Lmst

Here is my morning trying to convince Gemini that, no, the --discard option in Debian's losetup is pure hallucination.

#ai #ml #hallucination #linux #debian #opensource #google #gemini

Two command line screenshot stacked above each other. The first one is from Gemini 2.5 saying: "Given your 6.8 kernel, your losetup version (from the util-linux package) almost certainly supports the --discard flag.". The second one is from a bash shell with the commands:
"sudo losetup --discard -f --show file.img
losetup: unrecognized option '--discard'
Try 'losetup --help' for more information.
"

🎨 Tôi đã tạo app AI "Lumo" biến ảnh dễ الدماغ thành trang màu cho con Cara & Celine! ⚡️ Technologie: React Native, Firebase, ML. Features: chuyển ảnh > đường vẽ, tùy chỉnh độ chi tiết, xuất chất lượng.رزerved-child-safe, không eût. Mình muốn giáo dục và sức vui của trẻ! 👩💻👧👦 #AIVănBập #TrangMàu #ReactNative #Firebase #ML #BắtĐầuCông #BãoM Auseinanders

(499 ký tự)

https://www.reddit.com/r/SideProject/comments/1ob40ml/i_built_an_ai_app_that_turns_photos_into_coloring/

»OpenAI’s ‘embarrassing’ math« https://techcrunch.com/2025/10/19/openais-embarrassing-math/?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

»Opera’s Neon shows just how confusing AI browsers still are« https://www.theverge.com/tech/801899/opera-neon-ai-browser-trial-run?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

#ByteDance is promoting its #AIchatbot app #Cici in the #UK, #Mexico, and #Indonesia through ads and influencer partnerships. https://www.wired.com/story/bytedances-ai-chatbot-is-quietly-gaining-traction-around-the-world/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI

»The AI sexting era has arrived« https://www.theverge.com/column/802022/ai-sexting-openai-chatbot-chatgpt?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

[Перевод] Как шаблоны рассуждения учат ИИ думать: новая эпоха Pattern-Aware Learning (PARO)

TL;DR: В работе «Reasoning Pattern Matters: Learning to Reason without Human Rationales» авторы показывают, что для шаблонных задач рассуждения (patterned reasoning) ключ к качеству — не объём и не безупречность «рационалей», а явное усвоение шаблона рассуждения . Они демонстрируют это на двух финансовых задачах (NSM и TPC), вводят метрику/анализ «развилочных токенов» и предлагают PARO — способ автоматически генерировать рационали, следуя заранее заданному шаблону. Итог: SFT+RLVR, обученное на 10× меньшем количестве человеческих рационалей или частично «испорченных» рационалях, почти не теряет в качестве; а рационали, сгенерированные PARO, сопоставимы с крупными человеческими наборами.

https://habr.com/ru/articles/957992/

#ai #ml

Sử thích học tập tương tác thay place cho video! AP Lab Academy là nền tảng học thị myصوص à học nhỏ các khái niệm kỹ thuật qua tay Livre-topic (mật, hiển thị, laboratory adjustable). AI hỗ trợ phát triển công cụ. Có Handbook trên ML, agent, toán ฐ+N Марк. Vui lòng phản hồi if đã thử!

#APLabAcademy #InteractiveLearning #ML #AI #Education #TửViHọcTập #HọcMáyThuyết #ChiêuHọc #TưTrungGiáo

https://www.reddit.com/r/SideProject/comments/1oaj9it/interactive_learning_without_videos_my_experiment/

War ja gestern wieder bei "den Kommis" (schon paar mal von geschrieben. Bieten flinta boxen an, essen , mietrechtsberatung, helfen Leuten, sich zu organisieren , sprachencafe, etc)

Und alle hochmotiviert, finden ihre Arbeit toll, die sie teils seit 2 Jahren machen. (Womit sie recht haben, die vordergründige Stadtteilarbeit ist echt toll)

Aber. Ich habe gestern wieder mehrere gefragt, was sie von der Hinterorga wissen, wie ihre Meinung dazu ist. MLer.

Wie oft ich dann verunsicherte antworten bekam. "Nicht mit befasst". " Marx, Lenin....? Ääähh..."

