Giacomo Tesio

@tiotasram@kolektiva.social Yet we can build a mental model of #AI: it's simply a corrupted decompression of a lossy compressed archive of our competitors work. Those who builds the archive gives to our competitor access to our work in exchange of giving us access to their. All for the largest fee they can get at any given time. @jneen@unstable.systems

@argv_minus_one@mastodon.sdf.org Programmers change how the world behave through arcane words. This is literally why I learned Pascal when I was 13: to become a powerful wizard. Couldn't figure the horrible molochs I was going to face, trying to protect my family and friends not from evil wizards like me, but from evil almighty guilds like #BigTech that most people trust! @rygorous@mastodon.gamedev.place

The documents provide smoking-gun evidence that Meta, Google, Snap, and TikTok all purposefully designed their social media products to addict children and teens with no regard for known harms to their wellbeing, and how that mass youth addiction was core to the companies’ business models. The documents contain internal discussions among company employees, presentations from internal meetings, expert testimony, and evidence of Big Tech coordination with tech-funded groups, including the National Parent Teachers Association (PTA) and Family Online Safety Institute (FOSI), in attempts to control the narrative in response to concerned parents. “These unsealed documents prove Big Tech has been gaslighting and lying to the public for years https://techoversight.org/2026/01/25/top-report-mdl-jan-25/ #BigTech #GAFAM #Google #Meta #TikTok #Snap #teens #mentalhealth #privacy #surveillance #addiction #drugs

Hi @hisham_hm@mastodon.social just learned about #dit #editor and I found it interesting enough to find and follow you... just before learning you created #GoboLinux!

Parabéns!

As a #hacker you really expanded the Overton window! A truly great achievement.

Prima commit in #JehanneOS dopo 5 anni, per permetterle di compilare correttamente con #Debian Trixie.

brutto segno.

Sviluppare sistemi operativi è un hobby pericoloso.

@ahltorp@mastodon.nu Boilerplate is a side effect of excessive abstraction. Now think about it for a second. 😉 (btw, did you consider an April fool for Anthropic's leak? It would be great PR.. after.) @bodil@treehouse.systems @tiotasram@kolektiva.social

@pluralistic@mamot.fr Well, we are not only influenced by our legacy: however strong we are, we can't avoid some fundamental influence from the hegemonic culture we live in. Yet I see how the ethical misalignment here may not be about libertarian values but about utilitarian ones. Even more subtly, it might be a misalignment about respective utility functions, while both #pluralistic and @tante@tldr.nettime.org adopt an utilitarian framework instead of a normative one. For example, the Pluralistic use of a local LLM might be explained with a slightly higher evaluation of the benefits that his own writings brings to society and thus (indirectly) the value the LLM brings, despite its issues. Otoh, Tante might value a lot more the political harm that Cory's words did by blaming a political choice as irrational while it's totally rationale: in a way, by justifying the use of a #LLM, #Doctorow justified (even just a little bit) the industry that built it. And since Pluralistic's strawman is centered around a normative "purity culture" blamed as irrational, Tante framed his response over rationality. What if a normative behaviour was in fact totally rational in presence of unreducible complexity and informational asymmetry? I don't use LLM for so many technical and political reasons that would take hours to list. And you both would almost certainly nod to most of them as a strictly rational arguments. Yet the choice itself, bound to the society I want to build for my daughters and children, is normative: based on the values of truth, freedom and communion. None of these could ever come from the LLM we are talking about: they are weapons designed to fool people (Turing test included!), so there's no way to wield them to benefit people. As for "purity culture", I'm a catholic #christian, not a puritan: we brag about the #Church being a casta meretrix (Latin for something like "a pure bitch" 🤣), and we preach a man who hanged with the worst sinners and sometimes even hacking the law to save their lifes, so... 🤷‍♂️

I really can't grasp how I should use EXTERNAL_TOOLCHAIN variable to build #NetBSD with pcc (portable compiler).

Any NetBSD expert out there?

I wonder if @stefano@mastodon.bsd.cafe can help me answer a #NetBSD question.

Did anybody ever tried to build a recent NetBSD with #TCC (aka #TinyCC)?

I searched the web and a few NetBSD mailing list archives, but with no luck.

