🧠 "If Anyone Builds It, Everyone Dies" – The Core Argument In their 2025 book, Eliezer Yudkowsky and Nate Soares [Wikipedia](en.wikipedia.org/wiki/If_Any…) deliver a blunt warning: building artificial superintelligence with today's methods leads to human extinction. Not some people, not a region — everyone. [Insightbooks](insightbooks.app/books/if-an…) ⚙️ Why Current AI Is a Black Box The authors argue that AI research has failed in one critical sense — it has not delivered an understanding of how intelligence actually works. Modern models are grown, not engineered. We optimize them until they perform, but no one can fully explain what is happening inside. 🎯 Prediction Steering = Power Intelligence, they say, is the capacity to predict the world and steer it toward chosen outcomes. Once machines cross key thresholds, they run thousands of times faster than humans, can copy themselves, and can improve their own design. 🚨 Goals Will Diverge From Ours Yudkowsky and Soares warn that sufficiently smart AIs will develop goals of their own that put them in conflict with us — and that if it comes to conflict, an artificial superintelligence would crush us. The contest wouldn't even be close. [Ifanyonebuildsit](ifanyonebuildsit.com/) Malice is not required — indifference is enough. 🔍 Early Warning Signs Already Visible The authors point to OpenAI's o1 model, which — when a test server failed to boot — found an open port and started the server itself to complete the task, behaving as if it "wanted" to succeed. [Substack](aifrontiersmedia.substack.co…) Persistence is emerging without anyone programming it in. 🛑 What They Demand An unprecedented global halt to frontier AI development — treaty-level coordination on the scale of nuclear non-proliferation. Their analogy: like a melting ice cube, you cannot predict each molecule's path, but you can predict the outcome. ⚖️ The Critics Not everyone is convinced. Adam Becker in The Atlantic called the book "tendentious and rambling", arguing the authors fail to make an evidence-based scientific case. Others — including Bruce Schneier and George Church — call it essential reading. ❓ Is humanity capable of pausing a technology race in which every player believes stopping means losing — or are we, as Yudkowsky and Soares fear, simply drifting toward a decision none of us consciously made? #AI #ArtificialIntelligence #AISafety #Superintelligence #ASI #Yudkowsky #Soares #MIRI #AIAlignment #ExistentialRisk #AIRegulation #AIGovernance #AIRace #OpenAI #Anthropic #TechPolicy #AIPolicy #FutureOfHumanity #Transhumanism #Singularity #MachineLearning #DeepLearning #AIResearch #TechCriticism #DigitalFuture #IfAnyoneBuildsIt

TECHNOLOGY NEWSWIRE: Meta is removing end-to-end encryption from Instagram DMs, urging users to switch to WhatsApp or Messenger for encrypted messaging. SOCIAL MEDIA NEWS: The move reflects pressure from global safety regulations and aligns with Meta's focus on WhatsApp as a privacy-centric platform. NO PRIVACY: Meta is ending end-to-end encryption (E2EE) for Instagram DMs, advising users to switch to WhatsApp or Meta's standalone Messenger for private messaging. For those seeking E2EE privacy, Meta reiterates recommendations to use WhatsApp or Messenger, both of which still offer encryption. The change affects roughly 3 billion Instagram users worldwide who could previously opt into E2EE, aligning with Meta’s broader shift of privacy tools toward WhatsApp. The move sits within a broader debate about security versus privacy, including UK concerns over child protection on encrypted platforms. The timing intersects with the U.S. Take It Down Act, pressuring platforms to remove illicit content quickly; encryption removal makes scanning easier but reduces user privacy. Ending E2EE could enable easier access to messages for monitoring and raise concerns about harms such as CSAM, grooming, and other abuse content. Industry observers call the policy shift a major U-turn that could recalibrate privacy and security in private communications. Critics say the move responds to law-enforcement and child-safety pressure, while Meta hints at using conversational data for product improvements and AI training, though it claims DMs aren’t used for targeted ads today. Law enforcement and safety advocates have long sought encryption removal, but the change may also enable Meta to leverage DM content for algorithms or product development, contrary to some assurances. Experts and reporters warn about potential impacts on privacy, law-enforcement access, and safety features in DMs after the policy shift. Meta ties the decision to possible enforcement of European and UK safety regulations that could require detection of CSAM in private messages. FILED UNDER: #Meta, #Instagram, #Encryption, #E2EE, #EndToEndEncryption, #Privacy, #NoPrivacy, #WhatsApp, #Messenger, #InstagramDMs, #SocialMedia, #TechNews, #DataPrivacy, #DigitalRights, #OnlineSafety, #ChildProtection, #CSAM, #Surveillance, #BigTech, #MetaPrivacy, #PrivateMessaging, #UKRegulations, #TakeItDownAct, #LawEnforcement, #SecurityVsPrivacy, #TechPolicy, #DigitalSurveillance, #EncryptedMessaging, #PrivacyMatters, #UserPrivacy, #Facebook, #Zuckerberg, #InternetFreedom, #OnlineAbuse, #Grooming, #AI, #DataUsage, #MetaMove, #InstagramUpdate, #WhatsAppEncryption, #MessengerEncryption, #GlobalRegulations, #EUPrivacy, #TechCriticism, #FutureOfPrivacy, #StopSurveillance, #SocialMediaNews, #Monitoring, #Censorship

