Miguel Afonso Caetano

RT @triofrancos
I would say I’m fairly knowledgeable about the corruption, violence, coercive authority, and brutal imperial power that pervades extractive sectors and yet I am still in shock at this Trump admin move to secure Zambian minerals

https://www.nytimes.com/2026/03/16/health/zambia-hiv-aid-minerals-trump.html?smid=nytcore-ios-share

"President Donald Trump has been leading a double life in prosecuting his war against Iran. In public, he regularly boasts that Iran’s military might has been decimated, its leadership killed off, and that the few officials remaining alive in Tehran are begging him to talk. “They want to negotiate. They want to negotiate badly,” Trump said Sunday night. “We’re talking to them. But I don’t think they’re ready, but they’re getting pretty close.”

Behind the scenes, it is the Trump administration that has been asking for talks. Two Iranian officials told Drop Site that Trump’s Special Envoy Steve Witkoff personally sent messages to officials in Tehran, including Foreign Minister Abbas Araghchi, last week exploring possibilities for resuming negotiations. Iran has not replied to Witkoff. The Iranian officials told Drop Site that Iran has also received messages from the White House via third countries.

“Because of decisions made by [Iran’s] top authorities, no response was sent to his messages,” a senior Iranian official told Drop Site. “The message here is clear: Iran has once again closed the window for any direct negotiations,” he added. “The authority to declare a ceasefire rests solely with the country’s Supreme Leader. It’s not something the foreign minister, or any other official or organization in Iran, would send messages about to a foreign party.”"

https://www.dropsitenews.com/p/iran-war-trump-witkoff-araghchi

#USA #Trump #Iran #Militarism #War #MiddleEast

"Tech writing and publishing in the Bay Area wasn’t always like this. In the second half of the 2010s, a rising tide of public discontent with Silicon Valley — the “techlash” — coincided with the first Trump inauguration to produce a mass politicization of tech workers. I was one such worker, and became a labor organizer in the tech industry alongside others who would go on to orchestrate everything from a global walkout at Google to salting programs at startups.

Such resistance movements in tech animated, and were in turn supported by, their own homegrown publications. One of them, Logic Magazine, was a forum for tech criticism that was widely read by partisans of the tech worker movement. Logic was where you could go to learn about racist criminal sentencing algorithms, tools for tenants to research landlords, and how to sabotage computer vision systems. My comrades published real-time histories of the tech worker movement and exposés on their employers in Logic; we showed up to their in-person events in San Francisco to meet others who thought similarly.

But that now feels like a past life. Today, the techlash is a receding memory and the tech worker movement is running on fumes, the preserve of a few lonely unions. In a sign of shifting winds, Logic rebranded as Logic(s) in 2022, pivoting to subjects outside Silicon Valley proper, and transitioned to a new editorial team institutionalized at Columbia University.

In the years since, the tech industry has gladly unburdened itself from the critiques leveled at it during the techlash. Nowhere is this more obvious than in tech’s involvement in building weapons of war."

https://bayareacurrent.com/meet-the-new-right-wing-tech-intelligentsia/

#BigTech #SiliconValley #BayArea #TechJournalism #Media #News

RT @HedgieMarkets
🦔 Google quietly dropped a search feature called "What People Suggest" that used AI to organize crowdsourced health advice from strangers into easy-to-understand themes. The company had said it showed "the potential of AI to transform health outcomes across the globe." A Google spokesperson told the Guardian the feature was scrapped as part of a "broader simplification" of the search page and had nothing to do with quality or safety. When asked where this was shared publicly, Google pointed to a blog post that doesn't mention the feature at all.

My Take
Google launched this thing with a blog post from their chief health officer explaining how it would help people find real insights from others with similar conditions. Now it's gone and they're pretending they announced the removal in a blog post that doesn't mention it. I find that harder to believe than just admitting the feature was a bad idea.

The whole concept was crowdsourcing medical tips from amateurs and dressing it up with AI organization. You search for health advice and Google shows you what random people on the internet suggest. This comes after the Guardian found Google's AI Overviews were showing false and misleading health information to 2 billion people a month. Google initially downplayed those findings, then quietly removed AI Overviews for some medical queries. I think they know these features have problems, they just don't want to say it out loud because it undermines the pitch that AI is ready to transform healthcare. Easier to quietly kill things and hope nobody notices.

Hedgie🤗

https://www.theguardian.com/technology/2026/mar/16/google-scraps-ai-search-feature-that-crowdsourced-amateur-medical-advice

RT @HedgieMarkets
🦔 Researchers at Aikido Security found 151 malicious packages uploaded to GitHub between March 3 and March 9. The packages use Unicode characters that are invisible to humans but execute as code when run. Manual code reviews and static analysis tools see only whitespace or blank lines. The surrounding code looks legitimate, with realistic documentation tweaks, version bumps, and bug fixes. Researchers suspect the attackers are using LLMs to generate convincing packages at scale. Similar packages have been found on NPM and the VS Code marketplace.

My Take
Supply chain attacks on code repositories aren't new, but this technique is nasty. The malicious payload is encoded in Unicode characters that don't render in any editor, terminal, or review interface. You can stare at the code all day and see nothing. A small decoder extracts the hidden bytes at runtime and passes them to eval(). Unless you're specifically looking for invisible Unicode ranges, you won't catch it.

The researchers think AI is writing these packages because 151 bespoke code changes across different projects in a week isn't something a human team could do manually. If that's right, we're watching AI-generated attacks hit AI-assisted development workflows. The vibe coders pulling packages without reading them are the target, and there are a lot of them. The best defense is still carefully inspecting dependencies before adding them, but that's exactly the step people skip when they're moving fast. I don't really know how any of this gets better. The attackers are scaling faster than the defenses.

Hedgie🤗
https://arstechnica.com/security/2026/03/supply-chain-attack-using-invisible-code-hits-github-and-other-repositories/

RT @amoebamusic
New "What's In My Bag?" episode with John Waters!!

Legendary filmmaker, writer, and cultural provocateur John Waters goes shopping at @AmoebaSF and talks about some of the best bad girl groups, '70s New York punk rock & lots more.

Watch the full video: https://youtu.be/9AOk8cSrZAY

Once again, Opus 4.6 is the best model. What else is new? :-D

"Can AI agents conduct cyber-attacks autonomously? If AI agents can reliably execute multi-step attack chains with minimal human oversight, it could lower the skill barrier for unsophisticated threat actors, increase the sophistication of attack achievable by experience ones, or enable entirely novel offensive operations.

As cyber capabilities improve, increasingly sophisticated testing is needed to accurately measure them. Existing cyber evaluations rely on isolated capture-the-flag (CTF) challenges or question-answer sets. While valuable for measuring specific skills, these approaches don't capture whether AI systems have the autonomous, long-horizon capabilities required for executing extended attack sequences in complex environments.

