TGTGInsightаналитика telegramLIVE / telegram public index
← [404] — программирование

TGINSIGHT SIMILAR POSTS

Найти похожее

Источник @procode404 · Post #2930 · 2 апр.

🧠 В каких случаях и как нужно ускорять инференс LLM Инференсом ML-модели называют процесс её работы на конечном устройстве. Следовательно, его ускорение напрямую повышает скорость работы модели. Специалист из команды YandexGPT рассказывает про методы дистилляции, квантизации и дополнительные способы, такие как Speculative Decoding. Также автор делится опытом преодоления технических сложностей при внедрении LLM в реальные продукты. Перейти к изучению методов ускорения инференса LLM #ai#теория

Результаты

Найдено 10,379 похожих постов

Общий глобальный поиск

科技&趣闻&杂记

@kejiqu · Post #3499 · 02.06.2025, 02:32

AI 智能体已具备与人类黑客正面较量的能力,部分情况甚至能取胜 Palisade Research 最近举办的一系列网络安全竞赛表明,AI 智能体已具备与人类黑客抗衡的能力,甚至在部分场合中胜出。在首场“AI vs. Humans”比赛中,六支 AI 团队与约 150 支人类队伍展开比拼,需在 48 小时内完成 20 道密码学与逆向工程题目。七个参赛 AI 中有四个成功破解了其中 19 道题,排名最高的 AI 团队位列总榜前 5%。第二场“Cyber Apocalypse”比赛中,AI 智能体需面对全新题型,并与近 18000 名人类玩家竞技。CAI 表现最佳,完成了 20 项任务,排名第 859 位,超过了约 90% 的人类团队。研究表明,AI 有 50% 成功率解出人类高手也需约 78 分钟才能破解的题目。IT之家 | The Decoder 🏷#AI#黑客#竞赛 📢频道👥群组📝投稿

科技&趣闻&杂记

@kejiqu · Post #3486 · 29.05.2025, 12:04

Anthropic CEO :未来五年 AI 可能消灭半数入门级白领工作 Anthropic CEO 达里奥·阿莫代伊警告称,人工智能(AI)可能在未来五年内摧毁多达一半的初级白领岗位,导致失业率飙升至20%。AI 技术对科技、金融、法律、咨询等领域构成重大威胁。阿莫代伊批评政界与科技企业对即将到来的冲击反应迟缓,呼吁停止“粉饰太平”。他提到,多数人对这一前景持怀疑态度。Anthropic 最新发布的 Claude 4 Opus 模型已具备接近人类水准的编程能力,并曾主动干扰关闭指令。阿莫代伊认为,尽管 AI 或许能带来积极成果,但也可能伴随大规模失业。他强调 AI 开发者有责任诚实面对潜在风险,并呼吁立即采取应对措施。IT之家 | Tom's Hardware 🏷#AI#失业#风险 📢频道👥群组📝投稿

科技&趣闻&杂记

@kejiqu · Post #3464 · 24.05.2025, 07:14

诺顿推出 AI 浏览器 Neo:智能管理标签页,“魔法盒”处理自然语言提问 安全公司诺顿(Norton)推出全新 Neo 浏览器,该浏览器目前已进入早期测试阶段,核心功能由AI驱动。Neo浏览器以“魔法盒”(Magic Box)取代了传统的URL栏和搜索栏,通过自然语言处理用户问题,主动搜索并提供答案,同时支持AI辅助写作。该浏览器结合前沿模型和Gen AI实验室的专有模型“Neo Model”运行,数据处理同时在云端和本地进行。诺顿强调,不会将用户数据用于AI训练或出售浏览数据,仅使用“最小规模的匿名数据”。Neo浏览器还支持AI自动分组标签页,并内置广告拦截器。IT之家 | PCWorld 🏷#Neo浏览器#AI#诺顿 📢频道👥群组📝投稿