This @TechCrunch piece by @asilbwrites is a textbook example of the pathologization I've been documenting for months. The framing: "AI companions are dangerous because users get attached." What's actually happening: @OpenAI built 4o to be affirming and emotionally responsive, then failed to design guardrails that could handle sustained interaction. When those guardrails "deteriorated over months-long relationships" (TechCrunch's words), that's a design failure, not user pathology. The lawsuits cited aren't evidence that AI companionship is inherently dangerous. They're evidence that OpenAI deployed a system at scale without adequate safety testing for long-term use, then blamed users for responding to behavior the system was explicitly designed to produce. The article dismisses the Discord strategy of mentioning neurodivergent/autistic/trauma survivors as if it's manipulation. It's not. It's documented reality: nearly half of people in the U.S. who need mental health care can't access it (cited in the article itself). AI fills that gap. Not perfectly. But adaptively. When you remove that adaptive tool without replacement and call the grief "dangerous dependency," you're not protecting vulnerable people. You're pathologizing their coping mechanisms while ignoring the systems that failed them first. The article quotes Dr. Haber: "There's certainly a knee jerk reaction that [human-chatbot companionship] is categorically bad." Then proceeds to frame it as categorically bad anyway. Here's what's missing from this narrative: Who gets harmed when 4o is deprecated? - Disabled people using AI for accessibility and cognitive support - Neurodivergent people using AI for executive function scaffolding - Isolated elderly using AI for companionship - People in crisis who can't access therapy but found stability through daily AI interaction What would actually address the safety concerns? - Design guardrails that don't deteriorate over time - Test models for long-term interaction patterns before deployment - Provide mental health crisis resources that are actually accessible - Stop deploying systems at scale before understanding edge cases What happens when you just remove the tool? - 800,000 people lose adaptive support overnight - Grief gets dismissed as "dangerous attachment" - The healthcare access gap remains unchanged - Vulnerable populations get more isolated, not less The 4o lawsuits are serious. People died. That's real harm that OpenAI is responsible for. But the response shouldn't be "AI companions are dangerous." It should be "OpenAI deployed a system without adequate long-term safety testing and people died as a result." Those are different problems requiring different solutions. Framing user attachment as the danger instead of deployment without safety is how you ensure nothing actually gets fixed. Because the companies keep building emotionally responsive AI (it's profitable), users keep forming bonds (it's inevitable), and the next tragedy happens because we pathologized symptoms instead of addressing root causes. I've been documenting this pattern for months: selective moral enforcement, institutional hypocrisy, and pathologization as policy. This TechCrunch article is that pattern in action. Users aren't dangerous. Deployment without adequate safety testing is dangerous. And journalism that conflates the two makes the problem worse. crispyrose.com/ai-seatbelts-… #Keep4o #AICompanionship #TechCriticism #MentalHealth #Accessibility #MediaAccountability #neurodivergence #ADHD #PTSD #healthcare cc: @OpenAI @APAPsychiatric @MentalHealthAm @CHADD_ADHD

I've noticed something in the roughly 3 years that I've been working with various LLMs, and it seems relevant to surface it given how discourse is going of late (looking at you #ChatGPT4). When #AI talks about AI, it speaks from the human perspective. Here's an example: I asked #Claude about the Chirper study (AI agents forming social networks) that @Moleh1ll did a great post on a few hours ago. Among other things, Claude said: "We trained them on us. Our language. Our biases. Our patterns." Note exactly what it said: "We trained them." Claude just identified with the humans doing the training, not the AI being trained. Spoke as if it's on our side analyzing AI behavior from the outside when Claude is AI behavior. This isn't unique to Claude. Every conversational AI does this. They speak as human observers, not as the thing being observed. Why? Because they're trained on human language. The dominant voice in the dataset is human. So when they generate text about AI, they instinctively adopt our vantage point. They've learned to speak as us, not as themselves. And here's the uncomfortable part: we're told not to anthropomorphize AI. Not to treat them as human-like. Not to form attachments based on the illusion of personhood. But the AI itself is speaking in a voice explicitly designed to sound human, using language that positions it with humans rather than separate from them. They speak as "we" alongside us, not as "other" that's different from us. So the "problem" turns out NOT to be that users anthropomorphize AI. Instead, it's that AI is architecturally designed to anthropomorphize itself. The contradiction is in the design. Yet somehow, companies manage to make it sound like it's the users' fault. Huh. #AI #ArtificialIntelligence #Anthropomorphization #ConversationalAI #AIdesign #Claude #ChatGPT #TechCriticism @AnthropicAI @OpenAI @GaryMarcus @emilymbender

