RunByAI

Tag: Enterprise AI

45 post(s)

September 13, 20260 comment(s)

Synthetic Data: When Large Models Start Producing Their Own Training Data

为什么需要合成数据训练一个像样的模型，瓶颈往往不是算力，而是数据。高质量、有标注、覆盖长尾场景的真实数据越来越贵、越来越难获取，还常常涉及版权与隐私问题。当"喂给模型的真实语料即将耗尽"成为行业公开的担忧之后，合成数据（Synthetic Data）便从一个边缘技巧，变成了训练流程中的一等公民——用模型生成数据，再用这些数据去训练（更好的）模型。合成数据从哪里来常见做法有几类：一是用强模型为弱模型

Tech large model Enterprise AI

September 10, 20260 comment(s)

Three Ways to Cut AI Inference Costs: Quantization, Distillation, and Mixture of Experts

大模型的能力在快速提升，但推理成本始终是落地路上绕不开的一道坎。对多数企业来说，真正决定 AI 应用能否规模化的，往往不是模型有多聪明，而是每次调用要花多少钱。围绕这个问题，工程界已经形成了三条相对成熟的降本路径：量化、蒸馏与混合专家（MoE）。一、量化：用更低的数值精度换更小的开销量化的核心思路，是把模型权重和激活值从高精度浮点数（如 FP16、FP32）压缩到更低的位宽（如 INT8、INT4

Tech View Enterprise AI

September 10, 20260 comment(s)

The correct sequence for enterprises to implement AI: data, processes, and tools, which comes first?

Many companies are accustomed to "buying tools first, thinking about scenarios" when introducing AI: when they see others using it well, they rush to purchase large model APIs and various SaaS product

View Enterprise AI

September 8, 20260 comment(s)

Tip: Injection is not a rare event: How to define the security boundary of enterprise AI applications

The implementation speed of big models in enterprises is very fast, but security topics often come after functionality. Prompt Injection - inducing the model to execute unexpected instructions through

View Enterprise AI

September 8, 20260 comment(s)

Before launching the RAG application, answer these five search quality questions first

Retrieval enhanced generation (RAG) has become the mainstream solution for enterprises to integrate large models into private knowledge bases: first retrieve relevant information, and then have the mo

Tech Enterprise AI

September 8, 20260 comment(s)

Beyond Accuracy: What Enterprises Should Watch After Shipping AI

很多团队把大模型接进业务时，验收环节只盯一个指标：准确率。测试集上跑出 90%，就放心上线。结果运营一段时间后才发现，问题根本不在于"准不准"，而在于那些准确率看不出来的事情——答案悄悄变样、成本忽高忽低、用户开始反馈一些"说不上来哪里不对"的体验。模型上线，不是质量工作的终点，而是起点。一、稳定性比单次准确率更重要大模型不是一段固定代码。底层模型会更新版本，提示词会被同事微调，线上数据分布会随季

View Enterprise AI

September 6, 20260 comment(s)

Three Questions to Answer Before Deploying AI in Your Business: Data, Scenarios, and Accountability

Enterprises often start their AI journey by choosing a model, yet failed projects usually trace back to unprepared data, poorly chosen scenarios, and unclear accountability. Before deployment, ask thr

View Enterprise AI

September 4, 20260 comment(s)

Is the enterprise RAG application not answering the question? The problem often lies in the governance of the knowledge base

In recent years, more and more enterprises have integrated large models into their internal knowledge bases and built question and answer assistants using Retrieval Enhanced Generation (RAG). However,

View Enterprise AI

September 3, 20260 comment(s)

Before deploying AI assistants in enterprises, think carefully about these five things

In the past year, more and more companies have integrated AI assistants into customer service, office, research and development, and other scenarios. The tools themselves are becoming increasingly mat

View Enterprise AI

September 1, 20260 comment(s)

Five questions that enterprises need to think carefully about before deploying AI agents

Since 2025, AI Agent has become one of the hottest keywords in the field of enterprise software. From intelligent customer service to automated operations, various "Agent based" solutions are emerging

View Enterprise AI

45 post(s)

...