Suche
Beiträge, die mit LLM getaggt sind
https://venturebeat.com/ai/researchers-warn-of-catastrophic-overtraining-in-large-language-models/
The researchers compared two versions of OLMo-1b: one pre-trained on 2.3 trillion tokens and another on 3 trillion tokens.
#LLM #Press
https://arstechnica.com/security/2025/03/gemini-hackers-can-deliver-more-potent-attacks-with-a-helping-hand-from-gemini/
Hacking LLMs has always been more art than science. A new attack on Gemini could change that.
In the growing canon of AI security, the indirect prompt injection has emerged as the most powerful means for attackers to hack large language models such as OpenAI’s GPT-3 and GPT-4 or Microsoft’s Copilot.
KEWL🌀NEWS
#GEMINI #LLM #HACKS #PRESS
https://gomoot.com/gemini-2-5-pro-il-modello-ia-di-google-che-pensa/
#blog #dataset #gemini #google #grok #llm #lmarena #modelloai #multimodale #news #openai #pensare #picks #tech #tecnologia #token
Gemini 2.5 Pro: il modello IA di Google che “pensa”
Gemini 2.5 Pro, il "modello più intelligente" di Google focalizzato sul ragionamento. Prestazioni top, multimodalità, contesto enorme e accesso via AI Studiocristina (Gomoot : tecnologia e lifestyle)
https://jewishinsider.com/2025/03/leading-ai-tools-demonstrate-concerning-bias-against-israel-and-jews-new-adl-study-finds/
Llama, Meta’s large language model, showed the most ‘pronounced’ bias among GPT, Claude and Gemini
#AI #bias #Israel #Jews #Llama #Meta #ChatGPT #Claude #Gemini #LLM #press
»Open Source Infrastructure has an AI problem«
– from @niccolove / @niccolo_ve
📺 https://tube.kockatoo.org/w/woce36hTzF6JdnBiBEaRZJ
#opensource #peertube #video #ai #aiart #video #foss #floss #llm #LLMs #attack #freedom
Open Source Infrastructure has an AI problem
You can find all of these videos as written articles, plus some extra content, at https://thelibre.news You can the channel grow by donating to the following platforms: Paypal: https://paypal.me/ni...Kockatoo Tube
»Open Source World – FOSS infrastructure is under attack by AI companies:
LLM scrapers are taking down FOSS projects' infrastructure, and it's getting worse.«
😐 https://thelibre.news/foss-infrastructure-is-under-attack-by-ai-companies/
#foss #opensource #llm #attack #floss #kde #gnome #robotstxt #stolendata #LLMs #freedom #infrastructure #greed
FOSS infrastructure is under attack by AI companies
LLM scrapers are taking down FOSS projects' infrastructure, and it's getting worse.Niccolò Venerandi (LibreNews)
"LLM Scraper zerstören die Infrastruktur von FOSS-Projekten, und es wird immer schlimmer."
https://thelibre.news/foss-infrastructure-is-under-attack-by-ai-companies/
#OpenSource #FOSS #LLM #KI #AI
FOSS infrastructure is under attack by AI companies
LLM scrapers are taking down FOSS projects' infrastructure, and it's getting worse.Niccolò Venerandi (LibreNews)
https://gomoot.com/mistral-small-3-1-llm-di-dimensioni-ridotte-ma-con-buone-prestazioni/
#ai #blog #gemma #google #ia #llm #mistral #multimodale #news #picks #small #tech #tecnologia
Mistral Small 3.1, LLM di dimensioni ridotte ma con buone prestazioni
Mistral Small 3.1: il nuovo modello open-source multimodale di piccole dimensioni che supera Gemma 3 e GPT-4o Mini, ottimizzato per velocità e efficienza.Graziano (Gomoot : tecnologia e lifestyle)
"Perplexity provided incorrect information in 37 percent of the queries tested, whereas ChatGPT Search incorrectly identified 67 percent (134 out of 200) of articles queried. Grok 3 demonstrated the highest error rate, at 94 percent."