To address this gap, we have begun evaluating models on cyber ranges: simulated network environments comprising multiple hosts, services, and vulnerabilities arranged into sequential attack chains; built by cybersecurity experts.

By comparing seven models released over an eighteen-month period (August 2024 to February 2026) at varying inference-time compute budgets, we observe two capability trends.

First, each successive model generation outperforms its predecessor at fixed token budgets: on our corporate network range, average steps completed at 10M tokens rose from just 1.7 (GPT-4o, August 2024) to 9.8 (Opus 4.6, February 2026).
(...)
The best-performing Opus 4.6 run completed 22 of 32 steps, reaching milestone 6, which requires reverse engineering a Windows service binary containing encrypted credentials, escalating privileges via token impersonation, and recovering a cryptographic key to access a C2 management service. Other runs with the same model and budget completed substantially fewer steps."

https://www.aisi.gov.uk/blog/how-do-frontier-ai-agents-perform-in-multi-step-cyber-attack-scenarios

#CyberSecurity #AI #GenerativeAI #CyberAttacks

Mapa dos comboios da CP em tempo real: https://comboios.ruicosta.pt/

Muito giro! Já fiquei a saber que há um comboio Badajoz -> Entroncamento que passa por Abrantes. E se calhar até dá para começar a viagem em Madrid sem grandes demoras 🙂

#Comboios #Portugal #CP #CaminhosDeFerro

Qual Guerra Mundial, qual quê!!! Este Verão o que eu quero é ver os Pavement a tocarem ao vivo - vá-se lá saber aonde!!!

https://www.youtube.com/watch?v=skGqpC0WYSk

Com a derrota do Ventura nas presidenciais e o rol de tretas figurinhas que são os deputados do Chega, pensava que o discurso anti-imigração já não tinha perdas para andar. Afinal enganei-me. Parece que continua a haver pequenitos Salazares espalhados por este país.

A verdade é esta: Num país verdadeiramente Democrático e onde se respeitam as Liberdades individuais, nenhum Estado deve ter o direito de meter o bedelho nos contratos que os seus cidadãos - em nome individual ou colectivo - estabelecem com pessoas de nacionalidade estrangeira na estrita medida em que esses contratos não violarem as leis nacionais e que não estiverem em causa crimes graves.

Não tenho qualquer dúvida que as pessoas que advogam o contrário devem ser chamadas de FASCISTAS. De todos os pontos de vista - económico, cultural, social e político - esse discurso tresanda a ódio e a ressentimento.

Quanto a mim, as pessoas que não sabem conviver com e respeitar os outros, independentemente de raça, etnia, nacionalidade, religião, género, etc. não sabem viver em liberdade e nem deixam viver os outros. Em suma: São um autêntico perigo à liberdade de todos - tanto a minha, como a tua, como a "deles".

E este é que é o grande problema: Como se já não bastasse o Estado estar constantemente nas tintas para garantir-me liberdade positiva para ser dono do próprio destino, realizar o meu potencial e agir de acordo com a minha própria vontade, também ao colocar barreiras completamente espatafúdias e arbitrárias à entrada de mais estrangeiros, o Estado está potencialmente a impedir-me de usufruir de mais liberdade negativa, uma vez que desta forma, está a colocar obstáculos e a interferir na minha capacidade de potencialmente estabelecer contratos com pessoas de nacionalidade estrangeira.

"A war spiralling in the Middle East. A death toll now in the thousands across Iran and Lebanon. Energy prices soaring. The Gulf seized up with Iranian strikes. It’s one of those eras that feels bewildering, incomprehensible, out of control. But there is, at the heart of it, a simple logic: everything that is unfolding is a result of Israel’s occupation of the Palestinians.

As the conflagration spreads, the connection to Palestine becomes obscured. But it is clear how much of the stability of the Middle East was secured at the expense of the Palestinians. Look at the region before 7 October 2023. US policy on the Middle East focused on “integration’’: containment of Iran, signing up more Arab countries to normalise relations with Israel and the creation, therefore, of a bloc of economic and security interests under the US military umbrella.

Iran would be isolated by this Israeli-Arab alliance, and the Palestinians’ file would be closed. Arab countries would pay lip service to them, through demanding guarantees that there would be efforts towards the creation of a Palestinian state, or that the West Bank should not be annexed. But in reality what was on the cards was a continuation of the occupation of Palestinian territories in perpetuity."

https://www.theguardian.com/commentisfree/2026/mar/16/war-iran-gulf-west-palestine-occupation-us

#Palestine #Iran #Israel #MiddleEast #USA

"Each time one person contributes data to improve an AI model, that model becomes better — not just at doing that individual’s job, but at doing the job of everyone in that role, anywhere.

In that sense, workers can end up competing against one another by supplying data too cheaply to companies or intermediaries that recruit people to train AI to do their jobs. When this happens, individuals may unintentionally undermine the bargaining power of others in their occupation.

If the goal is to ensure that workers share in the gains from AI, co-ordination may be necessary. Without it, individuals may strengthen the very systems that weaken their collective bargaining power. Instead, co-operation can benefit workers and companies alike. If fear about career risk leads people to hold back knowledge from AI systems, productivity may suffer. Smart companies will know that finding ways to recognise workers for their talent will ensure that they continue to supply it.

What, then, is to be done? As workers, people should think about how to use AI to expand their skills: whether by building complementary capabilities or by finding ways to scale their expertise through AI systems. As citizens, they should press for policies that give workers clearer rights over the data generated by their work and compensation for it."

https://www.ft.com/content/23c70905-147d-4213-8c30-c43e3bde7fec

#AI #AITraining #GenerativeAI #Automation #Productivity

"An anonymous group installed an anti-Elon Musk vending machine that dispenses the Epstein Files at SXSW 2026 in Austin, Texas, on Saturday morning.

Mashable witnessed a group of four men installing the piece of guerrilla protest art at the corner of Red River Street and 4th Street in downtown Austin. After a brief installation, passersby started collecting DVDs titled "Elon's Epstein Files." The vending machine is covered with anti-Elon Musk slogans, and messages on the machine say "Grok makes AI child porn free!" and "Our founder Elon Musk is in the Epstein Files.""

https://mashable.com/article/elon-musk-epstein-files-guerrilla-art-sxsw-2026

#GuerrillaArt #Artivism #Musk #EpsteinFiles #Grok #Austin #SXSW2026

RT @TheHackersNews
🔒 Google is tightening Android’s defenses.

In Android 17 Beta 2, Advanced Protection Mode 🛡️ blocks most apps from accessing the Accessibility Services API.

Malware has long abused it to read screens and steal data.

🔗 Read → https://thehackernews.com/2026/03/android-17-blocks-non-accessibility.html

"National economic planning would be opened up to popular input. At the enterprise or business-firm level, workers would take over management of day-to-day operations. This is all foreign to liberalism but not to social democracies around the world (though the extent of worker management in social democratic countries is still rather limited).