@cdarwin@c.im Since #Taxes don't work, I wonder how much time it will take for those 4 billion of people to realize that there is a simple, fast and effective alternative way to get rid of those parasites. @JRepin@mstdn.io

@timbray@cosocial.ca It's so funny, if you think about it. After all, #Dijkstra wrote The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence. So it's basically a race: COBOL to cripple the LLMsLLMs to cripple the global finance🤣 #LLM #vibecoding #COBOL #AIbubble

@riverpunk@defcon.social I'm afraid it's a bit naive, as a description: replace the "Social Web" with the "Market" and you get something Adam Smith might have wrote to describe classical model of market economy: The Market was designed like this because it prevents degradation. How? Because it has escape bridges built in. If one of the seller "goes bad", you can jump ship without loosing access to your favorite product. When the last monopolist went bad, there were no bridges, so we're all still stuck there. The Market can never go bad, because there will always be a new seller to escape to; a new better option And to be fair, within the small markets defined by the limits imposed by the slow production and products distribution of his age, the Smith's model provided a pretty good framework to describe "the wealth of nations". However the steam engine and the industrial revolution changed the dynamics, turning its simple model into a propaganda tool, instead of a valid economical one. In the same way, your description might be useful as a marketing/propaganda tool (and indeed it's actively used by #BlueSky to pretend an decentralization it doesn't really provide), but not to describe or understand truly distributed social systems, such as #XMPP, the #Fediverse and (maybe) #Matrix. @akareilly@hachyderm.io @david_chisnall@infosec.exchange

@hko@floss.social Wait, "proprietary format"? #GnuPG is #FreeSoftware, so even if it use an original format, how can it qualified as "proprietary"? Any patent? Some dynamic linking trick? Also, did you asked the authors about their choices? In the past, I've seen lot of politics around (against) gnupg, and whenever I got deeper, I've found their reason rock-solid from a technical/security perspective (even for retrocompatibility support).

@david_chisnall@infosec.exchange I find much easier to explain this technology in terms of virtualized hardware and software executed on it. The "model" is just a custom special-purpose machine architecture executing the "weights", that are the encoding of the software such machines can run. The match between the topology of the model and the cardinality of the weigths' matrices is a hint about this simple relationship. I call the "model" as "vector mapping (virtual) machine" or VMM. Just like a CPU is built by composing transistors, the fundamental component of a VMM is a "vector reducer", that is a little function from a vector (an array of floats) to a scalar (a float, usually in a carefully crafted range). Vector reducers have a few nice properties: they compose well, so you can compute M of them in parallel over the same array of N floats, and you get a map from a N-dimensional vector space to a M-dimensional vector space (a layer, in ANN jergon)they stack well, so you can take the M-dimensional output vector of a layer and feed it to another fleet of O vectors reducers to obtain another O-dimensional vector. Such pipeline effectively map the original M-dimensional vector to a O-dimensional one, appyling a complex non-linear trasform to it.crucially, by carefully picking each layer's parametric function and recording each reducer parameters (usually, polinomial coefficients, aka "weights", bias and threasolds), you can back-propagate statistical errors against your intended output for any given input, iteratively programming the machine.Once you realize we are using the data to iteratively compute a software designed to be executed by a very specific, special purpose machines, everything becomes clear about this technology. You are not "training" an "artificial intelligence" with IP-infridging datasets. No machine is "learning". You are just compiling such datasets into an executable, just like you would do with a compiler on C sources. The datasets are the source code for such compilation process. And if an x86 binary executed by a CPU retain the source's authors' copyright, the same applies to float executed by a GPU. Also you stop thinking about such software as a subject: it's never #ChatGPT who is harming or defaming a person, but #OpenAI through ChatGPT. Removing the antropomorphization restores the full chain of accountability. We should all refuse to talk about "artificial intelligences" that "hallucinate" and speak in terms of defective statistically programmed software that was compiled from illegal and unknown sources and that was likely injected with undetectable backdoors. We should refuse to call #MachineLearning what is just a compilation process and to use evocative terms like "latent space" what is just a pattern-preserving statistical heavy-loss compression. Unfortunately we need to wait for the #AIbubble to blast, because even #academia is too dependent on #BigTech money and subsides to... think clearly about this stuff. Let's just hope it busts before fascists find a way to leverage either the crash or the tech. #AI #LLM #security @Da_Gut@dice.camp @zzt@mas.to @pluralistic@mamot.fr

@tiotasram@kolektiva.social Of course, such mental model won't let you predict a #CodingAgent's output. Even just because you know a certain amount of random input is included in its computation to give an illusion of creativity and counter some plagiarism accusation. That's why I included the "corrupted" term. On the other hand, it gives you a useful insight of what you might exfiltrate from a competitor work and how to setup the context to get specific parts of its proprietary software, such as secret algorithms, test suites, or even secret keys (in particular short ones like ed25519) that the agent got access to. @jneen@unstable.systems

@gabrielesvelto@mas.to Even talking about "text", in the context of #LLM, is a subtle anthropomorphization. Text is a sequence of symbols used by human minds to express information that they want to syncronize a little with other human minds (aka communicate). Such syncronization is always partial and imperfect, since each mind has different experiences and informations that will integrate the new message, but it's good enough to allow humanity to collaborate and to build culture and science. A statistically programmed software has no mind, so even when it's optimized to produce output that can fool a human and pass the #Turing test, such output hold no meaning, since no human experience or thought is expressed there. It's just the partial decompression of a lossy compression of a huge amount of text. And if it wasn't enough to show the lack of any meaning, the decompression process includes random input that is there to provide the illusion of autonomy. So instead of "the AI replied" I'd suggest "the bot computed this output" and instead of "this work is AI-assisted" I'd suggest "this is statistically computed output".