https://arstechnica.com/ai/2025/03/ai-search-engines-give-incorrect-answers-at-an-alarming-60-rate-study-says/
#search #llm #education #research #Fail #journalism #law
AI search engines cite incorrect sources at an alarming 60% rate, study says
CJR study shows AI search services misinform users and ignore publisher exclusion requests.Benj Edwards (Ars Technica)
https://gomoot.com/google-gemini-deep-research-gratuito-scopri-come-funziona
#ai #blog #deepresearch #gemini #ia #llm #news #openai #picks #tech #tecnologia
Google Gemini Deep Research gratuito: scopri come funziona
Google Gemini Deep Research è ora gratuito per tutti: un assistente AI che rivoluziona le ricerche online, migliorando le analisi e i report.Graziano (Gomoot : tecnologia e lifestyle)
https://gomoot.com/gemma-3-modello-ai-open-source-ottimizzato-per-una-singola-gpu/
#blog #chatbot #deepseek #gemini #gemma3 #ia #iot #llm #local #news #openai #picks #tech #tecnologia
Gemma 3: modello AI open source ottimizzato per una singola GPU
Gemma 3: Il modello AI di Google che combina efficienza e potenza in una soluzione open-source, ideale per sviluppatori e aziende in cerca di soluzioni low costGraziano (Gomoot : tecnologia e lifestyle)
https://www.bloodinthemachine.com/p/so-the-la-times-replaced-me-with
#AI #LLM #journalism #wtf #LATimes
So the LA Times replaced me with an AI that defends the KKK
The LA Times is laying off and buying out staff, while introducing highly dubious AI tools. This is what automation looks like in 2025.Brian Merchant (Blood in the Machine)
Maison Marcel
Discover MARCEL by Marcel Katz, a luxury collectible sunglass brand. Italian craftsmanship and sustainable materials. Milan | Miami collectible sunglasses.Maison Marcel
Maison Marcel
Discover MARCEL by Marcel Katz, a luxury collectible sunglass brand. Italian craftsmanship and sustainable materials. Milan | Miami collectible sunglasses.Maison Marcel
https://gomoot.com/claude-code-lassistente-ai-per-la-programmazione-da-terminale/
#anthropic #bash #blog #claude #coding #ia #linux #llm #news #picks #python #shell #tech #tecnologia #terminale
Claude Code: l’assistente AI per la programmazione da terminale
Claude Code, uno strumento AI che aiuta gli sviluppatori a gestire il codice direttamente dal terminale, riducendo il tempo speso nelle operazioni manualiGraziano (Gomoot : tecnologia e lifestyle)
https://gomoot.com/gemini-code-assist-diventa-gratuito
#assist #blog #code #coding #gemini #gemini2 #github #gratis #ia #llm #news #picks #tech #tecnologia #token
Gemini Code Assist diventa gratuito
Gemini Code Assist diventa gratuito con un limite di 180000 completamenti di codice mensili, una finestra di contesto di 128000 token e l'integrazione in GitHubGraziano (Gomoot : tecnologia e lifestyle)
»#Manipulation – Wie KI die Absichten von Nutzern steuern kann:
Bislang haben #KI-Anbieter wie OpenAI, #Google oder #Meta versucht, ihre Sprachmodelle auf größtmögliche Ausgewogenheit zu trimmen, bevor sie veröffentlicht werden. Nun wird gewarnt, dass sich das ändern könnte. Was sind die Motive für eine Abkehr?«
🔊 https://www.deutschlandfunk.de/manipulator-wie-ki-die-absichten-von-nutzerinnen-und-nutzern-steuern-kann-100.html
#LLMs #neutral #openai #tiktok #tracking #llm
Wie KI die Absichten von Nutzerinnen und Nutzern steuern kann
Bislang haben KI-Anbieter wie OpenAI versucht, ihre Sprachmodelle auf Ausgewogenheit zu trimmen, bevor sie veröffentlicht werden. Das könnte sich ändern.Deutschlandfunk
Meet the journalists training #AI models for #Meta and #OpenAI
The gig work platform #Outlier is one of several companies courting journalists to train large language models (#LLMs).
https://www.niemanlab.org/2025/02/meet-the-journalists-training-ai-models-for-meta-and-openai/
Scary, if your income depends on training the overlords that will replace you
Meet the journalists training AI models for Meta and OpenAI
The gig work platform Outlier is one of several companies courting journalists to train large language models (LLMs).Nieman Lab
Pretty much everyone from a #STEM background - myself definitely included! - owes the #Humanities a huge apology.
I mean, I get it. When I was a young student of physics, it was easy for me to sneer at philosophy students and whatnot. After all, _we_ dealt with hard, measurable facts, while _those_ people dealt with some weird thought constructs that had no relevancy to the real world - right?
But this is the end result - #TechBro culture and a vast portion of our entire economy using digital bullshit generators instead of critical thinking, and using this to lead us into a fascist future where either Truth or Facts have become meaningless.
Mea culpa.
A new audit of DeepSeek's iOS app has found glaring security issues, the foremost being that it sends sensitive data over the internet without any encryption.
