California's AI Safety Bill Faces Industry Pushback and Amendments
California's proposed legislation to regulate AI, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047), has been significantly revised in response to concerns from major AI companies and lawmakers. The bill, which originally aimed to establish strict safety regulations for advanced AI models, now includes amendments to address the tech industry's fears of stifling innovation and potential economic impacts.
State Senator Scott Wiener, the bill's author, noted that while the amendments don't fully satisfy all industry requests, they strike a balance between innovation and safety. Notably, the bill now limits enforcement penalties, removes criminal perjury provisions, and narrows whistleblower protections. The revisions reflect a compromise, ensuring that California remains a leader in AI development while addressing public safety concerns.
As AI continues to advance, the debate over regulation versus innovation will undoubtedly intensify. SB 1047's outcome could set a precedent for how other states and countries approach AI governance.
#AIRegulation#AIModels
Google Empowers Web Publishers with Data Control
Hello, everyone! Google has introduced a new tool called Google-Extended, allowing website publishers to opt out of using their data to train Google's AI models, all while maintaining accessibility through Google Search.
This tool provides a solution for publishers who wish to have their sites indexed by Google's web crawlers but prefer not to contribute their data to the development of AI models like Bard and Vertex AI generative APIs.
Bard, Google's AI chatbot, has been undergoing training using publicly available web data. With Google-Extended, publishers can now manage their site's participation in this process.
Google-Extended is integrated into the robots.txt file, which is used to inform web crawlers about site accessibility. It grants web publishers the ability to control AI model training access and content availability.
#AIandLaw#GoogleAI#WebPublishers#AIModels
⚡Gemini 3 Flash - быстрый ИИ нового поколения от Google
Gemini 3 Flash:
- это очень быстрая модель с минимальной задержкой
- при этом она сохраняет сильные способности к рассуждению
- Frontier-уровень на GPQA Diamond - рассуждения уровня PhD
- Хорошие результаты на Humanity’s Last Exam
- State-of-the-art на MMMU Pro - хорошо работает с видео и мультимодальными данными
- В целом, качество сопоставимо с Gemini 3 Pro
Стоит в четыре раза дешевле, чем Gemini 3.0 Pro, при этом показывает сопоставимые результаты почти во всех бенчмарках, включая HLE и ARC-AGI 2. На некоторых бенчмарках модель обходит GPT-5.2.
Более того, в ряде тестов модель даже превосходит более мощную версию Pro, оставаясь при этом значительно дешевле.
По сути, Flash - это попытка Google сбалансировать три вещи одновременно: скорость + интеллект + стоимость.
Цены:
- Text input: $0.30 per 1M tokens
- Text output: $2.50 per 1M tokens
- Cache read: $0.075 per 1M tokens
- Input audio: $0.999 per 1M tokens
- Input audio (cached): $0.249 per 1M tokens
- Web search: $0.035 per request
- Cache storage: $1 per hour per 1M tokens
https://blog.google/products/gemini/gemini-3-flash/
@ai_machinelearning_big_data
#AI#Gemini#Google#LLM#Multimodal#AIModels#MachineLearning
New high-volatility geopolitical question live ⚠️
Will a senior political leader become the target of an assassination attempt?
Markets react before media confirms.
AI tracks signal shifts in real time.
Watch sentiment move.
#Geopolitics#Prediction#MiddleEast#AIModels#Forecasting#RiskSignals#EdgeMarket
🚀 AI TRENDS | Meta Platforms Reallocates Engineers to Enhance AI Models
Meta Platforms is reportedly shifting its engineering resources to improve its AI models. According to Jin10, this strategic move aims to bolster the company's capabilities in artificial intelligence, reflecting its commitment to advancing technology in this field. The reallocation of engineers is part of Meta's broader initiative to enhance its AI infrastructure and deliver more sophisticated AI solutions.
#Meta#AI#ArtificialIntelligence#Engineering#TechInnovation#AIModels#MetaPlatforms
Zoom Addresses EU Privacy Concerns and Updates Terms of Service
Greetings! In response to discussions about potential EU privacy law implications, Zoom issues a statement and revises its Terms of Service. The focus? Ensuring customer data isn't utilized to train AI models.
Zoom's statement and Terms affirm that user-generated content, including audio, video, chat, and more, isn't employed for training Zoom's or any third-party AI models. This step aims to dispel any concerns.
Zoom initially shared its statement on August 7 and later updated it on August 11, aligned with the revised Terms. The shared stance now unequivocally states, "Zoom does not use any of your customer content to train AI models."
Earlier, a Stack Diary article flagged changes to Zoom's March Terms, raising potential concerns about broad data utilization for AI model training. Zoom's quick response aims to address these concerns and reaffirm privacy commitments.
#Zoom#PrivacyMatters#TermsOfService#AIModels#DataProtection#PrivacyLaw#TechUpdates
🚀 AI TRENDS | Moore Threads Completes Rapid Adaptation of MTT S5000 GPU for MiniMax M2.7
Moore Threads has announced the successful completion of the Day-0 rapid adaptation of its flagship AI training and inference all-in-one GPU, the MTT S5000, for the new generation large model, MiniMax M2.7. According to Odaily, this achievement further demonstrates the capability of domestically produced full-feature GPUs to quickly respond to and support cutting-edge AI models.