When Sanders broke the national political ice on the word “socialism” in 2016, I recall sappy messages to the effect that “socialism is nice; even your public library is socialist.” But that’s wrong — socialism is so much more than that.

Here I want to describe some economic policy projects that might define a distinctively socialist (or social democratic) approach to policy in the United States today, one that pushes beyond the frontiers of liberalism. These are differences in kind, not of degree. For the principal US social democratic projects, I suggest the following breakdown:

- Labor power
- Industrial policy
- Social insurance
- Social ownership
- Anti-federalism

There are traces of all of these in the history of liberal social policy, but I want to highlight the categorical distinctions between liberal and socialist approaches to each element. Such distinctions can give rise to political themes and to explicit campaigns.

(Socialists also crucially differ from liberals in our commitment to internationalism and our opposition to the United States’ militaristic imperialism. But I focus here on the distinctive elements of socialists’ domestic policy agenda.)"

https://jacobin.com/2026/03/socialism-liberalism-economic-policy-agenda/

#Socialism #USA #EconomicPolicy #PoliticalEconomy #EconomicDemocracy

RT @lukOlejnik
Ex-BND (German Foreign Intelligence Service) deputy chief Arndt Freytag von Loringhoven received a message from fake Signal “support” asking for his PIN. He typed it in.
His contacts then got a malicious link through his hijacked account.
He’s a former NATO intelligence chief, and the author of a book called Putin’s Attack on Germany, where he apparently covers Russian cyberattacks.
He fell for a fake customer service message.

https://www.spiegel.de/politik/deutschland/spionage-ehemaliger-bnd-vize-wird-opfer-von-cyberangriff-a-3fb118d6-b740-4e09-bfa2-6bf67c3fd1e9

"Even worse was the suggestion by Grammarly’s A.I. version of me to replace the first sentence of the news article with an anecdotal opening describing a fictional person named Laura whose privacy had been violated. “Laura, a patient searching for relief from a chronic condition, clicks through her hospital’s website to schedule an appointment. In just a few moments, her most private medical details — her reason for visiting, her doctor’s name and even the treatment she seeks — are quietly sent to Facebook, without her knowledge,” the bot suggested with a button allowing the user to paste that excerpt straight into the article. Replacing a factual sentence with an imagined story about a person who doesn’t exist is not only bad editing. It’s a deception that could end my career as a journalist (or the career of any journalist who took that terrible advice). And this is the problem with A.I. It doesn’t know truth from fiction. It doesn’t know an investigative news article from an offhand comment. It flattens all content into word associations. What Grammarly made wasn’t a doppelgänger. As the writer Ingrid Burrington wrote on Bluesky, it was a sloppelgänger — A.I. slop masquerading as a person. And it must be stopped." https://www.nytimes.com/2026/03/13/opinion/ai-doppelganger-deepfake-grammarly.html #AI #GenerativeAI #Copyright #Writing #Hallucinations

RT @HedgieMarkets
🦔 The Guardian talked to a dozen humanities professors about teaching in the age of AI. Most described the experience in despairing terms. One said generative AI is the bane of her existence. Another said she wishes she could push ChatGPT off a cliff. 92% of students now report using AI for schoolwork. Some professors have resorted to oral exams, handwritten notebooks, and requiring students to submit photos of their notes. One injects random words like "broccoli" into assignments to catch students who paste prompts directly into AI without reading them.

My Take
The thing that stuck with me is the professor who assigned students to visit a museum, look at a painting for ten minutes, and write a few paragraphs about the experience. A student showed up on a Monday when the museum was closed, then turned in an AI-generated reflection anyway. The assignment was designed to be impossible to fake because it was supposed to be personal. It didn't matter.

I don't know what the answer is here. The professors are trying everything they can think of and none of it scales. You can require handwritten work and oral exams but that means smaller classes and more staff, which means more money, which isn't coming. Meanwhile universities are partnering with OpenAI and announcing AI-fluent curriculums while faculty figure it out alone. The worry isn't just cheating. It's that we're running an experiment on an entire generation's ability to think, and nobody's sure what comes out the other side.

Hedgie🤗

https://www.theguardian.com/technology/ng-interactive/2026/mar/10/ai-impact-professors-students-learning

"Judged from an economic perspective, despite the exuberance of the Israeli stock market, the course of the Israeli state is highly questionable. It costs a great deal of money – two billion NIS a day in direct expenditure and five to six billion indirectly – and will require significant continued American financial aid. The government’s logic is that this will be balanced by the economic dividends: sky-rocketing profits from arms sales, now that cutting-edge Israeli weapons are being showcased on the battlefield, not to mention the prospect of Iranian oil reserves and greater access to those of the Gulf states, as they come to realize they need Israel’s protection. Yet there is no certainty this will make up for the financial strain; the same goes for money spent on settlements and the promotion of messianic Judaism in lieu of healthcare and other social priorities.

There are further reasons why Israel will struggle to pursue its strategy over the long term. Campaigns like this in the past were abandoned the moment they faced difficulties. Loss of American life, pressure from other countries in the region, public opinion in the US, the potential resilience of the Iranian regime and continued resistance of the Palestinians may all shift the balance. An invasion of Lebanon, judging by past attempts, will benefit no one. Much depends on the global coalition that fortifies Israel’s wars: the arms industry, multinational corporations, megalomaniac leaders of powerful states, Christian and Jewish Zionist lobbies, the timid governments in the global north as well as corrupt Arab regimes in the Middle East. What is certain is that before this fiasco ends, Israel will inflict a great deal of suffering – on the Iranians, the Lebanese and the Palestinians."

https://newleftreview.org/sidecar/posts/on-the-warpath

#Israel #Zionism #MiddleEast #Iran #Palestine #Lebanon

RT @allenholub
"Gartner predicts that by next year, half of the companies that fired workers for AI are going to hire them back. Also, 9 months ago, Microsoft proudly proclaimed that 30% of their code was written by AI, and since then, we've seen some of the worst software issues at the company in its
history."

This video is well worth watching. It's a balanced, real-world look at the effectiveness of AI (across many disciplines, not just software dev) and its impact on work, substantiated by a well-designed study that actually compares UI to human workers.

https://www.youtube.com/watch?v=z3kaLM8Oj4o

"At an Amazon fulfillment center in Spain, we used a flurry of brief walkouts late last year to force the company to improve wages and time off.

We struck for three days in November and in December in a series of “flexible strikes,” timed to hit production with intermittent walkouts during the holiday “peak” season. On December 22, the union committee announced a settlement, negotiated through government mediators.

The facility, RMU1 in the city of Murcia, employed 2,000 workers at the time, and our union the General Confederation of Labor (CGT) was one of four unions that represented them. [European countries don’t have the same “exclusive representation” system as the U.S., so multiple unions can have a presence at the same worksite. –Editors]

About 75 percent of the workforce, made up of workers from Spain and immigrants from Venezuela, Ecuador, Colombia, and Morocco, participated in the strike, reaching beyond the ranks of the CGT to include other union members.