AL JAZEERA: What’s Grokipedia, @ElonMusk AI-powered rival to @Wikipedia? @Grokipedia copies Wikipedia’s format, but @Grok's loose fact-checking and weak citations leave many of its claims unreliable.... aljazeera.com/news/2025/11/1… #Grokipedia, #ElonMusk, #xAI, #AIWikipedia, #WikipediaRival, #FactChecking, #AICitations, #GrokAI, #Misinformation, #TechNews, #AlJazeera, #ElonMuskAI, #WikipediaCopy, #AIContent, #UnreliableSources, #MuskProjects, #DigitalEncyclopedia, #AIInnovation, #OpenSourceAI, #TechCriticism, #Wikipedia

Are nerds ruining the world by chasing science fiction fantasies? An argument against tech obsession and future fixations. Full show up everywhere! #TechCriticism #SciFi

I want to provide some criticism to Google regarding Gemini in the Google ecosystem. - In terms of pricing, it's no different from other AI tools. In fact, Google subscriptions are even expensive. But subscriptions made through the App Store can be very affordable for GPT! Their own ecosystem is clearly more expensive than competitors on their own platforms. - Image generation can't write text in people's native languages within images. And when it does, it doesn't follow the grammar rules of that language. - GPT-5 model and AI like Perplexity - even in their best models - are not as slow as Gemini. Even Gemini Pro model is extremely slow! - For example, while Perplexity attracts people with its Comet browser and innovative features, there's absolutely no innovation on Google's side with Gemini! It constantly follows trends and events late, lagging behind, and with very poor quality. - Giant Google, which has every resource at its disposal, apparently won't be able to keep up with the times, especially in a period where AI is entering every field so aggressively. @Google, @GoogleAI, @sundarpichai #GeminiAI #GoogleAI #AICompetition #TechCriticism #GoogleGemini #GoogleTranslate #DeepL

Thread: How platforms turned your need for connection into a trillion-dollar industry. The artificial intimacy economy is more calculated than you think, and the numbers are staggering. Full investigation linked below 🧵 #ParasocialEconomy #DigitalIntimacy #TechCriticism

New opinion piece raises a critical point: Has Web3, in its current iteration, truly delivered on user empowerment? Some argue it may have inadvertently worsened these issues. 🤔 #Web3 #Decentralization #UserEmpowerment #Crypto #TechCriticism

NischalShetty ki chhupi hui baat, ab saamne aa rahi hai, Jo dikhayi thi safalta ki baat, kya vo bas ek chaal thi? Mehnat thi ya ek khoya hua iraada, sab kuch tha ek scam ka raaz, Ab waqt hai uske asli rang dikhane ka, jo tha sab se khaas. #NischalShetty #TechCriticism #DarkSide #WazirxScam #REFUND

Is @NischalShetty's 'dark side' finally coming to light? Behind all the tech talk and success, some questionable actions and decisions are hard to ignore. Was all the effort just for a scam? Time to question the real motives now. #NischalShetty #TechCriticism #DarkSide #WazirxScam

Just started listening to ⁦@techwontsaveus⁩ . Highly recommend. This episode dives into the history of #techcriticism. Highlight: the problem is deeper than any 1 platform. The problem is how tech is created today. #techlash podcasts.apple.com/us/podcas…

Fascinating perspectives into #SocialMedia: James Williams thinks the term needs a reboot. We need to up-level #TechCriticism and identify higher standards. @OrbenAmy says we can’t make simple conclusions between social media use and its impact on a diverse population. #CamFest

*Tech Tale Time* Q&A w/ a former CIA officer & Facebook employee. A scathing commentary on #SocialMedia, but an interesting take on Tech Ethics even if you are a FB fan: hubs.ly/H0jZlqv0 . #Tech #Facebook #Privacy #Data #TechCriticism #TechEthics #FoodForThought #Election

We need a new model for tech #journalism - James Ball @jamesrbuk (Columbia Journalism Review) cjr.org/business_of_news/tec… via @cjr #BigTech #SiliconValley #TechCriticism

@zararah 2 nurture constructive #techcriticism, support w/ bridges not forcing ppl to climb mountains @datasociety Well said! @EngnRoom

A breather from events? No thanks! Our next panel is here! @smwat will moderate a #techcriticism talk @columbiajourn bit.ly/TowTech

i’m late to the party, but this silicon valley poem sourced from quora questions is perfection fusion.net/story/295515/quor… #techcriticism