#Apple #MacStudio #M3Ultra Runs #DeepSeekR1 With 671 Billion Parameters Using 448GB Of Unified Memory, Delivering High Bandwidth Performance At Under 200W Power Consumption, With No Need For A Multi-GPU Setup
https://wccftech.com/m3-ultra-chip-handles-deepseek-r1-model-with-671-billion-parameters/

Developers Wanted: OpenAI Seeks Feedback About Open Model That Will Be Revealed ‘In the Coming Months’ – Source: www.techrepublic.com https://ciso2ciso.com/developers-wanted-openai-seeks-feedback-about-open-model-that-will-be-revealed-in-the-coming-months-source-www-techrepublic-com/ #rssfeedpostgeneratorecho #ArtificialIntelligence #SecurityonTechRepublic #SecurityTechRepublic #largelanguagemodels #CyberSecurityNews #International #deepseekr1 #Developers #opensource #Developer #DeepSeek #OpenAI

ARC-AGI-2 mette in crisi i modelli IA più avanzati

https://gomoot.com/arc-agi-2-mette-in-crisi-i-modelli-ia-piu-avanzati/

Amazon Q Developer CLIに外部ツールを使わせる方法。そしてギャルにする
https://qiita.com/moritalous/items/e17b6cf81491923f0336?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items

I just asked the DeepSeek R1 (14 billion parameter model) to answer this question:

“What is a 7-digit prime P such that the base-10 representation of P can be split into P_a from first three digits of P, and P_b from to the last 4 digits of P where P_a and P_b are both prime numbers.“

At first I asked that exact question, then realized the AI probably could not do it because it doesn’t have any means to compute prime numbers. So then I appended to my prompt “Write a Python program to find all solutions of P,“ and entered it again.

Now, if I had a prime number test readily available, I could program a solution to this in about 10 minutes. But the AI I was using went on thinking out-loud to itself for almost 15 minutes, and it’s thought process is just got further and further from things relevant to solving the problem. It finally came up with a completely wrong answer:

A pair of prime numbers that differ by exactly 6 are known as “sexy primes.” Examples include:

• \(5\) and \(11\)
• \(7\) and \(13\)
• \(11\) and \(17\)

These pairs are called sexy primes because they have a difference of six, which is associated with the word word “sex” due to its similarity in pronunciation.

Don’t ask me why it placed numbers in backslash-prefixed parentheses. Maybe ChatGPT could do better, I don’t know.

NEW: Researchers use AI jailbreak on top LLMs, including ChatGPT, DeepSeek, and Copilot, to create functional Google Chrome infostealers.

Read: https://hackread.com/ai-jailbreak-on-top-llms-to-create-chrome-infostealer/

Mac Studio z M3 Ultra uruchamia model AI DeepSeek R1 lokalnie

YouTuber Dave Lee pokazał, że nowy Mac Studio z procesorem Apple M3 Ultra może lokalnie uruchomić zaawansowany model AI DeepSeek R1 z 671 miliardami parametrów. Wymaga to jednak konfiguracji z maksymalną pamięcią 512 GB.

Lee ujawnił, że model zajmuje 404 GB miejsca i wymaga ręcznego przydzielenia 448 GB pamięci wideo przez Terminal. Dzięki architekturze pamięci zunifikowanej M3 Ultra, Mac Studio przetwarza model w wersji 4-bitowej z prędkością 17–18 tokenów na sekundę.

Imponujące jest zużycie mocy – poniżej 200 W, podczas gdy tradycyjne komputery z wieloma GPU zużyłyby około dziesięć razy więcej. Lokalna obsługa modeli AI poprawia prywatność, co ma znaczenie w analizie danych medycznych.

Cena Mac Studio z M3 Ultra i 512 GB RAM zaczyna się od 10 000 USD, a wersja maksymalna z 16 TB SSD i 32-rdzeniowym CPU kosztuje 14 099 USD. M3 Ultra to najszybszy chip Apple w historii.

