I Will Fucking Piledrive You If You Mention AI Again
(link)
So it is with great regret that I announce that the next person to talk about rolling out AI is going to receive a complimentary chiropractic adjustment in the style of Dr. Bourne, i.e, I am going to fucking break your neck. I am truly, deeply, sorry.

Pourquoi je n’utilise pas ChatGPT
(link)
Plus le temps passe, moins je suis tentée d’utiliser ChatGPT ou d’autres outils d’IA générative. Le rythme effréné des annonces et la vision du monde des promoteurs de ces outils m’ont définitivement vaccinée contre le moindre frémissement d’intérêt qui aurait pu subsister. Et je n’ai même pas abordé ici les questions de biais, de sécurité, de protection de la vie privée, …

Réjouissance
(link)
Je vois une régression de la pensée et une manipulation à grande échelle rendue tellement aisée que pas grand monde ne pourra y résister. C’est du Cambridge Analytica on cocAÏne.

Programming Still Sucks.
(link)
The Guy at the party is still waiting for an answer. I'm too drunk now to lie. I tell him: AI didn't take our jobs. Greed did. Same greed that moved factories to Bangladesh and keeps slaves in cobalt mines in the Congo, wearing a new mask. Tell the nephew to do something else. Anything. It won't save him either, but at least he won't have to pretend the thing destroying his life is a robot.

Taking AI Doom Seriously For 62 Minutes
(link)
I can't really tell you how to weight all these things, there is a lot of uncertainties in all of them, but I'll give you my overall hot take which is that (...) may be we could just slown down a little bit and concentrate on understanding on how these things actually work that we are making, and also making sure that we can control them.

Have I hardened against LLMs?
(link)
The more I wrote about generative models, the more appalled I became at the response from the industry, to both my writing and that of others actively highlighting the risks. Few people who have any influence in tech and software seem to care about the harms, the political manipulation, the outright sabotage of education, the association with extremism, or the literal child abuse.

The West Forgot How to Make Things. Now It’s Forgetting How to Code
(link)
When juniors skip debugging and skip the formative mistakes, they don’t build the tacit expertise. And when my generation of engineers retires, that knowledge doesn’t transfer to the AI. It just disappears.

AI Search Has a Citation Problem. We compared eight AI search engines. They’re all bad at citing news.
(link)
The findings of this study align closely with those outlined in our previous ChatGPT study, published in November 2024, which revealed consistent patterns across chabots: confident presentations of incorrect information, misleading attributions to syndicated content, and inconsistent information retrieval practices. Critics of generative search like Chirag Shah and Emily M. Bender have raised substantive concerns about using large language models for search, noting that they “take away transparency and user agency, further amplify the problems associated with bias in [information access] systems, and often provide ungrounded and/or toxic answers that may go unchecked by a typical user.”

Contributor Poker and Zig's AI Ban
(link)
For us the ability to provide contributors with an engaging ecosystem where they can improve their systems thinking and interact with other competent, trusted and prolific engineers is a critical aspect of our business model.

The machines are fine. I'm worried about us.
(link)
The problem isn't that we'll decide to stop thinking. The problem is that we'll barely notice when we do.

Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender
(link)
A key prediction of the theory is "cognitive surrender"-adopting AI outputs with minimal scrutiny, overriding intuition (System 1) and deliberation (System 2).

Adults Lose Skills to AI. Children Never Build Them.
(link)
Adults who offload thinking to AI lose capacity they built. Children may never build it at all. When students process information through the same model, the result may be similar minds. Auditing AI output requires expertise the child is still supposed to be developing. In a study, developers who delegated coding to AI produced working code but failed conceptual understanding.

Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task
(link)
Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI's role in learning.

The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers
(link)
Moreover, while GenAI can improve worker efficiency, it can inhibit critical engagement with work and can potentially lead to long-term overreliance on the tool and diminished skill for independent problem-solving.

GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation
(link)
Academic journals, archives, and repositories are seeing an increasing number of questionable research papers clearly produced using generative AI. They are often created with widely available, general-purpose AI applications, most likely ChatGPT, and mimic scientific writing. Google Scholar easily locates and lists these questionable papers alongside reputable, quality-controlled research. Our analysis of a selection of questionable GPT-fabricated scientific papers found in Google Scholar shows that many are about applied, often controversial topics susceptible to disinformation: the environment, health, and computing. The resulting enhanced potential for malicious manipulation of society’s evidence base, particularly in politically divisive domains, is a growing concern.

Agentic Coding is a Trap. Remaining vigilant about cognitive debt and atrophy.
(link)
Despite the countless failed attempts at trying to democratize coding while not understanding coding, we're faced with the reality that you cannot understand code without engaging with it. And it's become clear that if you don't keep engaging and writing it, you can lose touch with that understanding, which will in turn make you a less capable orchestrator in the first place, rendering this phase of AI coding a strange and needlessly stressful interlude.

How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt
(link)
As generative and agentic AI accelerate development, protecting that shared theory of what the software does and how it can change may matter more for long-term software health than any single metric of speed or output.