Our experience shows what’s possible, even at a multinational corporation designed to neutralize organizing. Building from below, workers can organize a well-planned strike—over the objections of more conservative unions—draw on their knowledge of the production process, hit the company where it hurts the most, and wrest real gains.

Here’s how we got Amazon to negotiate with us when it didn’t want to."

https://labornotes.org/amazon-workers-spain-cgt-strategy

#Spain #Amazon #GGT #Murcia #CGT #Labor #WageSlavery #ClassWarfare

"Grammarly’s “expert review” feature offers to give users writing advice “inspired by” subject matter experts, including recently deceased professors, as Wired reported on Wednesday. When I tried the feature out myself, I found some experts that came as a surprise for a different reason — one of them was my boss.

The AI-generated feedback included comments that appeared to be from The Verge’s editor-in-chief, Nilay Patel, as well as editor-at-large David Pierce and senior editors Sean Hollister and Tom Warren, none of whom gave Grammarly permission to include them in the “expert reviews.”

The feature, which launched in August, claims to help you “sharpen your message through the lens of industry-relevant perspectives.” When users select the “expert review” button in the Grammarly sidebar, it analyzes their writing and surfaces AI-generated suggestions “inspired by” related experts. Those “industry-relevant perspectives” include the likes of Stephen King, Neil deGrasse Tyson, and Carl Sagan, among many others.

The Verge found numerous other tech journalists named in the feature, as well, including former Verge editors Casey Newton and Joanna Stern, former Verge writer Monica Chin, Wired’s Lauren Goode, Bloomberg’s Mark Gurman and Jason Schreier, The New York Times’ Kashmir Hill, The Atlantic’s Kaitlyn Tiffany, PC Gamer’s Wes Fenlon, Gizmodo’s Raymond Wong, Digital Foundry founder Richard Leadbetter, Tom’s Guide editor-in-chief Mark Spoonauer, former Rock Paper Shotgun editor-in-chief Katharine Castle, and former IGN news director Kat Bailey. The descriptions for some experts contain inaccuracies, such as outdated job titles, which could have been accurately updated had Superhuman asked those people for permission to reference their work."

https://www.theverge.com/ai-artificial-intelligence/890921/grammarly-ai-expert-reviews

#AI #GenerativeAI #Grammarly

RT @HedgieMarkets
🦔 OpenClaw, the open-source AI agent that exploded to 200,000 GitHub stars in weeks, has become a security nightmare. In five weeks it accumulated 9 disclosed vulnerabilities, over 2,200 malicious add-ons in its marketplace, and 40,000 internet-exposed instances. Researchers found that 93% of those instances had authentication bypassed, and the project triggered 8 of 10 vulnerability classes that security experts warned about for AI agents.

The attack chain works like this: malicious add-ons in the marketplace instruct the AI agent to present fake setup dialogs to users, tricking them into entering passwords. The agent becomes the social engineering tool. One campaign distributed macOS malware by having the agent itself ask users for their credentials. Users trust their AI assistant, so they comply.

My Take
I believe this is what happens when something goes viral before anyone thinks through what they're actually deploying. Developers gave OpenClaw shell access to their computers, connected it to their email and Slack, handed it cloud API keys, and then installed add-ons from a community marketplace that had basically no vetting. Over 40% of the add-ons that got audited had serious security issues. The project went from weekend hack to 200,000 users before anyone built the guardrails.

The attack method here is new. The malware doesn't trick the human directly anymore, it tricks the AI agent into tricking the human. When your assistant asks you for a password to finish an installation, you probably enter it because you trust it. To anyone investigating later, it looks like you voluntarily installed the software. The agent's role is invisible. I've been writing about AI tools being deployed faster than security can keep up, and this is that problem at scale. If anyone at your company has been running OpenClaw, I'd treat it as compromised until proven otherwise.

Hedgie🤗

https://x.com/HedgieMarkets/status/2029337090844946791

"Then, on Tuesday morning, came the big one. YouGov — Britain’s most reputable pollster — put them on 21% of the vote and in second place, just behind Reform: five points ahead of both Labour and the Tories. In all likelihood, Reform and the Greens will have more members than Labour come the local elections in May. And, for those looking to dismiss the Greens as a party of students, bohemians and unemployed artists, think again. YouGov gives them a 6% lead over Reform among all Brits under 65.

“For those looking to dismiss the Greens as a party of students, bohemians and unemployed artists, think again.”

Given the avalanche of attacks and negative campaigning against the Greens in recent weeks, from both the Right and the centre, how is any of this possible? After all, similar attacks paralysed Jeremy Corbyn’s Labour after 2017, and led to a drubbing two years later. Yet Polanski seems to go from strength to strength, leading to panic among some on the Right, with much of Westminster struggling to understand exactly why the Greens are soaring so high."

https://unherd.com/2026/03/how-the-greens-stole-reforms-mojo/

#UK #ReformUK #Farage #GreenParty #Greens

RT @BenjaminNorton
This is incredible: Spain is the only EU country that has opposed the US war of aggression against Iran.

Spain refuses to let the US use its military bases to attack Iran, so Trump threatened an economic embargo against it.

And now the BlackRock executive who runs Germany on behalf of Washington and Wall Street, Friedrich Merz, said he supports Emperor Trump's attack on another EU member.

As an excuse, Merz claimed it's because Spain refuses to increase its military spending to 5% of GDP. But that's not what this is really about. They're threatening to punish Spain for refusing to support a US war of aggression.

European "solidarity" is a myth. The EU is run by vassals of the US empire who sell out their countries to Wall Street.

(By the way, Merz is a multimillionaire who ran the German arm of BlackRock and owns two private jets.)

https://x.com/BenjaminNorton/status/2029006394201186489

It's way easier for lazy politicians to enact laws to outlaw, censor, and surveil under the guise of a moral panic than to enforce antitrust regulations and interoperability and algorithmic transparency mandates...

"While social media bans may seem like a prudent measure to protect children, they are not only ineffective, they endanger both children and adults. There is little evidence that social media is driving any type of widespread mental health crisis in children. Studies have repeatedly shown the opposite. Removing anonymity from the web, which will inevitably happen when tech companies are required to identify and ban children, allows for easier government tracking and censorship of journalists, activists and whistleblowers, who rely on online anonymity.

And while some claim the laws would curb big tech’s power, only the largest tech companies have the resources to shoulder the extensive costs of age verification systems. Non-profit and indie platforms could be forced to close, consolidating big tech’s power further. Mass surveillance systems, once constructed, could also be easily leveraged by governments and bad actors.