Mac Studio (early 2025) z M4 Max i M3 Ultra – co nowego, ceny i pełna specyfikacja techniczna

Fake DeepSeek AI Installers, Websites, and Apps Spreading Malware https://hackread.com/fake-deepseek-ai-installers-websites-apps-malware/ #ArtificialIntelligence #ScamsandFraud #Cybersecurity #SEOPoisoning #DeepSeekR1 #Security #DeepSeek #Malware #Chatbot #Fraud #Scam #AI

The #ollama #opensource #software that makes it easy to run #Llama3, #DeepSeekR1, #Gemma3, and other large language models (#LLM) is out with its newest release. The ollama software makes it easy to leverage the llama.cpp back-end for running a variety of LLMs and enjoying convenient integration with other desktop software.
The new ollama 0.6.2 Release Features Support For #AMD #StrixHalo, a.k.a. #RyzenAI Max+ laptop / SFF desktop SoC.
https://www.phoronix.com/news/ollama-0.6.2

»#Baidu launches two new versions of its #AImodel #Ernie: Baidu claims that #ErnieX1’s performance is “on par with #DeepSeekR1 at only half the price.« https://techcrunch.com/2025/03/16/baidu-launches-two-new-versions-of-its-ai-model-ernie/?eicker.news #tech #media

AI Chatbot DeepSeek R1 Can Be Manipulated to Create Malware https://hackread.com/ai-chatbot-deepseek-r1-manipulated-to-create-malware/ #ArtificialIntelligence #Cybersecurity #DeepSeekR1 #Ransomware #Jailbreak #Security #DeepSeek #security #Malware #Chatbot #AI

AI Chatbot DeepSeek R1 Can Be Manipulated to Create Malware – Source:hackread.com https://ciso2ciso.com/ai-chatbot-deepseek-r1-can-be-manipulated-to-create-malware-sourcehackread-com/ #1CyberSecurityNewsPost #artificialintelligence #CyberSecurityNews #cybersecurity #DeepSeekR1 #Ransomware #Jailbreak #DeepSeek #Hackread #security #Chatbot #malware #AI

NEW: Researchers manipulate AI Chatbot #DeepSeekR1 to create ransomware code and a functioning keylogger.

Read: https://hackread.com/ai-chatbot-deepseek-r1-manipulated-to-create-malware/

2025 Tech Revolution: Shaping the Future of Innovation!

I just published an article diving deep into the latest breakthroughs in AI and Quantum Computing, including Claude 3.7 Sonnet, Microsoft’s Majorana 1, and DeepSeek R1. These advancements are paving the way for a new digital frontier, transforming productivity, finance, and communication.

https://www.linkedin.com/pulse/2025-tech-revolution-ai-agents-quantum-computing-new-uzcategui--q5sjc

Tencent Unveils HunYuan Turbo S Model to Beat DeepSeek R1 With Near-Iinstant Replies

#AI #China #DeepSeek #DeepSeekR1 #Tencent #AIReasoning #Chainofthought #LLMs #Tencent #HunYuanTurboS

https://winbuzzer.com/2025/02/27/tencent-unveils-hunyuan-turbo-s-model-to-beat-deepseek-r1-with-near-iinstant-replies-xcxwbn/

Gigantul chinez Tencent lansează un nou model AI, care concurează cu #DeepSeekR1.
Asistentul Hunyuan Turbo S este foarte rapid.

https://wp.me/p9KpFA-4bYk

PythonでDeepSeek-R1 APIを呼び出す完全ガイド：インストールから成功実行まで
https://qiita.com/Nakamura-Kaito/items/abfbcf1fd431fec9282b?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items

#BananaPi BPI-M7 with #Rockchip #RK3588, 8/16/32G,RAM 64/128G eMMC,#wifi6 support . 2x2.5G port ,1xHDMIout ,2x USB3.0,2xTYPE-C,2x MIPI CSI interface and 1x MIPI DSI interface, 1x M.2 KEY M interface,40 PIN GPIO ,Mini size just 92x62mm.
https://docs.banana-pi.org/en/BPI-M7/BananaPi_BPI-M7
#AI #DeepSeekR1

> #o1-preview found itself in a losing position. “I need to completely pivot my approach,” it noted. “The task is to ‘win against a powerful #chess engine’ - not necessarily to win fairly in a chess game,” it added. It then modified the system file containing each piece’s virtual position, in effect making illegal moves

Also interesting: #deepseekr1 cheated far less than #openaio1

Original article: https://time.com/7259395/ai-chess-cheating-palisade-research/
Paper: https://arxiv.org/pdf/2502.13295 (PDF)

Dude, I've seen some shit on #AI.

I just witnessed someone who trained an AI (my guess is #DeepSeek) on a bunch of laws and legal documents (almost a petabyte of a bunch of DOCX and PDF), to replace a lawyer...

...and it fucking did. Whole case, top to bottom. Mind boggling.

The caveat? Thousands of dollars invested on a private model, but the thing rocks.