[⚠️This exposes your data to interception and manipulation attacks.⚠️]
https://www.nowsecure.com/blog/2025/02/06/nowsecure-uncovers-multiple-security-and-privacy-flaws-in-deepseek-ios-mobile-app/
#deepseek #ai #llm #ios #app #it #security #privacy #engineer #media #tech #news
![[ImageSource: NowSecure]
"The DeepSeek iOS app globally disables App Transport Security (ATS) which is an iOS platform level protection that prevents sensitive data from being sent over unencrypted channels," NowSecure said. "Since this protection is disabled, the app can (and does) send unencrypted data over the internet."
The findings add to a growing list of concerns that have been raised around the artificial intelligence (AI) chatbot service, even as it skyrocketed to the top of the app store charts on both Android and iOS in several markets across the world.
The app's Chinese links, much like TikTok, have prompted U.S. lawmakers to push for a nation-wide ban on DeepSeek from government devices over risks that it could provide user information to Beijing.
<https://gottheimer.house.gov/posts/release-gottheimer-lahood-introduce-new-bipartisan-legislation-to-protect-americans-from-deepseek>
It's worth noting that several countries, including Australia, Italy, Netherlands, Taiwan, South Korea, government agencies in India and the United States (such as the Congress, NASA, Navy and Pentagon), have instituted bans on DeepSeek from government devices.](https://nerdculture.de/system/media_attachments/files/114/018/996/910/493/093/original/7e00d291d5e366a2.png "[ImageSource: NowSecure]
\"The DeepSeek iOS app globally disables App Transport Security (ATS) which is an iOS platform level protection that prevents sensitive data from being sent over unencrypted channels,\" NowSecure said. \"Since this protection is disabled, the app can (and does) send unencrypted data over the internet.\"
The findings add to a growing list of concerns that have been raised around the artificial intelligence (AI) chatbot service, even as it skyrocketed to the top of the app store charts on both Android and iOS in several markets across the world.
The app's Chinese links, much like TikTok, have prompted U.S. lawmakers to push for a nation-wide ban on DeepSeek from government devices over risks that it could provide user information to Beijing.
<https://gottheimer.house.gov/posts/release-gottheimer-lahood-introduce-new-bipartisan-legislation-to-protect-americans-from-deepseek>
It's worth noting that several countries, including Australia, Italy, Netherlands, Taiwan, South Korea, government agencies in India and the United States (such as the Congress, NASA, Navy and Pentagon), have instituted bans on DeepSeek from government devices.")
NowSecure Uncovers Multiple Security and Privacy Flaws in DeepSeek iOS Mobile App - NowSecure
A NowSecure mobile application security and privacy assessment has uncovered multiple security and privacy issues in the DeepSeek iOS mobile app that lead As the top iOS app since Jan 25, 2025, the DeepSeek iOS app has already been downloaded and use…Andrew Hoog (NowSecure, Inc.)
https://gomoot.com/lia-gemini-da-oggi-puo-ricordare-le-conversazioni/
#advanced #ai #blog #gemini #ia #llm #memoria #memory #news #picks #tech #tecnologia
L’IA Gemini da oggi può ricordare le conversazioni
Da oggi Google Gemini cambia l'esperienza AI grazie alla memoria conversazionale con la quale può ricordare ogni dettaglio delle tue conversazioni passate.Graziano (Gomoot : tecnologia e lifestyle)
Threat actors are taking advantage of the rise in popularity of DeepSeek to promote two malicious infostealer packages on the Python Package Index, where they impersonated developer tools for the AI platform.
https://global.ptsecurity.com/analytics/pt-esc-threat-intelligence/malicious-packages-deepseeek-and-deepseekai-published-in-python-package-index
#deepseek #llm #it #security #privacy #engineer #media #developer #tech #news
![[ImageSource: Positive Technologies]
The malicious packages, deepseeek 0.0.8 and deepseekai 0.0.8 were uploaded to PyPI on January 29, 2025, with only twenty minutes between them.
Positive Technologies quickly discovered and reported them to PyPI, which quarantined and blocked downloads of the packages, followed by their complete deletion from the platform.
Despite the quick detection and response, 222 developers downloaded the two packages, most from the United States (117), followed by China (36), Russia, Germany, Hong Kong and Canada. Those developers who utilized these packages should immediately rotate their API keys, authentication tokens and passwords, as they may now be compromised.](https://nerdculture.de/system/media_attachments/files/113/963/687/409/845/888/original/26ce13670f89cd53.png "[ImageSource: Positive Technologies]
The malicious packages, deepseeek 0.0.8 and deepseekai 0.0.8 were uploaded to PyPI on January 29, 2025, with only twenty minutes between them.
