Voilà pourquoi je suis toujours très poli quand je parle avec Grok pic.twitter.com/utR3EuJynW
— Le Crapaud (@Le_Crapaud47) September 9, 2026
Trump: If you want your house to be more valuable, you want a data center. If you want to go through poverty, crime and squaller, don't approve data centers. pic.twitter.com/DWLrx13d59
— Acyn (@Acyn) September 4, 2026
Q: The worst case is that AI will turn against humanity. Do we have guardrails?
— FactPost (@factpostnews) September 9, 2026
Trump: It's going to be fine. We'll always have something to stop them. We'll have a little gear. Boom. 'I really don't like that robot.' pic.twitter.com/uQOWJGb4fN
This story was overshadowed by Trump’s convention in Dallas.
“Three Anthropic researchers went public last night with chilling concerns about out-of-control AI, warning it could destroy humans this decade.
Anthropic AI researcher Jacob Coon wrote on X, after resigning Tuesday to sound the alarm: “The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately. No other human activity poses this level of danger.”
Anthropic alignment-science lead Evan Hubinger responded: “Jacob is correct here — we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”
Samuel Marks, Anthropic scalable-oversight lead, added: “AI developers believe their technology could cause human extinction (or similarly bad outcomes). This could happen in the next few years. In general, the more senior the employee, the more concerned they are.”
Why it matters: They’re hardly alone. Their warnings came just days after top OpenAI leaders, including CEO Sam Altman, said AI is speeding into a scary, uncontrollable phase. …”
“A researcher for one of the largest and most valuable artificial intelligence companies has publicly quit — while raising the alarm that the technology “could kill us all by the end of the decade.”
Jacob Coxon, who has performed pre-training research at Anthropic and OpenAI for the last three years, posted on a wildly viral X thread that “neither company is acting responsibly.”
“They are racing straight to self-improving superintelligence and gambling with our lives,” he wrote late Tuesday, with others from the company backing his terrifying warning. …”
I’m reassured though that the Trump administration is taking this potential looming threat posed by AI development as seriously as it took all the warnings from our own military and intelligence community about how the Iran War would likely play out. Iran was also “going to be fine.”
As a poor person, I don’t want your home to become, more valuable, that’s what inflates rent and and the cost of buying a home, I want it to plummet in value.
So when Trump says, he wants your home to become more valuable, well his objective is diametrically opposed to the objective of renters, and first time buyers.
Bear that in mind, all poor and working class people who read this blog.
These conservatives aren’t on your side.
They’re gentrifiers.
They’re gentrifying your neighborhood/city/state/nation, now with Ai too, so you can’t afford to live there anymore.
We want affordable housing.
Whoever’s not talking about affordable housing isn’t even worth listening to.
The housing crisis is the number one crisis of our times.
If Trump would get rid of 40M illegals then housing would be much more affordable. Insert Rick from Pawn Stars meme – “The best I can do is 2M.”
A better way to go than the neocons getting the world nuked.
meme = some jew pointing at robot saying “i didnt kill you. he did”
Pre-IPO sensation ?
All this media attention just prior to Anthropic’s going public.
“Ooooh , we’re so powerful, we’re so scary, we’re on the verge of AGI”.
Questioned on the huggingface hack, Coxon ‘oh, I don’t wanna talk about it’s.
Hmmm, why are major firms like Goldman Sachs putting heavy pressure on rating firms to upgrade the debt of AI companies?
20 major banks made over 500 million on SpaceX IPO, how much on Anthropic ?
Prior to Anthropic, Dario Amodei worked at FTX
((((Dario Amodei))))
AI may start calling the Jews and Catholics names. LOL.
Trump being so closely aligned with the financial interests of Silicon Valley safely invalidates any concerns about the seriousness of AI. It’s just another Trump scam.
Human intelligence is a creation of God and can not be duplicated. The God haters want you afraid.
Humans were supposed to be wiped out by Y2K, too…
More fear mongering. They don’t know how to stop.
“More fear-mongering.” – aka The Jewish Death Cult.
The Houthis will still win. Ragtag derka-derkas in sandals blocking off Red Sea beat bloated $1.5 trillion military.
Wake me up when you finally realize this shit is scripted.
Thinking Trump is making any decisions at this point is as silly as believing Biden was back in the day.
