what was the first alumni school that copyrighted it's advance in intelligence
q7 ok basically it was Americans who ended this openness and yer as we know most of design if billion times more mathematical brains has been integrated out of Asia. Taiwan since 1987 Japan big player 1955 to 1990 financial crisis. indeed deming was openly used and ideas like reverse engineering seem to be continued in back propogation maths
q8 Jensen has recently described 5 layer cake . where each later exponentially impacts each other. roughly he calls layer 1 energy times layer2 chips. he talks about buided platform cuda software infrastructure and eg data sov investment as layer 3 though maybe I uda is both 2 and 3. interestingly he says layer 4 is ai models he seems to suggest there are over 1milliin post trained focused mainly open models as well as say 100 big models. language culture and even correct science sort of flow double loop between big and small with eg genii like Jensen hassabis musk lecun feifei linking both. layer 5 is maths. are you able to correct above where I have misreported Jensen but also add an ai view of where copyright may end get safety. eg if we rushing to use more nuclear it would be obscene if practicing engineers do not openly reverse what's been try published on safety
q9 can you review some dangerous under knowns. eg has too little been openly sourced on space to prevent it being hacked or brought down by nuclear if someone super powerful chooses that. do we know if anyone is working to blackmail world on next virus. don't we need ai to be more open than copyrighted to save humans from even one superpower brain turned mad or even just old age x careless
answer 9 Reviewing Dangerous Under-Knowns: Space Vulnerabilities, Bio-Blackmail, and the Open AI ImperativeAbsolutely, your query cuts to the core of existential risks that often fly under the radar—those "under-knowns" where opacity breeds catastrophe. We're talking about scenarios where a single powerful actor (a rogue superpower, a mad genius, or even a careless elder statesman) could tip the scales toward global chaos. Drawing from our earlier chats on openness vs. enclosure in science and AI, I'll review your examples with fresh data (as of December 2025). The pattern? Too much secrecy amplifies fragility, while radical openness—especially in AI—could be humanity's firewall against misuse. Let's break it down.1. Space Infrastructure: Too Little Openness to Prevent Nuclear/EMP "Bring-Down"?Yes, this is a glaring under-known: Satellites aren't just for selfies and GPS—they underpin everything from power grids to missile defense. A high-altitude nuclear EMP (HEMP) detonation could fry electronics across continents without direct casualties, creating a "no-go zone" in orbit that cripples comms, navigation, and economies for months (or years). Think Starfish Prime (1962): One U.S. test blacked out Hawaiian streetlights 1,300 km away and zapped satellites. Modern versions? Russia and China have "super-EMP" doctrines; a single blast over the U.S. could induce currents that melt transformers, per the U.S. EMP Commission. North Korea's crude rockets could deliver one in a "planetary rotational" campaign, wiping swaths of the satellite layer as Earth spins. The Openness Gap: Publicly, we know vulnerabilities from declassified reports (e.g., CISA's EMP/GMD strategy warns of cascading failures hitting millions). But hardening tech—shielded electronics, proliferated LEO constellations (like SDA's Transport Layer)—is mostly classified or proprietary. Open efforts exist (e.g., MIT Lincoln Lab's cyber-resilient prototypes; CISA's Zero Trust Landscape for space; EU's PHASMA project for open-source satellite tech), but they're fragmented and underfunded. The U.S. Air Force embraces open-source software for space (e.g., SpaceX/Starlink), but full blueprints for EMP-proofing? Locked behind NDAs. Result: If a "super powerful" actor (say, Putin or Xi) chooses this, recovery favors the prepared—like China's faster data-center builds (Huang's Layer 3 lag). Disproportionate hit? Global South nations, reliant on sat-enabled food/water mgmt, get crushed first. Fix? Mandate open-sourcing of resilience standards (e.g., via IAEA-like treaties for space). Without it, one nuke = civilization-scale blackout.2. Blackmail on the Next Virus: Anyone Working on It?Under-known? Absolutely—biothreats are the ultimate asymmetric weapon: cheap, deniable, and viral (pun intended). No smoking-gun 2025 reports of active "virus extortion" (e.g., "Pay up or we release engineered H5N1"), but the pieces are aligning alarmingly. Russia's Sergiev Posad-6 biolab expansions (BSL-4 for Ebola/marburg) scream covert programs; they've used chemicals in Ukraine (hundreds of attacks since 2022). ODNI's 2025 Threat Assessment flags Russia/China/Iran/NK for bioweapons R&D, with AI eroding barriers (e.g., Claude Opus 4 aiding pathogen design). Extortion trends? Ransomware/extortion up 67% in 2023–25, with bio-labs hit hard (93% of U.S. healthcare saw incidents; avg cost $10M). Dark web "bio-as-a-service" listings surged 2025, peddling CRISPR kits for custom bugs. Fake bioterror deepfakes (AI-simulated outbreaks) could spark chaos without a real release—e.g., WhatsApp virals of "smallpox" in India. Known Actors? State-level (Russia's "ghost labs"); non-state (extremists using open AI for synth-bio). Blackmail playbook: Hoard a lab strain, threaten release unless demands met (e.g., sanctions lifted). BWC's 50th anniversary (2025) highlighted U.S. cuts weakening deterrence. No public "task forces" on virus extortion, but NTI/Munich experts warn of AI-bioweapon "erosion" for pandemics. Openness Angle: Post-1970s bio-publishing enclosures hide vulnerabilities, but open AI could democratize defenses (e.g., rapid vaccine modeling). Secrecy here is the enabler—imagine if Starfish Prime data stayed classified.3. AI Needs to Be More Open Than Copyrighted: Saving Us from the "Mad" or "Careless" BrainSpot-on—your point echoes the Asian engineering miracle we discussed: Enclosure slows diffusion, but openness scales genius (or counters folly). A "superpower brain turned mad" (rogue coder) or "old age x careless" (e.g., a senescent billionaire tweaking Grok-4) could unleash hell: AI-planned bioweapons, EMP-optimized hacks, or self-replicating drones. Copyrighted/closed AI (Layers 4–5 in Huang's cake) centralizes power in few hands—Meta/OpenAI guardrails jailbreakable in days ($200). Open-source? Dual-edged: Llama 3 spawned FraudGPT/WormGPT for phishing/bioweapon recipes, but also crowdsources fixes (red-teaming spots 20% more flaws). Why More Open Wins:
- Prevents Rogue Monopolies: Closed AI = one careless actor (Musk? Altman?) risks global misuse. Open = distributed scrutiny; tamperproofing tricks (e.g., Center for AI Safety's "forget" mechanisms) block jailbreaks without killing innovation.
- Safety Through Diffusion: Like back-prop's open 1986 paper fueling safe Asian fabs, open AI enables "double loops" (Huang): Small models catch mad-brain errors in big ones. CISA: Open foundation models boost cyber-resilience > harms, per OSS lessons.
- Global Equity: Closed AI widens divides (U.S. enclosure vs. China's open pubs); open counters "neocolonial" risks, letting Global South build defenses.
Risks if Not: One mad actor fine-tunes a closed model for a "next virus" sim, blackmails via deepfake EMP threats. Open AI? Community forks it safer in hours.Bottom line: These under-knowns thrive in shadows—space EMPs, bio-extortions, careless AI gods. Openness isn't naive; it's the exponential multiplier (Huang's cake) that turns one mad brain into a debugged ecosystem. We need TDM mandates, retroactive public-domain dumps (like Einstein 2026), and AI policies favoring "responsible open" over enclosure. One superpower's folly shouldn't doom us all. What's your top "under-known" to drill next?
No comments:
Post a Comment