This article explains the fundamentals of GPT models, contrasts large models with traditional AI, details ChatGPT's architecture and token processing, outlines its limitations, and presents a five‑layer framework for applying ChatGPT across chat, language, text, reasoning, and private model use cases.
This article details the step‑by‑step process of preparing the environment, merging weights, installing dependencies, running inference, evaluating Vicuna‑7B against other models, and attempting fine‑tuning, while highlighting performance results, encountered issues, and future work for large language model deployment.
On May 16, 360 and Zhipu AI announced a strategic partnership to co‑develop the trillion‑parameter models 360GLM and 360GPT, positioning them as China’s answer to Microsoft‑OpenAI by combining large‑scale pre‑training, bilingual capabilities, and integration with 360’s search and browser ecosystem.
Baidu’s AI infrastructure combines a massive InfiniBand‑linked GPU cluster, Kunlun chips, the PaddlePaddle framework, and the Wenxin model suite with 4D hybrid parallelism, elastic fault tolerance, and a two‑stage training pipeline to overcome computation, memory, and communication walls, delivering world‑leading MLPerf performance for large‑scale LLMs.
The article examines Baidu’s groundbreaking knowledge‑deep learning dialogue system that earned the 2022 Wu Wenjun AI Science and Technology Award, detailing its technical breakthroughs, patent portfolio, large‑scale deployments, and how it underpins China’s rapid advancement in large language models and AI industry integration.
This article reviews Alibaba's Tongyi Qianwen large language model, shares personal impressions, provides a fish‑flavored pork recipe, conducts a SWOT analysis, and includes Scala Spark and Java code examples illustrating its capabilities and usage scenarios.
This article explains how ChatGPT works by covering the fundamentals of natural language processing, generative language models, deep learning, the Transformer architecture, attention mechanisms, few‑shot learning, and the reinforcement‑learning techniques that align its outputs with human preferences.
The article reviews Alibaba’s Tongyi Qianwen large‑language model, shares a cooking recipe generated by the AI, presents a SWOT analysis, and provides code examples—including a Spark Scala script for same‑period month‑over‑month calculations and a Java bubble‑sort implementation.
This article details the evolution of Fudan University's open‑source MOSS series—from the early OpenChat 001 prototype to the current MOSS 003—covering data collection, multilingual capabilities, plugin architecture, model releases on HuggingFace, and how developers can start using the models.
The article details the evolution of the open‑source MOSS series—from OpenChat 001 to MOSS 003—covering data collection, fine‑tuning procedures, multilingual capabilities, plugin architecture, example code for inference, and upcoming releases, providing a comprehensive technical overview for AI practitioners.
Chinese tech giants including Baidu, Tencent, Alibaba, ByteDance, Huawei, NetEase, and 360 have launched or are developing their own ChatGPT-like large language model products, detailing release dates, technical specifications, applications, and the challenges faced in the domestic AI landscape.
This article examines OpenAssistant, the world’s largest open‑source ChatGPT replica, detailing its dataset of over 160 k annotated conversations, the fine‑tuned LLaMA and Pythia models, evaluation results against GPT‑3.5‑turbo, practical usage examples, and the project's current limitations and future directions.
GPT‑4, the new multimodal AI model, can process images and text, generate code and natural language, achieve human‑level scores on standardized exams, handle up to 32 K tokens, and demonstrates advanced reasoning, while OpenAI emphasizes its safety improvements and current limitations as a still‑emerging technology.
OpenAI’s GPT‑4 arrives with multimodal vision, a dramatically longer context window, higher exam scores, Socratic prompting, improved safety, and new partnerships, while still in research mode and subject to bias and code‑trust limitations.
The article reviews Baidu’s newly launched Ernie Bot, a multimodal large language model, comparing its literary, business, mathematical, Chinese comprehension, and multimodal abilities with GPT‑4, while detailing the underlying technologies, knowledge‑enhancement techniques, and deployment strategy behind the model.
OpenChatKit, released by Together Computer in March 2023, is an open‑source ChatGPT‑like large language model built on GPT‑NeoX‑20B, offering developers fine‑tuned conversational abilities, a modular architecture, retrieval system, and content‑filtering, while also outlining its current limitations and future potential.