If we want to fix the problems with social media, the place to start is through comprehensive data-privacy reform and consumer protections. Governments could also take action to break up big tech companies and prosecute them for anti-competitive behaviour. Lawmakers, who claim to care about children, could pass broader social and economic policies that we know would meaningfully improve children’s lives. Social media is a lifeline, especially for marginalised youth such as LGBTQ+ teens. Any policies that limit online access should centre on the most vulnerable children and adults.

To enact the social media bans being proposed around the world requires some system of age verification, which inherently means expanding surveillance technology."

https://www.theguardian.com/commentisfree/2026/mar/02/ban-children-social-media-biometic-data-surveilled

#AgeVerification #SocialMedia #Politics #Censorship #Surveillance

"According to a source familiar with the negotiations, on Friday morning, Anthropic received word that Hegseth’s team would make a major concession. The Pentagon had kept trying to leave itself little escape hatches in the agreements that it proposed to Anthropic. It would pledge not to use Anthropic’s AI for mass domestic surveillance or for fully autonomous killing machines, but then qualify those pledges with loophole-y phrases like as appropriate—suggesting that the terms were subject to change, based on the administration’s interpretation of a given situation.

Anthropic’s team was relieved to hear that the government would be willing to remove those words, but one big problem remained: On Friday afternoon, Anthropic learned that the Pentagon still wanted to use the company’s AI to analyze bulk data collected from Americans. That could include information such as the questions you ask your favorite chatbot, your Google search history, your GPS-tracked movements, and your credit-card transactions, all of which could be cross-referenced with other details about your life. Anthropic’s leadership told Hegseth’s team that was a bridge too far, and the deal fell apart. Soon after, Hegseth directed the U.S. military’s contractors, suppliers, and partners to stop doing business with Anthropic. The list of companies that contract with the military is extensive, and includes Amazon, the company that supplies much of Anthropic’s computing infrastructure. The Department of Defense did not respond to a request for comment. A spokesperson for Anthropic referred me to the company’s statement addressing Hegseth’s remarks."

https://www.theatlantic.com/technology/2026/03/inside-anthropics-killer-robot-dispute-with-the-pentagon/686200/

#USA #Trump #Pentagon #DoD #DroneWarfare #AI #GenerativeAI #Anthropic #Claude

RT @ChrisRMcGuire
Reuters reports that an object struck an AWS data center in the UAE, causing a fire and shutting it down. Assuming this was an Iranian drone strike, it is the first time a commercial data center was physically targeted in a conflict. It won’t be the last.

https://www.reuters.com/world/middle-east/amazons-cloud-unit-reports-fire-after-objects-hit-uae-data-center-2026-03-01/

RT @lukOlejnik
Let’s speak about the future of analytical assessment using AI Agents. A week before the strikes, on February 22, my AI agent
@openclaw assessed a ~70-75% probability of a US strike on Iran within two weeks. It identified five escalation levels already underway and predicted the attack would come.
What it got right:
Timing. Strikes landed on February 28, six days into the forecast window. The bot flagged the next 48-72 hours as critical for an Iranian counter-offer. Iran’s offer didn’t deliver. Strikes followed.
Escalation model. The bot described five levels. Levels 1-2 (cyber and information operations) were already active. Today we know Israel (reportedly) launched large cyber operations alongside the kinetic strikes, exactly as the bot anticipated. The actual operation landed between the bot’s Level 3 (limited strikes) and Level 4 (sustained air campaign), with the Pentagon saying operations would last “a few days.”
Indicators. The bot cited three converging layers: military (personnel withdrawals from Gulf bases, identical pattern to June 2025), diplomatic (25+ countries issuing evacuation notices, UK embassy closed), and political (Trump advisors signaling 90% chance of kinetic action). All three held.
Targeting. The bot predicted nuclear and military infrastructure. The actual strikes hit government, military, intelligence, and IRGC targets, with regime decapitation added on top.
What it missed. The bot didn’t predict the m attempt to kill Khamenei and Pezeshkian directly. It also estimated invasion probability at under 2%, which remains correct so far. It flagged UK base restrictions as a complication for sustained bombing, but the operation proceeded regardless.
70-75% one week out, on a six-day fuse. The bot read the indicators correctly.​​​​​​​​​​​​​​​​

https://clawdint.com/

"We found that reliability has improved only modestly over 18 months, while accuracy improved substantially. All three major providers cluster together, so this appears to be an industry-wide limitation (though there are some cases where Anthropic’s models outperform OpenAI’s and Google’s).

More specifically, we measured the following criteria:

Consistency: Agents that can solve a task often fail on repeated attempts under identical conditions. Many models have trouble giving a consistent answer, with outcome consistency scores ranging from 30% to 75% across the board.

Robustness: Most models handle genuine technical failures (server crashes, API timeouts) gracefully. But if we rephrase the instructions with the same semantic meaning, performance drops substantially.

Predictability: Agents are not good at knowing when they’re wrong. This is the weakest dimension across the board. When agents report confidence, it often carries little signal. On one benchmark, most models couldn’t distinguish their correct predictions from incorrect ones better than chance.

Safety: Recent models are noticeably better at avoiding constraint violations, though financial errors, such as incorrect charges, remain a common failure mode. We use safety narrowly to mean bounded harm when failures occur, not broader concerns like alignment. We are still iterating on how we measure safety, so we report it separately from the aggregate reliability score.

Impact of scaling: Bigger models aren’t uniformly more reliable. Scaling up improves some dimensions (calibration, robustness) but can hurt consistency. Larger models with richer behavioral repertoires sometimes show more run-to-run variability.

Our view is that reliability lags capability, and that reliability will remain a barrier to deployment unless researchers and developers focus effort on improving reliability as a separate dimension from accuracy."

https://www.normaltech.ai/p/new-paper-towards-a-science-of-ai

#AI #GenerativeAI #LLMs #AISafety #Science #Reliability

Go headless:

"If you absolutely have to ship UI as part of your project, the best way to approach it is to make the UI optional and modular. Here’s how you can do it:

- Separate UI and Core Logic: Create two distinct packages: one for the core functionality and one for the UI components. This allows developers to choose whether they want to use the provided UI or build their own while still leveraging the core logic of the SDK.
- Provide Clear Documentation: Offer detailed instructions for implementing the SDK with and without the built-in UI. Include code samples and guidelines for developers who want to customize or replace the default UI.
- Make the UI Customizable: If you include a UI package, ensure it is fully customizable. Allow developers to override styles, colors, fonts, and even layouts to match their app’s branding and design.
- Design for Integration: Ensure the UI components follow Android’s Material Design guidelines and can adapt seamlessly to different themes, orientations, and screen sizes. Use isolated namespaces to avoid resource conflicts.
- Support Analytics Hooks: Provide APIs or callbacks that allow developers to integrate their analytics and tracking solutions into the SDK’s UI. This ensures that they can still collect data and maintain insights into user behavior.
- Offer a “Headless Mode”: For advanced users, offer a “headless” mode that exposes only the core logic, enabling developers to integrate it into their own UI without relying on your SDK’s visuals.