Let’s talk about LLMs
(link)
Not only is there no silver bullet, there especially is no quick or magical gain to be had from rushing to adopt LLM coding without first working on those fundamentals. In fact, the evidence we have says you’re more likely to hurt than help your productivity by doing so.

I finally turned off GitHub Copilot yesterday.
(link)
So, after giving it a fair try, I have concluded that it is both a net decrease in productivity and probably an increase in legal liability.

Mirage: The Illusion of Visual Understanding
(link)
First, Frontier models readily generate detailed image descriptions and elaborate reasoning traces, including pathology-biased clinical findings, for images never provided; we term this phenomenon mirage reasoning. Second, without any image input, models also attain strikingly high scores across general and medical multimodal benchmarks, bringing into question their utility and design. (...) Third, when models were explicitly instructed to guess answers without image access, rather than being implicitly prompted to assume images were present, performance declined markedly. (...) These findings expose fundamental vulnerabilities in how visual–language models reason and are evaluated, pointing to an urgent need for private benchmarks that eliminate textual cues enabling non-visual inference, particularly in medical contexts where miscalibrated AI carries the greatest consequence.

Traditional models still ‘outperform AI’ for extreme weather forecasts
(link)
The authors tested how well both AI and traditional weather models could simulate thousands of record-breaking hot, cold and windy events that were recorded in 2018 and 2020. They find that AI models underestimate both the frequency and intensity of record-breaking events. A study author tells Carbon Brief that the analysis is a “warning shot” against replacing traditional models with AI models for weather forecasting “too quickly”.

AI layoffs backfire as cutting staff doesn't cut it, firms warned
(link)
Replacing meatbags with failure prone agents isn't the gold mine some CEOs hoped for

We did the math on AI’s energy footprint. Here’s the story you haven’t heard.
(link)
By 2028, the researchers estimate, the power going to AI-specific purposes will rise to between 165 and 326 terawatt-hours per year. That’s more than all electricity currently used by US data centers for all purposes; it’s enough to power 22% of US households each year. That could generate the same emissions as driving over 300 billion miles—over 1,600 round trips to the sun from Earth.

Officials hugely underestimated impact of AI datacentres on UK carbon emissions
(link)
To waste what little bandwidth we have left – when 750 million people worldwide lack access to electricity – assisting some of the richest men ever to hone their plagiarism bots would be a historic idiocy that future generations are unlikely to forgive today’s leaders for.

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
(link)
We have identified a wide variety of costs and risks associated with the rush for ever larger LMs, including: environmental costs (borne typically by those not benefiting from the resulting technology); financial costs, which in turn erect barriers to entry, limiting who can contribute to this research area and which languages can benefit from the most advanced techniques; opportunity cost, as researchers pour effort away from directions requiring less resources; and the risk of substantial harms, including stereotyping, denigration, increases in extremist ideology, and wrongful arrest, should humans encounter seemingly coherent LM output and take it for the words of some person or organization who has accountability for what is said.

The Low-Paid Humans Behind AI’s Smarts Ask Biden to Free Them From ‘Modern Day Slavery’
(link)
A typical workday for African tech contractors, the letter says, involves “watching murder and beheadings, child abuse and rape, pornography and bestiality, often for more than 8 hours a day.” Pay is often less than $2 per hour, it says, and workers frequently end up with post-traumatic stress disorder, a well-documented issue among content moderators around the world.

Flaws in Kenya’s AI-driven health reforms driving up costs for the poorest
(link)
An AI system used to predict how much Kenyans can afford to pay for access to healthcare, has systemically driven up costs for the poor, an investigation has found.

The AI Layoff Trap
(link)
If AI displaces human workers faster than the economy can reabsorb them, it risks eroding the very consumer demand firms depend on. We show that knowing this is not enough for firms to stop it. In a competitive task-based model, demand externalities trap rational firms in an automation arms race, displacing workers well beyond what is collectively optimal. The resulting loss harms both workers and firm owners.

AI Slop Is Polluting Bug Bounty Platforms with Fake Vulnerability Reports
(link)
This could easily kill the whole concept of bug bounties," he said. "Why? Genuine researchers quit in frustration as they don't get proper reward for their hard work, and see AI slop scoop the money. Orgs/projects abandon bug bounty programs since they get mostly AI Slop reports. Financial backing (as donations or investment) for bug bounty programs disappears as the money is paid to scammers.

AI bots are destroying Open Access
(link)
We are headed for a world in which all good information is locked up behind secure registration barriers and paywalls, and it won't be to make money, it will be for survival.

The physics slop that YouTube wants me to make
(link)
The use that we see here doesn't seem to align with good intentions. (...) And I wonder if that use can be actually harmful.

Word frequency tool ‘wordfreq’ stops updates, overwhelmed by AI spam
(link)
“The world where I had a reasonable way to collect reliable word frequencies is not the world we live in now,” says author Robyn Speer.

Amazon restricts authors from self-publishing more than three books a day after AI concerns
(link)
The new sets of rules come after Amazon removed suspected AI-generated books that were falsely listed as being written by the author Jane Friedman. Earlier this month, books about mushroom foraging listed on Amazon were reported as likely to have been AI-generated and therefore containing potentially dangerous advice. AI-generated travel books have also flooded the site.