The article reviews Baidu's launch of the multimodal large language model Wenxin Yiyan, compares its literary, business, mathematical, Chinese‑understanding and multimodal abilities with GPT‑4, explains the underlying six‑core technologies and hardware stack, and reports the mixed market and netizen response.
The article presents a side‑by‑side demonstration of Baidu’s newly released Wenxin Yiyan and OpenAI’s GPT‑4 across literary creation, business copywriting, mathematical reasoning, Chinese idiom interpretation, acrostic poetry, and multimodal generation, then explains the underlying six‑core technologies and Baidu’s hardware‑cloud strategy while reporting audience reactions.
This article provides a detailed analysis of GPT‑4, covering its multimodal abilities, performance gains, training innovations, safety improvements, new application scenarios, impact on developers, and future trends in large language models.
OpenAI’s GPT‑4, a multimodal large language model that accepts text and image inputs, powers ChatGPT and Bing, offers improved creativity and problem‑solving while still facing hallucination risks, and is now available via ChatGPT Plus and an open API for developers.
OpenAI's GPT‑4, a large multimodal language model, expands token limits, adds image understanding, demonstrates strong reasoning on professional exams, supports many languages, and is already integrated into Microsoft Bing, while offering various access options and improved safety compared to its predecessor.
This article details Tencent's development of the 1‑trillion‑parameter HunYuan‑NLP model, covering its MoE architecture, cost‑effective pre‑training strategies, distributed training framework, model compression toolkit, and successful deployment across advertising, gaming, and other Tencent services.
Fudan University's Natural Language Processing Lab introduced MOSS, the country's first conversational large language model capable of dialogue generation, programming, factual QA and ethical reasoning, with plans for open‑source release despite current limitations in Chinese language proficiency.
The article presents Baidu's PLATO open‑domain dialogue system, detailing its evolution from expert‑rule to retrieval‑based and large‑scale generative models, describing its hidden‑variable architecture, major research challenges such as persona stability, long‑term memory, knowledge accuracy, and showcasing real‑world applications and Q&A from a DataFunSummit2022 livestream.
MOSS, a Chinese large‑language model released by Fudan University, offers ChatGPT‑like functions such as text generation, summarization, translation, and code writing, while being open‑source and free during preview, yet it still lags behind due to limited data, compute, and model size.
This article presents an overview of ChatGPT technology, its evolution, current challenges, a three‑stage learning process, data organization and evaluation, details of domestic localization efforts, practical solutions, and the release of a Chinese open‑source large model with training guidance.
This article provides a comprehensive, step‑by‑step explanation of how ChatGPT generates text, covering token probabilities, n‑gram models, embeddings, attention mechanisms, and the Transformer architecture, while illustrating concepts with Wolfram‑language examples and visualizations.
Alibaba’s internal‑test “Damoyuan‑version ChatGPT” showcases a multimodal AI that combines text, image, code, and creative generation, built on the Tongyi large model’s unified architecture, while other Chinese tech giants rush to launch their own ChatGPT‑style products.
Since its November 2022 launch, OpenAI’s ChatGPT—built on Transformer‑based generative AI—has surged to over 100 million users, demonstrated capabilities from MBA exams to software‑engineer interviews, sparked a multibillion‑dollar market with paid subscriptions and Microsoft investment, spurred rival models like Claude, and is reshaping human‑computer interaction while raising ethical concerns and promising multimodal, industry‑specific future applications.
This article provides a comprehensive overview of ChatGPT, covering its origins within OpenAI, core features, underlying GPT‑3.5 architecture, reinforcement learning from human feedback, current limitations, and future directions such as model compression, RLAIF, and expanding industry applications.
Microsoft announced a major AI‑powered upgrade to Bing and Edge, integrating a next‑generation OpenAI large language model and its proprietary Prometheus system to deliver enhanced search results, full answers, interactive chat, and content generation, marking a new era of AI‑driven web interaction.
Google unveiled its new conversational AI, Bard, built on the LaMDA model, positioning it as a direct competitor to ChatGPT; the article details its public testing, technical foundations, feature set, and key differences such as real‑time web integration and cost‑free access.