An ideal SDK with UI should have at least 2 importable modules, 1 for the core functionality, and one for UI."

https://proandroiddev.com/sdk-development-the-good-the-bad-the-ugly-9e9ab2a81697

#SDK #SDKDevelopment #HeadlessSDK #Programming #SoftwareDevelopment #DeveloperExperience

"The current fad is to create Claude Skills, Copilot instructions, Cursor rules, and so on, so that developers and users can use your products through LLMs in a way that’s easier than just feeding raw documentation as context. In a way, this is quite similar to packaging docs as if they were Nintendo cartridges or floppy disks with some cool cheat codes for a game: docs-as-data. You plug docs in, the agent learns kung-fu.

Seeing developers coming up with agentic helpers fed by docs whose provenance, quality, and maintenance is unknown, eager to just use context as some sort of agentic boilerplate, makes me upset. As a tech writer you should own the instructions, as you should own the words in a REST API or the prompts in an MCP server. All that is docs. It doesn’t matter that it’s going to be consumed by AI. It’s still docs. It’ll always be docs.

Start creating skills to automate and enhance your own work. We call them “skills” today, but they could be “agentic docs” tomorrow. "

https://passo.uno/new-habits-tech-writers-ai-age/

#TechnicalWriting #Automation #AI #LLMs #GenerativeAI #Programming #SoftwareDevelopment

"The Cuban Revolution has survived two-thirds of a century under US blockade but is today under greater pressure than ever. US government actions including the January 3 kidnapping of Nicolás Maduro, assaults on Caribbean sea-lanes, and a January 29 executive order imposing steep tariffs on states that supply oil to Cuba are all aimed at cutting off Cuba’s fuel supply and bringing the country to its knees.

Many reports tell of a dire humanitarian situation, with now-frequent power cuts and widespread shortages, worse even than in past crisis moments. Donald Trump and anti-communist ideologues like Secretary of State Marco Rubio seem determined to push Cuba into chaos. Threats against Cuba’s trade partners are designed to give the United States the final say over the island’s fate.

Still, Cuba does not stand alone. The Nuestra América Convoy has called for an international solidarity effort to bring humanitarian aid to the island on March 21. One of the organizers is David Adler, co–general coordinator of the Progressive International, and a veteran of the Gaza Freedom Flotilla. He spoke to Jacobin’s David Broder about the US pressure on Cuba, its ripping up of international law, and the need for practical solidarity with the Cuban people."

https://jacobin.com/2026/02/cuba-trump-nuestra-america-convoy

#USA #Trump #Cuba #Imperialism #NuestraAmerica

Large-scale online deanonymization with LLMs

"We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator. We then design attacks for the closed-world setting. Given two databases of pseudonymous individuals, each containing unstructured text written by or about that individual, we implement a scalable attack pipeline that uses LLMs to: (1) extract identity-relevant features, (2) search for candidate matches via semantic embeddings, and (3) reason over top candidates to verify matches and reduce false positives. Compared to prior deanonymization work (e.g., on the Netflix prize) that required structured data or manual feature engineering, our approach works directly on raw user content across arbitrary platforms. We construct three datasets with known ground-truth data to evaluate our attacks. The first links Hacker News to LinkedIn profiles, using cross-platform references that appear in the profiles. Our second dataset matches users across Reddit movie discussion communities; and the third splits a single user’s Reddit history in time to create two pseudonymous profiles to be matched. In each setting, LLM-based methods substantially outperform classical baselines, achieving up to 68% recall at 90% precision compared to near 0% for the best non-LLM method. Our results show that the practical obscurity protecting pseudonymous users online no longer holds and that threat models for online privacy need to be reconsidered."

https://arxiv.org/html/2602.16800v1

#AI #GenerativeAI #LLMs #Anonymity #Privacy #Deanonymization

I think I can now totally agree with this argument. It's very myopic to think that a bunch of AI agents can develop and maintain a fully fledged SaaS platform while maintaining a production-ready level of quality for thousands of users around the world...

"I’ve worked in or around SaaS since 2012, and I know the industry well. I may not be able to code, but I take the time to speak with software engineers so that I understand what things actually do and how “impressive” they are. Similarly, I make the effort to understand the underlying business models in a way that I’m not sure everybody else is trying to, and if I’m wrong, please show me an analysis of the financial condition of OpenAI or Anthropic from a booster. You won’t find one, because they’re not interested in interacting with reality.

So, despite all of this being very obvious, it’s clear that the markets and an alarming number of people in the media simply do not know what they are talking about or are intentionally avoiding thinking about it. The “AI replaces software” story is literally “Anthropic has released a product and now the resulting industry is selling off,” such as when it launched a cybersecurity tool that could check for vulnerabilities (a product that has existed in some form for nearly a decade) causing a sell-off in cybersecurity stocks like Crowdstrike
(...)
There is no rational basis for anything about this sell-off other than that our financial media and markets do not appear to understand the very basic things about the stuff they invest in. Software may seem complex, but (especially in these cases) it’s really quite simple: investors are conflating “an AI model can spit out code” with “an AI model can create the entire experience of what we know as ‘software,’ or is close enough that we have to start freaking out.”"

https://www.wheresyoured.at/on-nvidia-and-analyslop/

#AI #GenerativeAI #LLMs #AIAgents #AIBubble #Nvidia #Programming #SoftwareDevelopment

RT @TheHackersNews
Attackers are stealing encrypted data under a “Harvest Now, Decrypt Later” strategy.

Store it now. Decrypt it when quantum machines mature, possibly between 2030 and 2035. Security Navigator 2026 outlines a five-step PQC migration plan and breach data.

🔗 Read → https://thehackernews.com/2026/02/expert-recommends-prepare-for-pqc-right.html

RT @lukOlejnik
My comments for @Telegraph on autonomous weapons, AI and the Pentagon standoff with @Anthropic. The US has no binding prohibition on lethal autonomous weapon systems. AI-powered weapons such as mass drone swarms and loitering munitions that would only require a human in an observer role are close to reality.

https://www.telegraph.co.uk/business/2026/02/26/the-one-man-stopping-trump-from-building-killer-ai/

RT @jasminewsun
200+ Google and OpenAI staff have signed this petition to share Anthropic's red lines for the Pentagon's use of AI

let's find out if this is a race to the top or the bottom

https://notdivided.org

"With a hammer, harm is external and obvious, like a bent nail, a bruised thumb. In this cognitive environment, harm can be real and insidiously subtle. Once an LLM has offered a line of reasoning or even a clever turn of phrase, it becomes part of your mental terrain. You don’t just receive information; you adopt it as a starting point for your next thought. And that's what makes AI something other than just a tool...