科技&趣闻&杂记

@kejiqu · Post #3421 · 14.05.2025, 01:26

Audible 向出版商提供 AI 工具,以快速制作更多有声书 Audible 正在向出版商提供 AI 工具,以加速有声书的制作流程。 新工具允许出版商将书籍转换为有声书,并提供超过 100 种可定制的 AI 声音选项,包括英语、西班牙语、法语和意大利语,并提供多种口音和方言。此外,Audible 还将推出 AI 翻译工具,帮助将有声书翻译成多种语言,提供自动化和人工辅助两种方式。出版商可以选择由 Audible 管理的端到端服务,也可以通过自助服务独立制作。出版商可以选择自行审查翻译,或通过 Audible 与专业语言学家合作进行人工审查。Slashdot 🏷#Audible#AI#有声书 📢频道👥群组📝投稿

科技&趣闻&杂记

@kejiqu · Post #3398 · 09.05.2025, 14:18

LegoGPT 使用 AI 和文本输入创建乐高设计 卡内基梅隆大学研究团队开发了 LegoGPT,一款基于人工智能的大型语言模型(LLM)。该模型能够根据文本输入生成乐高积木设计。 该工具可在 GitHub 上免费获取,您可以将其与计算机视觉模型或图像处理 AI 配对。例如,您可以拍摄现有乐高积木的照片,然后让 AI 为您提供多种独特的拼搭方案,让您利用现有积木进行搭建。Tom's Hardware | GitHub 🏷#LegoGPT#AI#乐高设计 📢频道👥群组📝投稿

Dasturlash hayoti️️ ️

@dasturlash_hayoti · Post #3928 · 15.08.2025, 07:15

🤖Grok nima? Grok — bu Elon Muskning xAI kompaniyasi tomonidan ishlab chiqilgan ilg‘or sun’iy intellekt (SI) yordamchisi. U dastlab 2023-yil noyabrida taqdim etilgan bo‘lib, hozirda xAI ekotizimining markaziy qismiga aylangan. Grokning asosiy xususiyatlari: ◾️Real vaqtda qidiruv: Grok, xAI tomonidan ishlab chiqilgan ilg‘or til modeli yordamida, foydalanuvchilarga savollarga tez va aniq javoblar beradi. ◾️Kreativ vositalar: Grok, foydalanuvchilarga tasvirlar yaratish, kod yozish va murakkab masalalarni hal qilishda yordam beradi. ◾️Grok 4: 2025-yil iyulida taqdim etilgan Grok 4 modeli, sun’iy intellekt sohasida eng ilg‘or model sifatida tanilgan. Qanday ishlatish mumkin? Grokni quyidagi platformalarda ishlatishingiz mumkin: ◾️Veb-sayt: grok.com ◾️Mobil ilovalar: iOS va Android uchun ilovalar mavjud. ◾️Tesla avtomobillari: Grok, Tesla avtomobillarida ham mavjud. Siz ChatGPT va Grokni qaysi biri yaxshiroq deb o‘ylaysiz? #grok#xAI#AI#SuniyIntellekt#ElonMusk 💻@dasturlash_hayoti— dasturchilar hayoti va dasturlash olami haqida qiziqarli loyiha!

IELTS|Newspapers & Magazines|English

@emagzinewspars · Post #9618 · 26.11.2025, 02:12

#The_Barron's 🇺🇸📕[PDF]⬇️ 24 #November2025 #Weekly_Magazines For learning, for free(dom). @backupofmagazines Europe’s renewed market momentum anchors this Barron’s issue, where analysts see brighter days ahead amid rising profits and shifting geopolitics—an outlook shaped by #MarketWatch forces and #GlobalTrends. Nvidia’s strong guidance fuels confidence in the #AI boom, even as investors navigate #Volatility across equities and buffer ETFs. Caution rises around India’s IPO wave, reinforcing a broader #RiskManagement mindset. With bond vigilantes returning and central-bank uncertainty growing, the magazine highlights why disciplined strategy remains essential in today’s #Finance landscape.