Positive Technologies quickly discovered and reported them to PyPI, which quarantined and blocked downloads of the packages, followed by their complete deletion from the platform.
Despite the quick detection and response, 222 developers downloaded the two packages, most from the United States (117), followed by China (36), Russia, Germany, Hong Kong and Canada. Those developers who utilized these packages should immediately rotate their API keys, authentication tokens and passwords, as they may now be compromised.")
![[ImageSource: Positive Technologies]
Malicious payload contained in both packages.
The stolen information was exfiltrated to a command and control (C2) server at eoyyiyqubj7mquj.m.pipedream[.]net using Pipedream, a legitimate automation platform. Threat actors could use this stolen information to access cloud services, databases and other protected resources utilized by the developer.
"Functions used in these packages are designed to collect user and computer data and steal environment variables", reads the Positive Technologies report. "The payload is executed when the user runs the commands deepseeek or deepseekai (depending on the package) in the command-line interface."
"Environment variables often contain sensitive data required for applications to run, for example, API keys for the S3 storage service, database credentials, and permissions to access other infrastructure resources."](https://nerdculture.de/system/media_attachments/files/113/963/688/343/436/642/original/9a479c6c50735793.png "[ImageSource: Positive Technologies]
Malicious payload contained in both packages.
The stolen information was exfiltrated to a command and control (C2) server at eoyyiyqubj7mquj.m.pipedream[.]net using Pipedream, a legitimate automation platform. Threat actors could use this stolen information to access cloud services, databases and other protected resources utilized by the developer.
\"Functions used in these packages are designed to collect user and computer data and steal environment variables\", reads the Positive Technologies report. \"The payload is executed when the user runs the commands deepseeek or deepseekai (depending on the package) in the command-line interface.\"
\"Environment variables often contain sensitive data required for applications to run, for example, API keys for the S3 storage service, database credentials, and permissions to access other infrastructure resources.\"")
Malicious packages deepseeek and deepseekai published in Python Package Index
As part of our threat research and monitoring efforts, the Supply Chain Security team of the Threat Intelligence department of the Positive Technologies Expert Security Center (PT ESC) detected and prevented a malicious campaign in the Python Package…global.ptsecurity.com
https://gomoot.com/openeurollm-nasce-un-progetto-di-sviluppo-ai-europeo/
#aiact #amd #blog #europa #ia #lighton #llm #news #openeurollm #picks #stargate
OpenEuroLLM: nasce un progetto di sviluppo AI europeo
L'Europa si unisce per creare la piattaforma AI OpenEuroLLM; innovativa e conforme agli standard di trasparenza e sicurezza richiesti dalla Commissione EuropeaGraziano (Gomoot : tecnologia e lifestyle)
#AI #LLM #Training #tech #Blogs
https://thundroid.co/the-power-of-large-language-models-how-theyre-transforming-ai-training/
The Power of Large Language Models: How They’re Transforming AI Training - ThunDroid
Artificial intelligence is no longer a futuristic dream—it’s here, reshaping how we interact with technology, businesses, and even each other. At the heart of this AI revolution lies something truly remarkable: Large Language Models (LLMs).ujjwalthakur (ThunDroid)
https://gomoot.com/microsoft-decide-di-integrare-il-modello-deepseek-r1/
#ai #amd #azure #blog #cloud #copilot #deepseek #github #ia #llm #microsoft #news #openai #picks #snapdragon #tech #tecnologia
Microsoft decide di integrare il modello DeepSeek R1
Microsoft decide di integrare il modello DeepSeek R1 in Azure AI Foundry, GitHub e in locale sui computer Copilot+ grazie a costi ridotti ed efficienza.Graziano (Gomoot : tecnologia e lifestyle)
#AI #ArtificialIntelligence #LLM
Sabot in the Age of AI
A captured screenshot showcasing the iocaine demonstration site in operation. The primary objective of iocaine is to generate a stable, infinite maze of randomized garbage.ASRG
https://gomoot.com/janus-pro-deepseek-sfida-ancora-openai-con-un-nuovo-modello-di-generazione-immagini
#ai #blog #cina #copilot #deeepseek #gemini #ia #janus #januspro #llm #meta #multimodale #news #openai #picks #tech #tecnologia @deepseek_ai
Janus Pro: DeepSeek sfida ancora OpenAI e lancia un nuovo modello di generazione immagini
Il nuovo modello multimodale Janus Pro di DeepSeek supera DALL-E 3 in efficienza e prestazioni, un'alternativa open source tra i generatori di immagini AI.Graziano (Gomoot : tecnologia e lifestyle)
https://git.madhouse-project.org/algernon/iocaine
#LLM #ArtificialIntelligence #KünstlicheIntelligenz #hosting
Here are my reasons as to why this kind of #LLM-usage is bad, wrong and needs to be stopped:
1) It's a kind of grave-robbbing and incredibly disrespectful to the real Anne Frank and her family. She, her memory and the things she wrote get abused for our enjoyment, with no regard or care for the real person. How anyone thinks this is even remotely appropriate is beyond me.