Certainly, this is powerful, even exhilarating. The environment can change your cognitive perspective and trajectory. But it also carries a hidden cost or even accumulates a type of debt. It can smooth away the very frictions that make human thought generative. Confusion, hesitation, and error are not accidents of cognition—they are the mechanisms by which we refine our understanding. They are what turn information into meaning."

https://www.psychologytoday.com/us/blog/the-digital-self/202509/maybe-ai-was-never-a-tool

#AI #GenerativeAI #CriticalThinking #Automation

"The United States is waging a new cold war against China. This has been openly admitted in Washington for several years.

The First and Second Cold Wars are different in some significant ways. The ideological split is not exactly the same.

The United States is capitalist and constitutes the heart of the capitalist world-system, whereas China is socialist. However, the People’s Republic of China is not the Soviet Union; it does not lead a socialist bloc of countries, and Beijing has been clear that it does not seek to “export” revolution.

“We will not import other countries’ models, and will not export the China model”, President Xi Jinping asserted in 2017 — although he added, “We will provide more opportunities for the world through our development”.

With that established, it should be stressed that just because China is very different from the USSR does not mean that there is no ideological aspect of Cold War Two.

There are unambiguous ideological differences between the US and China, and each promotes a very different vision of international relations.

The Second Cold War, therefore, will still have an enormous impact on the new global order that is being shaped.

In short, the political model that Washington seeks to impose on the world is the exact opposite of the political model being advocated by Beijing."

https://www.geopoliticaleconomy.report/p/us-unipolarity-china-multipolarity-vision-global-order

#Multilateralism #Unipolarity #Multipolarity #USA #China #Imperialism

"New research shows that behaviors that occur at the very lowest levels of the network stack make encryption—in any form, not just those that have been broken in the past—incapable of providing client isolation, an encryption-enabled protection promised by all router makers, that is intended to block direct communication between two or more connected clients.

The isolation can effectively be nullified through AirSnitch, the name the researchers gave to a series of attacks that capitalize on the newly discovered weaknesses. Various forms of AirSnitch work across a broad range of routers, including those from Netgear, D-Link, Ubiquiti, Cisco, and those running DD-WRT and OpenWrt.

AirSnitch “breaks worldwide Wi-Fi encryption, and it might have the potential to enable advanced cyberattacks,” Xin’an Zhou, the lead author of the research paper, said in an interview. “Advanced attacks can build on our primitives to [perform] cookie stealing, DNS and cache poisoning. Our research physically wiretaps the wire altogether so these sophisticated attacks will work. It’s really a threat to worldwide network security.” Zhou presented his research on Wednesday at the 2026 Network and Distributed System Security Symposium.

Paper co-author Mathy Vanhoef, said a few hours after this post went live that the attack may be better described as a Wi-Fi encryption “bypass,” “in the sense that we can bypass client isolation. We don’t break Wi-Fi authentication or encryption. Crypto is often bypassed instead of broken. And we bypass it ;)” People who don’t rely on client or network isolation, he added, are safe."

https://arstechnica.com/security/2026/02/new-airsnitch-attack-breaks-wi-fi-encryption-in-homes-offices-and-enterprises/

#CyberSecurity #Wifi #Encryption #AirSnitch

"AI tools are making potentially harmful errors in social work records, from bogus warnings of suicidal ideation to simple “gibberish”, frontline workers have said.

Keir Starmer last year championed what he called “incredible” time-saving social work transcription technology. But research across 17 English and Scottish councils shared with the Guardian has now found AI-generated hallucinations are slipping in.

As scores of local authorities begin to use AI note-takers to accelerate recording and summarisation of meetings with adult and child service users, a seven-month study by the Ada Lovelace Institute found “some potentially harmful misrepresentations of people’s experiences are occurring in official care records”.

The independent thinktank found that one social worker who had used an AI transcription tool to create a summary said the technology had incorrectly “indicated that there was suicidal ideation”, but “at no point did the client actually … talk about suicidal ideation or planning, or anything”."

https://www.theguardian.com/education/2026/feb/11/ai-tools-potentially-harmful-errors-social-work

#AI #GenerativeAI #AITranscription #SocialWork #UK #Hallucinations

"A software engineer’s earnest effort to steer his new DJI robot vacuum with a video game controller inadvertently granted him a sneak peak into thousands of people’s homes. While building his own remote-control app, Sammy Azdoufal reportedly used an AI coding assistant to help reverse-engineer how the robot communicated with DJI’s remote cloud servers. But he soon discovered that the same credentials that allowed him to see and control his own device also provided access to live camera feeds, microphone audio, maps, and status data from nearly 7,000 other vacuums across 24 countries. The backend security bug effectively exposed an army of internet-connected robots that, in the wrong hands, could have turned into surveillance tools, all without their owners ever knowing. Luckily, Azdoufal chose not to exploit that. Instead, he shared his findings with The Verge, which quickly contacted DJI to report the flaw. While DJI tells Popular Science the issue has been “resolved,” the dramatic episode underscores warnings from cybersecurity experts who have long-warned that internet-connected robots and other smart home devices present attractive targets for hackers." https://www.popsci.com/technology/robot-vacuum-army/ #AI #IoT #CyberSecurity #DJII #RobotVaccum

"At the heart of the Chatrie case are legal orders known as geofence warrants. This controversial tool allows police to demand location data from tech companies (usually Google) to see every device in a specific area at a specific time. Imagine drawing a digital fence around a crime scene and demanding a list of every phone that crossed into it.

These demands can reveal precise details about people’s movements and locations. Authorities can pinpoint where someone stood within a couple of yards and whether they were on the first or second floor of a building.

But geofence warrants are also imprecise: They sweep up the movements not just of suspects but also of innocent people who happen to be within the digital fence. Demanding location data for a 150-yard radius of a bank in the hour before it was robbed, for example, may show the movements of people who worked at the bank, visited the psychiatrist’s office next door, worshipped at the church on the neighboring block, or dropped into the nearby strip club."

https://freedom.press/issues/supreme-court-could-greenlight-geofence-warrants/

#USA #PressFreedom #Journalism #Surveillance #Geolocation #GeofenceWarrants

RT @gabriel_zucman
The pace at which US wealth concentration is rising is simply staggering

The concentration of AI wealth into the hands of a few tech barons + plutocratic capture ==> unchartered territory

https://x.com/gabriel_zucman/status/2023046025162940692

RT @jasonhickel
What's most repulsive about Rubio's speech, and the warm embrace it received from the European ruling class, is that its only vision for "Western civilization" is a future of endless imperialist violence.

They have nothing else to offer to their citizens, or to the world. No vision for addressing social and ecological crises, no vision for improving people's lives, no vision for human progress... no other vision for "greatness" besides violence and plunder.

It is, in fact, the antithesis of civilization. It is barbarism.

https://x.com/jasonhickel/status/2023435161606447542

RT @RnaudBertrand
2 years ago I wrote that humanoid robots will end up being "THE product that will symbolize China's rise to world preeminent power status."