Gestern das erste mal nach nem Jahr rumfragen(!!!!) wen getroffen, die Person wusste Bescheid.

Ich find das einfach krass.

(Gespräch war interesssant, hat viel nachgedacht die Person. Wurde aber immer unterbrochen von dem Ende der Pause 😂) kamen zufällig ins Gespräch und als das Diskussionslevel plötzlich stieg über menschliches Sein dachte ich, ich Frage mal nach machtvorstellungen die "gebraucht" werden der Person nach. Solche Unterhaltungen lieb ich ja, wenn es nicht nur Geschwafel ist.:gribz_3:

Naja ach .einfach Lust zu teilen.

Wer geht mit mir zu denen und klärt mal die ganzen antiautoritären auf, wem sie da zu arbeiten? 😂

#ml #kommis #anarchie

»Wikipedia says traffic is falling due to AI search summaries and social video« https://techcrunch.com/2025/10/18/wikipedia-says-traffic-is-falling-due-to-ai-search-summaries-and-social-video/?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

✨ Impulsione a Economia com Modelos de Machine Learning!
📝 Descubra como modelos de machine learning estão se tornando essenciais para revitalizar a economia em tempos desafiadores. Neste artigo do INK|DESIGN NEWS, exploramos as inovações em IA que prometem transformar setores inteiros e otimizar processos. Não fique de fora dessa revolução digital! Clique e saiba mais!
.
.
.#AI #ML #Dee...
https://inkdesign.com.br/modelos-de-machine-learning-podem-salvar-a-economia-de-ai/?fsp_sid=220665

»Too burned out to travel? This new app fakes your summer vacation photos for you« https://techcrunch.com/2025/10/18/too-burned-out-to-travel-this-new-app-fakes-your-summer-vacation-photos-for-you/?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

Heute n Workshop besucht, ging um Kritik und Selbstkritik .

Wurde anfangs beschrieben, wie das Patriarchat Nutzen zieht aus schlechter Kritik und Unfähigkeit , Kritik anzunehmen, etc.

Vereinzelung.

Der war voll nice.

Und von den MLern.

........

Warum kommt so was nicht von uns 😭

#anarchie #ML

NEW BIML Bibliography entry

https://www.nature.com/articles/s41586-025-08661-4

Mert Yuksekgonul et al

New computational paradigm of semantics of language. Auomatic differentiation by analogy. A very naive approach to feedback loops ignoring all kinds of known problems including: looping, non-convergence, and local minima.

#MLsec #ML

https://berryvilleiml.com/references/

NEW BIML Bibliography entry

https://www.sciencedirect.com/science/article/pii/S0167404824002931
Adversarial Machine Learning in Industry: A Systematic Literature Review

Felix Viktor Jedrzejewski et al

Wonder why science journals are useless? This paper from 2024 is already seriously out of date. Results are circa 2020 (a year after BIML was formed). Pre-LLM adversarial examples are the main subject. Building Security In is barely mentioned.

DON'T BOTHER

#MLsec #ML

https://berryvilleiml.com/references/

»WhatssApp changes its terms to bar general purpose chatbots from its platform« https://techcrunch.com/2025/10/18/whatssapp-changes-its-terms-to-bar-general-purpose-chatbots-from-its-platform/?AIagent.at #AIagent #AI #ML #NLP #LLM #GenAI

NEW BIML Bibliography entry

https://arxiv.org/pdf/2501.09798v1

Computing Optimization-Based Prompt Injections Against Closed-Weights Models
By Misusing a Fine-Tuning API

Andrey Labunets, et al

Lack of coordination betweem designers of ML systems leads to attacks like this. Lots of leaky interfaces and poor security design. Malicious input is still a major problem. This one is fun because it uses the fine-tuning interface.

#MLsec #ML

https://berryvilleiml.com/references/

NEW BIML Bibliography entry

https://arxiv.org/pdf/2503.18813

Defeating Prompt Injections by Design

DeepMind (Shumailov, Carlini, Kerns, Tramer)

Authored by some of our favorite people in MLsec, this paper is a beacon of hope for security engineering. Wrapper with proof. Securing natural language interfaces.

#ML

Client Info