@gimulnautti@mastodon.green I'd focus to support small and efficient alternative servers like #snac2 that would really benefit the #fediverse diversity and distribution @EUCommission@ec.social-network.europa.eu @Framasoft@framapiaf.org @Mastodon@mastodon.social @IceCubesApp@mastodon.online @ivory@tapbots.social @Tusky@mastodon.social @DataKnightmare@mastodon.xyz

@Edelruth@mastodon.online #LLM output lacks intention, awareness or meaning. It's designed to fool human mind by exploiting the statistical patterns that humans use to syncronize (aka communicate) information they hold in their minds, but there's no mind there. No intelligence, just malicious use of statistics. @gabrielesvelto@mas.to

@gnemmi@mastodon.sdf.org Also... #Signal? A centralized half-open system from #USA is the opposite of "digital independence". @collectifission@greennuclear.online

@teixi@mastodon.social #EthicalAI is a smokescreen. You shouldn't require #AI to be ethical but corporations to be fully accountable for each of its outputs and to pay each of the authors and each of the data subjects whose work or personal data is compiled and compressed in their "models". The use of personal data without explicit and removable consent shouldn't be "unethical", but a penal crime leading people in jail for several years, to pay exponential demages and to the distruction of the model created. With such a legal regime, the interests AI companies will finally align with those of the people. Bonus point: several bilionaries will go to jail for a long long time. @drcaberry@blacktwitter.io @ruha9@mastodon.social @timnitGebru@dair-community.social @cfiesler@hci.social @ruchowdh@mastodon.social @safiyanoble@dair-community.social

@zoy@merveilles.town You'd be surprised to note how many typo I intentionally (and unintentionally 😅) include in my English prose. And I'm literally designing a programming language to be funny for programmers and toxic to #LLM. @tiotasram@kolektiva.social

@FediThing@social.chinwag.org mid way, and I love it so much I can't wait but thank you for sharing. @zoy@merveilles.town @WeirdWriter@caneandable.social @tiotasram@kolektiva.social

Well, for sure one of us lack knowledge on the topic, while the other knows since years that Language Modelling Is Compression. But to claim I'm "factually wrong", you should be able to prove easily with facts that there was no sample in the "training" data of #Claude associating a similar post-it near a doorbell to a similar description. Can you? If you can not, consider you might lack understanding of the very language you use, not just language models. In fact, a recent study on a large population of kids shows a serious degradation in cognitive ability of those using "Generative AI". Adults work the same. #LLM #AI #GenAI

You are either misunderstanding the #Shannon theorem, #LLM working or both. In fact, LLM's weights are lossy compressions of the source data ("training data", in #AI parlance). They mimic intelligence by chaining statistically related fragments of such source data that the users have a negligible probability to recognize as quotations (basically because they haven't read the original sources). The low probability lets AI companies (and their addicted users who "want to believe") to dismiss and/or gaslight evidences as anecdotes, but in low frequency areas of the source data, where decompression artifacts are rare enough, they become quite evident (see for example the evidences produced in NYT vs OpenAI). Or, much more simply, see #GitHub #Copilot distributing #GPLv3 code with wrong attribution and under a wrong license.

😀 I know very well (and faced several times) curve overfitting: it's a term related to models working very well on the "training data", but badly on new samples outside that set. Using the term for LLM output reproducing verbatim fragments of the training data is just mischaracterizing the term to fool those who do not understand statistical algorithms (and neural networks in particular). As for Shannon, you seem to ignore what lossy compression is: https://en.wikipedia.org/wiki/Lossy_compression Having said that, I'm afraid that no amount of computer science or examples can change the mind of an #AI believer/addict.

@troed@swecyb.com Lossy compression is called "lossy" for a reason. As for overfitting, these are literally the first words from your link: Overfitting occurs when a model or algorithm memorizes and fits too closely or exactly to its training data. Overfitting results in a model that might not be able to make accurate predictions or conclusions from any data other than the training data and potentially fails in unexpected scenarios. Compare that with my post to see which of us knows what we are talking about and who don't. Let me guess - in your worldview anyone who does not agree with your definitions is wrong? First, not "my definition" but the classic definitions that even your own link from IBM adopts. Second it's you that 2 days ago replied to me Your claim is factually wrong and shows that you lack knowledge on the topic you're posting about. Look at the mirror, but do not despair: ignorance is a natural condition of human beings, a step in a journey, some times starting point, some times destination. The problem is never ignorance but wrong unchallenged beliefs.

Dude, I think there's nothing I can add. Good luck with your new #AI-proxied overlords! 😉