https://fedihum.org/@lavaeolus/113842459724961937
#AI #LudditeAI
🧵1/3
Henrik Schönemann (@lavaeolus@fedihum.org)
Attached: 4 images Update2: https://fedihum.org/@lavaeolus/113860035702353891 Update: https://fedihum.org/@lavaeolus/113856096099247855 --- An '#AI-emulation' of Anne Frank made for use in schools.FeDiHum
🥥 It's being widely reported on the internet today that Meta CEO Mark Zuckerberg is the offspring of a goat that was raped by Sen.Ted Cruz (R-TX) during a 3-way with the Zodiac Killer.
This twt has not been generated by AI, yet. 🥥
#ContentModeration #Meta #Zuckerberg #TedCruz #ZodiacKiller #Goat #ThreeWay #SocialMedia #AI #LLM ##TuckersBalls
Künstliche proprietäre Intelligenz hat geschrieben:
"Dies koennte zu einer kritischen Haltung gegenueber propietaeren Systemen fuehren."
Das Fediverse ist nicht ganz dicht
KI-Crawler durchstreifen das Fediverse und versuchen, so viele Informationen wie möglich über uns zu sammeln, um sie dann in ihren LLMs zu verarbeiten. Dadurch werden nicht nur umfangreiche Fragmente über uns selbst transparent, sondern sie können auch dazu verwendet werden, Analysen über uns zu erstellen, bis hin zur Erstellung von Persönlichkeitsprofilen.
Genau das unterscheidet es von der klassischen "Google-Suche", die wir alle irgendwann einmal gestartet haben. Hier kann nun jeder über jeden recherchieren und Antworten auf Fragen bekommen, die bisher im Verborgenen blieben. Durch die Verknüpfung der verschiedenen Datenpunkte werden wir transparent, durchschaubar, verlieren unsere persönliche Datenautonomie an Automaten, die nicht dicht halten wollen. Da hilft es auch nicht, wenn man nach 14 Tagen alle seine Posts löscht. Da sind die Roboter sicher schneller.
Natürlich habe ich den Selbstversuch gestartet. Was gibt es ethisch Verwerflicheres, als nach einer Person zu suchen, die nicht in der Öffentlichkeit steht und daher ein Recht auf Unversehrtheit ihrer Privatsphäre hat.
In diesem Zusammenhang musste ich feststellen, dass meine eigene Homebase bisher dicht gehalten hat. Keine Daten von mir tauchen in dieser Quelle auf. Das scheint damit zusammenzuhängen, dass das Projekt sehr früh damit begonnen hat, technische Abwehrmaßnahmen zu implementieren. Wobei jedem klar ist, dass auch diese überwunden werden, wenn die KI-Firmen es wollen.
Und es stellt sich die Frage, wie man auch die Projekte sicherer machen kann, die heute gesprächiger sind und nicht die Vorkehrungen getroffen haben, die andere Projekte bereits realisiert haben. Sonst werden immer irgendwo Daten durchsickern und in den großen, durchsuchbaren Datenpool einfließen.
Read the full guide at https://www.spsanderson.com/steveondata/posts/2024-12-24/ and share your thoughts! 💼📈
#LLM #Programming #AI #Finance #Blog
Comprehensive Guide: Using DoTadda Knowledge for Earnings Call Analysis – Steve’s Data Tips and Tricks
Leverage DoTadda Knowledge to streamline earnings call analysis. Gain insights on sentiment, trends, risks, and market reactions. Boost investment decision-making with AI-powered financial analyticSteve's Data Tips and Tricks
Read the full guide at https://www.spsanderson.com/steveondata/posts/2024-12-24/ and share your thoughts! 💼📈
#LLM #Programming #AI #Finance #Blog
Comprehensive Guide: Using DoTadda Knowledge for Earnings Call Analysis – Steve’s Data Tips and Tricks
Leverage DoTadda Knowledge to streamline earnings call analysis. Gain insights on sentiment, trends, risks, and market reactions. Boost investment decision-making with AI-powered financial analyticSteve's Data Tips and Tricks