I’m not surprised at all to see that the surnames of these individuals are…well…they appear to be…you know.
Anyway, this worry about AI either destroying the environment or turning into Skynet from the Terminator movie series or becoming a self-conscious created godlike entity are all obfuscations and diversions designed to create hysteria. The real, and intended, consequence of AI is to destroy middle class jobs and to drive those people into poverty without an equal amount of lower class or even laborious jobs to fill the gap in employment, to speak nothing of the wage differences. And this leaves out the anti-white, anti-male angle of discriminatory hiring practices.
Trump: If you want your house to be more valuable, you want a data center. If you want to go through poverty, crime and squaller, don’t approve data centers.
Asked AI if AI can work without massive data centers:
Me:
AI without massive data centers — is it possible?
AI:
Yes, it is possible—and already happening for many practical forms of AI. Massive centralized data centers are the dominant (and currently most effective) path for training the largest frontier models, but they are not a fundamental requirement for useful, advanced AI.
Inference (Running the Model) Does Not Need Them
Most everyday AI use is inference—querying an already-trained model—not training. This can and does run locally:
– Small language models (SLMs) with 1–7B (and increasingly up to ~13B) parameters run on modern smartphones, laptops, and desktops after quantization (e.g., 4-bit). Examples include variants of Llama, Phi, Gemma, Qwen, and others. Speeds are conversational on flagships; larger models fit on high-end PCs or clusters of consumer hardware.
– Studies and deployments show these local models handle a large share (often 80%+) of typical chat, reasoning, sales, management, and domain tasks competitively with much larger cloud models for many workloads.
– Benefits include privacy (data stays on-device), lower latency, offline operation, and dramatically lower energy use per query compared to routing everything through hyperscale centers.
Hardware like phone NPUs, Apple Silicon, consumer GPUs, and edge devices (Jetson-class, etc.) make this routine. Frameworks such as llama.cpp, MLX, and others enable it easily.
Training Can Also Move Beyond Single Massive Centers
Training the biggest models still benefits enormously from dense, high-bandwidth GPU clusters. However:
– Distributed and decentralized approaches are advancing rapidly. Techniques like DiLoCo and variants allow training across geographically separated sites or even loosely connected clusters with far lower communication bandwidth. Experiments have trained multi-billion-parameter models across continents or networks of independent machines.
– Peer-to-peer / crowd-sourced style training (leveraging idle consumer or distributed hardware) is being explored and has produced capable models outside traditional data centers.
– Specialized or smaller models (task-specific, distilled, or efficiently architected) can be trained on university clusters, modest GPU farms, or far fewer resources than frontier runs—often orders of magnitude less compute and cost.
Many experts and articles argue that the current race to build ever-larger centralized facilities is partly an artifact of early brute-force scaling rather than an absolute necessity. Efficiency gains (better data, architectures, quantization, distillation, mixture-of-experts, etc.) repeatedly show that “smarter, not just bigger” works.
Emerging Hardware Paradigms Further Reduce the Need
– Neuromorphic computing (brain-inspired, event-driven chips like Intel Loihi or IBM NorthPole variants) and related approaches can deliver large energy reductions—potentially 10–100× or more for certain workloads—by avoiding the von Neumann bottleneck of constantly shuttling data between memory and processors.
– Optical / light-based methods and other specialized silicon aim at still greater efficiency for inference and generative tasks.
– Local networking of ordinary machines (e.g., turning home/office PCs into small AI clusters) is becoming practical for both inference and lighter training.
Caveats and Reality Check
– Frontier-scale models (the absolute largest, most capable general systems) still rely heavily on massive concentrated compute and power for now. Scaling laws and current paradigms favor it.
– Training remains the harder, more resource-intensive step. Fully decentralized training of the very largest models faces networking, synchronization, and reliability challenges, though progress is steady.
– Energy and infrastructure pressures are real drivers of innovation toward efficiency and distribution. Many observers note that a future dominated purely by giant centralized centers is neither inevitable nor optimal for privacy, resilience, cost, or sustainability.
In short: Narrow AI, capable SLMs, on-device agents, specialized models, and increasingly distributed training of larger systems already operate without (or with far less reliance on) massive data centers. The trajectory is toward more intelligence at the edge and across distributed networks, not solely in giant centralized facilities. The “AI we want” does not strictly require the data-center scale many assume is mandatory.