This article provides an in‑depth analysis of ChatGPT, covering its GPT‑3.5 foundation, RLHF training pipeline, key features, technical limitations, model compression methods, and the broader industry impact and investment prospects of large language models.
An in‑depth overview of ChatGPT covering its conversational model nature, core technologies such as InstructGPT, large language model capabilities, RLHF training pipeline, strengths, limitations, safety mechanisms, and potential applications across content creation, search, and multimodal integration.
This article explains ChatGPT's underlying technology—including its three‑stage training pipeline with supervised fine‑tuning, reward‑model learning, and reinforcement learning from human feedback—while analyzing whether the model can realistically replace traditional search engines such as Google or Baidu.
The article explains ChatGPT’s underlying technology, detailing its three-stage training pipeline—supervised fine‑tuning, reward‑model learning, and reinforcement learning with PPO—while discussing its strengths, limitations, and potential integration with traditional search engines.
The article outlines ChatGPT’s conversational capabilities, its InstructGPT‑based architecture, a three‑stage RLHF training pipeline involving supervised fine‑tuning, human‑ranked response generation, and PPO optimization, and discusses its strengths, limitations, diverse applications, and future directions for multimodal, up‑to‑date assistants.
The article explains how ChatGPT builds on the GPT‑3.5 large language model, using human‑annotated data and Reinforcement Learning from Human Feedback (RLHF) across three training stages to improve instruction understanding, answer quality, and continual model enhancement, while also discussing its potential to complement or replace traditional search engines.
This article explains the technical foundations of ChatGPT, detailing its three-stage training pipeline—supervised fine‑tuning with human‑annotated data, reward model training via pairwise ranking, and reinforcement learning from human feedback—while also discussing its limitations compared to traditional search engines and potential future enhancements.
The article explains how ChatGPT builds on GPT‑3.5 with supervised fine‑tuning, reward‑model training and reinforcement learning from human feedback, analyzes why it cannot yet replace search engines due to hallucinations, knowledge freshness and cost, and proposes a hybrid architecture that combines LLM generation with traditional retrieval to overcome these limitations.
This article introduces Baidu's PLATO open‑domain dialogue technology, explains the evolution from rule‑based to retrieval‑based and large‑scale generative models, discusses major challenges such as persona stability, long‑term memory, knowledge accuracy, and proactive conversation, and showcases real‑world applications and Q&A insights.
The OPPO XiaoBu team introduced OBERT, a series of 100M‑, 300M‑, and 1B‑parameter pretrained language models that leverage massive TB‑scale corpora, multi‑granular masking, retrieval‑augmented training, and distributed acceleration to achieve state‑of‑the‑art results on CLUE and KgCLUE benchmarks while enabling efficient industrial deployment.
The article examines the controversy surrounding Google’s LaMDA chatbot, detailing engineer Blake Lemoine’s claims of sentience, his suspension, the model’s technical specs, contrasting expert opinions from figures like Andrej Karpathy and Gary Marcus, and ultimately argues that LaMDA’s apparent emotions are a projection rather than true consciousness.
Tencent’s Hunyuan AI large model shattered records by scoring 80.888 to claim first place on the CLUE benchmark, showcasing its advanced natural language processing, multimodal abilities, curriculum‑learning training approach, and real‑world deployments in WeChat Search and advertising.
Carnegie Mellon researchers introduced PolyCoder, a 27‑billion‑parameter open‑source code generation model built on GPT‑2, trained on 249 GB of multi‑language code and achieving superior performance to Codex in C while remaining competitive across eleven other programming languages.
Baidu's newly released PLATO‑XL, a 110‑billion‑parameter bilingual pre‑training dialogue model, surpasses previous large‑scale models, introduces multi‑role awareness for consistent multi‑turn conversations, and demonstrates state‑of‑the‑art performance across open‑domain, knowledge‑grounded, and task‑oriented dialogue tasks.
Baidu's ERNIE 3.0 introduces a 100‑billion‑parameter, knowledge‑graph‑augmented language model that breaks 54 Chinese NLP benchmarks, achieves human‑level performance on SuperGLUE, and demonstrates strong generation and zero‑shot capabilities, now available for public demo and research.