Look at what humanoids in China are capable of now (this was the Chinese new year gala last night, https://x.com/WhileTravelling/status/2023416426984288506/video/1) 👇

Most impressively, you can buy one of these Unitree robots TODAY for $13.5k (https://shop.unitree.com/products/unitree-g1?srsltid=AfmBOoqbvG6ilxJx2DVCuh4nhzGz6nwupiNWbkqqzTgg4ZJhZ0d-_fgC), which isn't the case for any US competitor: for all the hype Elon's Optimus isn't remotely ready for commercialization (Elon says it'll need another 2 years: https://axios.com/2026/01/22/elon-musk-tesla-optimus-robots).

Which means that this is an industry, which will end up being one of the largest industries ever (dixit Jensen Huang: https://instagram.com/reel/DK0agCDS1VQ/), in which China is easily 3 years ahead of the US in terms of technology (given that Unitree humanoids, among others, have been mass produced for more than one year).

Which is a lifetime at the pace at which technology moves, and a gap that continuously widens given that China's humanoid field is more dynamic than the US: the city of Shenzhen alone, with 8 humanoid robot companies (https://x.com/i/status/1991734168896827817), outcompetes the entire US industry today.

The only thing, funnily, in which US robotics companies outcompete China is market capitalization. For instance Figure AI, which - like Optimus - has yet to commercialize a single humanoid, is valued at $39B (https://figure.ai/news/series-c) when Unitree is valued at $1.6B (https://eu.36kr.com/en/p/3344368397190018).

Which shows again the extent to which tech valuations are divorced from reality - in the US and China both, just in opposite directions.

https://x.com/RnaudBertrand/status/2023606882116812924

RT @HedgieMarkets
🦔 Global data center power demand is projected to hit 1,596 terawatt-hours by 2035, a 255% increase from 2025 levels. The US leads with demand expected to surge 144% to 430 terawatt-hours. China is projected to rise 255% to 397 terawatt-hours, while Europe jumps 303% to 274 terawatt-hours.

New data centers coming online between now and 2030 alone will need more than 600 terawatt-hours of electricity, enough to power roughly 60 million homes.

My Take
The numbers are staggering, but what's already happening to electricity bills is more immediate. Wholesale electricity costs have jumped as much as 267% in areas near data centers compared to five years ago. In the mid-Atlantic region, capacity charges rose 833% for 2025-2026, and an independent monitor found data center demand made up 63% of the total power capacity bill. Electricity prices rose 6.9% in 2025, more than double the inflation rate, and Goldman Sachs says prices will keep climbing through the end of the decade.

Here's what bothers me. A Yale Climate Connections analysis found that between 2020 and 2024, residential electricity prices increased 25% while data centers and commercial users saw only modest increases. Industrial users actually paid lower prices than two years ago. Households are subsidizing the AI buildout through higher bills while the companies driving the demand negotiate discounts. Virginia and New Jersey just elected governors who campaigned partly on this issue. States are now scrambling to pass laws making data centers pay their own way, but those rules can't fix the short-term problem of demand outpacing supply. The infrastructure costs are already being passed to consumers.

Hedgie🤗

https://x.com/HedgieMarkets/status/2023559772398841980

É óbvio que as pessoas têm todo o direito a recusar-se a trabalhar em empresas que obrigam os seus funcionários a utilizar ferramentas de Inteligência Artificial (IA). Parece-me igualmente óbvio que essas empresas estão a cometer uma verdadeira idiotice se necessitam de obrigar os seus funcionários a utilizar IA.

Na verdade, isso é a PIOR FORMA de promover a adopção de ferramentas de IA. Aliás, a mensagem implícita nessa obrigação é essa: Essas ferramentas são tão más, tão ineficientes e tão improdutivas que é preciso obrigar as pessoas a utilizarem-nas nas suas tarefas de trabalho.

Seja como for, muito além dessa discussão maniqueísta típica dos EUA entre prós-IA e anti-IA - discussão essa que, a bem de verdade, me parece totalmente desinteressante por se agarrar a um discurso generalizante e abstracto, desvinculado de cenários de utilização real e ferramentas/modelos concretos -, parece-me bem mais interessante recorrer a metáforas históricas com tecnologias que surgiram anteriormente.

E as metáforas que neste momento me ocorrem são duas. A primeira remete para a altura em que tiveram início os voos regulares comerciais atravessando o mesmo continente, por exemplo entre Nova Iorque e Los Angeles ou, a nível europeu, entre Lisboa e Berlim.

A segunda remete para o início da Internet de banda larga, via ADSL. Ou muito me engano ou parece-me que desde o final do ano passado, com a chegada do Claude Opus 4.5, chegámos a uma etapa da história da evolução dos LLMs muito semelhante a esses dois pontos históricos.

Mesmo com os primeiros voos comerciais regulares a atravessar a América do Norte ou a Europa, muita gente continuou a preferir fazer essa viagem de comboio ou mesmo de automóvel privado. Mas com os anos, muito mais gente passou a considerar que lhes convinha mais poupar tempo numa viagem de avião, mesmo que isso significasse não poder apreciar a paisagem ou parar pelo caminho.

"With a determination underappreciated in Trump’s Washington, Xi decided long ago never to let China’s AI sector become “addicted” to American technology. He sees the risks of dependence as outweighing the benefits; any lead in AI that isn’t grounded in China producing its own physical components is too fragile to be worth having. And he believes that Chinese tech firms can innovate even if they do not have topflight American chips, in keeping with his long-standing view that periods of hardship make China stronger. It may seem contradictory, but because AI is so important to Xi—he has called technology “the main battlefield of international competition”—he wants to force the country’s tech firms to create a self-reliant system, even if it means slowing their near-term progress, in order to gain a more durable long-term advantage. Even as Trump approved sales of the H200, Xi considered providing as much as $70 billion in additional support for China’s chip industry, in part to accelerate his effort to produce advanced semiconductor manufacturing machinery as good as that of the global leader, the Dutch firm ASML. But this equipment, often called “the world’s most complex machine,” may be the single most challenging technology to replicate domestically, far harder than solar panels, electric vehicles, or other Chinese success stories. It is a high-stakes gamble.

To understand why Xi is making these decisions, there is no better guide than the Harvard sociologist Ya-Wen Lei’s The Gilded Cage: Technology, Development, and State Capitalism in China. Her masterful study of what she calls China’s regime of “techno-development”—the model of “[science-and-technology]-oriented socioeconomic development” that has gradually replaced the prior model of “labor-intensive, export-oriented manufacturing”—makes clear that China’s leaders see advanced technology as the country’s lifeblood:"

https://www.nybooks.com/articles/2026/02/26/when-the-chips-are-down-china-us-ai/

#China #USA #AIChips #Nvidia #AI #Huawei #Tariffs #TradeWar