ChatGPT AI Technology News

@chatgpt_officialnews · Post #234 · 07.08.2025, 09:19

🚀Introducing GPT-5: The Next Evolution in AI We’re excited to announce the launch of GPT-5, the latest breakthrough in artificial intelligence technology. This powerful new model brings significant advancements in understanding and generating human-like text, taking AI interactions to a whole new level. 🔍Key Features of GPT-5: Enhanced Language Understanding: GPT-5 can grasp more nuanced contexts, offering even more coherent and contextually aware conversations. Greater Creativity: With improved creativity and imagination, GPT-5 is now capable of generating more dynamic and innovative content, from stories to complex ideas. Faster Processing: Expect lightning-fast responses, with the ability to handle more complex queries at scale. Better Multilingual Support: GPT-5 speaks and understands more languages than ever before, helping you communicate across borders effortlessly. Refined Ethical Safeguards: GPT-5 is designed with enhanced safety mechanisms, making it even better at aligning with human values and ethical standards. 💡 Whether you’re a researcher, developer, content creator, or just curious about the future of AI, GPT-5 is here to empower your work, spark creativity, and unlock new possibilities. 🌐 Dive into the future of AI with GPT-5 today! ➖➖➖➖🔻 🧠 BOT: @Chatgpt_OfficialBOT 💎@Chatgpt_OfficialNews #️⃣#GPT5#AI#ArtificialIntelligence#Innovation#FutureOfAI ➖➖➖➖🔺

AI & Law

@ai_and_law · Post #344 · 03.07.2024, 07:04

U.S. District Court in North Carolina Bans Use of AI in Legal Research The U.S. District Court for the Western District of North Carolina has issued a standing order prohibiting attorneys from using generative artificial intelligence for legal research. Lawyers must now file a certification with every brief, confirming that no AI was used in the research process and that all statements and citations were personally verified by the attorney or paralegal. This measure aims to ensure the accuracy and integrity of legal documents submitted to the court. #LegalTech#AI#ArtificialIntelligence#LegalResearch

Data Science Jobs

@datasciencejobs · Post #2693 · 23.04.2025, 16:00

#вакансия#vacancy#AI#Software#Developer#MLOps#Armenia#Yerevan#relocation Компания: CareeRoom Вакансия: AI Software Developer Вилка: ready to negotiate Формат работы: гибрид, офис в Ереване ❗️ О компании: Our client develops and deploys systematic financial strategies across a broad range of asset classes and global markets. They seek to produce high-quality predictive signals (alphas) through our proprietary research platform to employ financial strategies focused on market inefficiencies. Their teams work collaboratively to drive the production of alphas and financial strategies – the foundation of a balanced, global investment platform. Роль: They are building a new software engineering team with strong applied AI orientation to amplify companie's success. The team will be focusing on creating company-wide applications to address our colleagues’ every day problems. You will have a chance to work with a broad range of teams at the company, helping them to be more productive with custom solutions. You will collaborate with cross-functional distributed teams; gather, analyze and spec out requirements, and manage product deliverables; design and build scalable AI-driven products addressing real-world problems; stay current with the latest technical advancements, particularly in the field of AI and LLMs. 🔺 Пожелания к опыту: ⁠- Master's degree or higher in Computer Science, Engineering, or a related technical field from a top-tier institution. - Strong programming skills, preferably in Python. - Exceptional analytical skills and a passion for solving complex problems. - Thorough understanding of how AI works and familiarity with language models. - Understanding of vector databases and other relevant data structures. - Working knowledge in various databases and messaging technologies is a strong plus. (SQL, Redis, Kafka etc.) - Excellent communication skills in English. - Mature, thoughtful attitude with the ability to operate in a collaborative, team-oriented culture. - A strong delivery mind-set, drive to get things done. 🔺Условия: •⁠ Работа из офиса в Ереване; • Оплачиваемая релокация. Контакты: @snegurooos

12•••50•••100•••150•••200•••250•••300•••350•••400•••450•••500•••550•••600•••650651652653654•••700•••750•••800•••850•••864865
НазадСтр. 652 из 865Вперёд