#MooreThreads#MTTS5000#AItraining#GPU#MiniMaxM2.7 #AImodels#domesticGPU
⚡️ Omni-Embed-Nemotron - новая единая модель от NVIDIA для поиска по тексту, изображениям, аудио и видео
Модель обучена на разнообразных мультимодальных данных и может объединять разные типы входных сигналов в общее векторное представление.
- Поддержка всех типов данных: текст, изображение, аудио, видео.
- Основана на архитектуре Qwen Omni (Thinker-модуль, без генерации текста).
- Контекст - до 32 768 токенов, размер embedding — 2048.
- Оптимизирована под GPU, поддерживает FlashAttention 2.
Это делает её идеальной для:
- кросс-модального поиска (поиск текста по видео или изображению);
- улучшения RAG-проектов;
- систем мультимодального понимания контента.
Просто, быстро и эффективно - всё в одном открытом решении.
🌐 Открытая модель: https://huggingface.co/nvidia/omni-embed-nemotron-3b
@ai_machinelearning_big_data
#crossmodal#retrieval#openAI#NVIDIA#OmniEmbed#multimodal#AIModels#OpenSource#Search#UnifiedEmbedding
Bitcoin to Hit $150,000 Next Year
🔭 Galaxy Research predicts Bitcoin will cross $150,000 in H1 2024 and could reach $185,000 by Q4 2025. Analysts note Bitcoin has consistently outperformed other asset classes, expecting its market cap to reach 20% of gold's value.
🔗 Read more: Leviathan News
📉 Total market cap of AI tokens has dropped by 28% recently, now at $50.93 billion.
#Bitcoin#AI#Crypto#Galaxy#MarketTrends
#Bitcoin#AI#Crypto#Galaxy#MarketTrends#DeFi#Investing#Trends#Predictions #2024 #2025 #Gold#InstitutionalInvestors#TechNews#AIModels#Innovation#MarketCap#S&P500 #Cryptocurrency#Blockchain
⚡️ HunyuanOCR: открытая OCR-модель, которая рвёт бенчмарки при размере всего 1B
Tencent выложила в open-source новую модель HunyuanOCR.
Это компактная, быстрая и полностью готовая end-to-end система для OCR, построенная на мультимодальной архитектуре Hunyuan.
Главное - при размере только 1 миллиард параметров она показывает результаты уровня крупных моделей и стоит в разы дешевле в запуске.
⚡ Топ по бенчмаркам
• 860 на OCRBench среди всех моделей до 3B
• 94.1 на OmniDocBench - лучший результат в задачах распознованяисложных документов
🌐 Что умеет HunyuanOCR
Модель закрывает практически все типы OCR задач
• текст на улицах, витринах, табличках
• рукописный текст и художественные шрифты
• сложные документы: таблицы, формулы, встроенный HTML и LaTeX
• субтитры в видео
• перевод текста на фото end-to-end сразу на 14 языков
Это не каскадный пайплайн, а единое решение
Один запрос и одно инференс-прогон дают готовый результат.
Это быстрее, надёжнее и удобнее, чем традиционные OCR-цепочки.
📌 Project Page
web: https://hunyuan.tencent.com/vision/zh?tabIndex=0
mobile: https://hunyuan.tencent.com/open_source_mobile?tab=vision&tabIndex=0
🔗GitHub
https://github.com/Tencent-Hunyuan/HunyuanOCR
🤗 Hugging Face
https://huggingface.co/tencent/HunyuanOCR
📄 Technical Report
https://github.com/Tencent-Hunyuan/HunyuanOCR/blob/main/HunyuanOCR_Technical_Report.pdf
@ai_machinelearning_big_data
#HunyuanOCR#TencentAI#OCR#VisionAI#DeepLearning#Multimodal#AIModels#OpenSourceAI#ComputerVision#DocumentAI
Weekly AI Digest: Key Developments
🔹 Anthropic raises $3.5B through incremental funding rounds. Read more
🔹 QwQ 32B launched, slightly trailing top performers. Details here
🔹 Wan 2.1 remains a top open-source model amid competition. Learn more
🔹 Hunyuan Image2Video: Tencent's response to Alibaba's offering. Explore here
🔹 SourceCraft introduces cloud-based team development as VM replacement. More info
🔹 Apple's Mac Studio handles demanding models and LLMs. Discover more
🔹 SpeechSense analyzes customer conversations using LLM technology. Details
🔹 RL wins Turing Award, recognized with a $1M prize. Full story
🔹 Bitcoin drops to $82,223; Ethereum to $1,998, with $243M in liquidations. Read market update.
#AI#Crypto#VC#Anthropic#Bitcoin#Ethereum#Hunyuan#Tencent#AIModels#OpenSource#MacStudio#SpeechSense#Investments#Funding#TuringAward#Liquidations#QwQ#SourceCraft#SmartTech
AI Startup Anthropic Expands Features
🚀 Anthropic to launch a two-way voice mode for Claude chatbot and enhance memory features for better user interaction, announced CEO Dario Amodei. Google has invested an additional $1 billion, totaling $3 billion, to support the company's operations.
📈 Expect smarter models in the coming months as Anthropic scales up to meet growing AI demand.
🔗 Read more: Forklog
#AI#Investments#Google#Tech#Startup#Claude#VoiceTechnology#MachineLearning#Innovation#Funding#Chatbots#DarioAmodei#AIModels#Scalability#UserExperience#NeuralNetworks#Computing#AIIndustry#AIExpansion#FutureTech#VC