Google Gemini tried to kill me.
(link)
Turns out I had just grew a botulism culture and garlic in olive oil specifically is a fairly common way to grow this bio-toxins. Had I not checked on it 3-4 days in I'd have been none the wiser and would have Darwinned my entire family. Prompt with care and never trust AI dear people...

South Africa yanks AI policy after AI-assisted drafting invents citations
(link)
All in all, it's a great look for a government trying to set the rules on AI when its own policy can't clear a basic fact check. And it's not exactly a one-off either. As The Register reported last year, Deloitte had to help clean up a government report in Australia after AI-generated citations and even a made-up court quote slipped through, a reminder that letting the machine do the writing is one thing, checking it is another.

Radical Right-Wing Political Deepfakes Can Successfully Delegitimize Targeted Political Actors: Evidence From Three-wave Experiments in the US and The Netherlands
(link)
Despite the relatively low credibility of deepfakes, however, our findings show that deepfakes may be successful in lowering support for the targeted political actor in both a bi-partisan and multiparty political context. Thus, even when people seem to doubt the truth value of AI-generated deepfakes, it still has the intended delegitimizing impact on the attacked politician. This discrepancy between credibility and delegitimization is crucial to consider: Although the deception involving AI-generated deepfakes attacking an opposed political candidate is likely to be detected by the audience, it can still pay off electorally as it may lower the support for the targeted politician, even more so for those who initially were most inclined to support the actor that was attacked in the deepfake. Targeting deepfakes to partisans supporting the delegitimized politician may thus be a successful disinformation technique in electoral periods.

LLMs can't stop making up software dependencies and sabotaging everything
(link)
The problem is, these code suggestions often include hallucinated package names that sound real but don’t exist. I’ve seen this firsthand. You paste it into your terminal and the install fails – or worse, it doesn’t fail, because someone has slop-squatted that exact package name.

A Single Operator, Two AI Platforms, Nine Government Agencies: The Full Technical Report
(link)
AI has collapsed the cost and complexity of reaching those systems. The gap between what attackers can do and what defenders can protect is widening.

AI fails to make inroads with cybercriminals
(link)
Cybercriminals are experimenting with these tools, but as far as we can tell it’s not delivering them real benefits in their own work. Our message to industry is: don't panic yet. The immediate danger comes from companies and members of the public adopting poorly secured AI systems themselves, opening them up to catastrophic new attacks that can be performed by cybercriminals with little effort or skill.

Elon Musk’s xAI powering its facility in Memphis with ‘illegal’ generators
(link), and more here
The 35 generators xAI is using are “illegal” and a “major source of air pollution”, the law center wrote in a letter to the Shelby county health department on Wednesday. It says these high emission rates violate the Clean Air Act, including specified limits on toxic and carcinogenic pollution.

The hidden cost of Google’s AI defaults and the illusion of choice
(link)
What we’re seeing in both free and paid Google accounts is the power of defaults in the AI era. The default is sharing data for AI training. The default is AI summaries in your email. The default is AI-powered document creation. You can change these settings, but Google has to know most people won’t do that, because the options are hard to find and don’t work as they should.

Google Chrome silently installs a 4 GB AI model on your device without consent. At a billion-device scale the climate costs are insane.
(link)
In light of what is increasingly becoming default behaviour, one has to ask a very simple question. When will the Regulators and Public Prosecutors start to enforce the law which has been in place since 2002 - or are global tech corporations exempt from criminal and civil statutes?
About that one, also read: "Google participates in the web standards process the way a bear participates in the “camping” process."

IA : Apple accepte de payer 250 millions de dollars après des plaintes l’accusant d’avoir trompé des millions d’acheteurs d’iPhone sur ses capacités
(link)
Les plaignants reprochaient à l’entreprise californienne d’avoir « promu des capacités d’IA qui n’existaient pas au moment des faits, n’existent pas aujourd’hui, et n’existeront pas avant deux ans ou plus », afin de stimuler les ventes d’iPhone, rappelle le document, consulté par l’Agence France-Presse.

Mark Zuckerberg ‘personally authorized’ Meta’s copyright infringement, publishers allege
(link)
“Defendants reproduced and distributed millions of copyrighted works without permission, without providing any compensation to authors or publishers, and with full knowledge that their conduct violated copyright law,” the complaint reads in part. “Zuckerberg himself personally authorized and actively encouraged the infringement.”

AI chatbots are giving out people’s real phone numbers
(link)
AI researchers and online privacy experts have long warned of the myriad dangers generative AI poses for personal privacy. These cases give us yet another scenario to worry about: generative AI exposing people’s real phone numbers.

Generative AI: What You Need To Know
(link)
"Generative AI: What You Need To Know is a free resource that will help you develop an AI-bullshit detector."

AI companies will fail. We can salvage something from the wreckage
(link)
AI is asbestos in the walls of our tech society, stuffed there by monopolists run amok. A serious fight against it must strike at its roots

AI 2027
(link)
We predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution. We wrote a scenario that represents our best guess about what that might look like. It’s informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting successes.