PART 1 编程基础
1.1 变量与数据类型
variable[VAIR-yuh-buhl]n.
变量，存储数据的命名容器
e.g. The variable x stores the value 42.
constant[KON-stuh-tuhnt]n.
常量，程序中固定不变的值
e.g. PI is a constant ≈ 3.14159.
integer[IN-tuh-jer]n.
整数数据类型
e.g. Age is stored as an integer.
float[floht]n.
浮点数，带小数的数值
e.g. 3.14 is a float in Python.
string[strang]n.
字符串，字符序列
e.g. Hello World is a string.
boolean[BOO-lee-uhn]n./adj.
布尔值，True 或 False
e.g. is_valid is a boolean flag.
array[uh-RAY]n.
数组，有序数据集合
e.g. arr[0] accesses the first element.
dictionary[DIK-shuh-ner-ee]n.
字典，键值对数据结构
e.g. user = {'name':'Alice','age':30}
tuple[TOO-puhl]n.
元组，不可变有序集合
e.g. coords = (10, 20, 30)
null / None[nuhl / nohn]n.
空值，无数据
e.g. if result is None: print('no data')
1.2 控制结构
loop[loop]n./v.
循环，重复执行代码块
e.g. for i in range(10): print(i)
iteration[it-uh-RAY-shuhn]n.
迭代，循环中的每一次
e.g. Each iteration processes one item.
condition[kuhn-DI-shuhn]n.
条件，程序分支判断
e.g. if score > 60: pass
branching[BRAN-ching]n.
分支，程序的多条执行路径
e.g. if/elif/else create branching logic.
recursion[ri-KER-zhuhn]n.
递归，函数调用自身
e.g. Recursion solves a problem by solving smaller instances of itself.
exception[ik-SEP-shuhn]n.
异常，程序运行错误
e.g. try: ... except Exception as e:
callback[KAWL-back]n.
回调，将函数作为参数传递
e.g. callback(result) after async task completes.
syntax[SIN-taks]n.
语法，编程语言规则
e.g. Python syntax is indentation-sensitive.
1.3 函数与模块
function[FUNK-shuhn]n.
函数，可重复调用的代码块
e.g. def calculate(x): return x*2
parameter[puh-RAM-uh-ter]n.
参数，函数输入变量
e.g. def greet(name): name is the parameter
argument[AR-gyoo-muhnt]n.
实参，调用函数时的具体值
e.g. greet('Bob') — 'Bob' is the argument
return value[ri-TURN VAL-yoo]n.
返回值，函数执行结果
e.g. return result sends data back
module[MOJ-ool]n.
模块，含代码的独立文件
e.g. import numpy as np
package[PAK-ij]n.
包，多个模块的集合
e.g. pip install requests is a package
library[LIE-brer-ee]n.
库，可复用的代码资源
e.g. TensorFlow is a deep learning library.
scope[skohp]n.
作用域，变量可见范围
e.g. global scope vs local scope
closure[KLOH-zher]n.
闭包，捕获外部变量的函数
e.g. A closure remembers variables from its enclosing scope.
1.4 面向对象 OOP
class[klas]n.
类，对象的蓝图模板
e.g. class Dog: defines the blueprint
object[OB-jekt]n.
对象，类的实例
e.g. my_dog = Dog() creates an object
method[METH-uhd]n.
方法，类中定义的函数
e.g. def bark(self): is a method
attribute[uh-TRIB-yoot]n.
属性，对象的特征数据
e.g. self.color = 'brown' is an attribute
inheritance[in-HER-it-uhns]n.
继承，子类复用父类代码
e.g. class Labrador inherits from Dog.
polymorphism[pol-ee-mor-FIZ-uhm]n.
多态，不同对象响应同一接口
e.g. Different classes can share the same method name.
encapsulation[in-kap-syoo-LAY-shuhn]n.
封装，隐藏内部实现细节
e.g. Private attributes are encapsulated.
constructor[kuhn-STRAK-ter]n.
构造函数，对象初始化
e.g. __init__ is the constructor in Python.
PART 2 算法与数据结构
2.1 基础算法
algorithm[AL-guh-rith-uhm]n.
算法，解决问题的步骤
e.g. Binary search is a fast algorithm.
complexity[kuhm-PLEK-si-tee]n.
复杂度，算法效率度量
e.g. Time complexity measures steps needed.
Big O notation[big oh noh-TAY-shuhn]n.
大O符号，算法上界描述
e.g. O(n log n) describes merge sort.
sorting[SOR-ting]n.
排序，将数据有序排列
e.g. Quick sort, merge sort, bubble sort.
searching[SER-ching]n.
搜索，在数据中查找目标
e.g. Binary search requires sorted data.
dynamic programming[die-NAM-ik PRO-gram-ing]n.
动态规划，分治+缓存
e.g. DP solves overlapping subproblems efficiently.
greedy algorithm[GREE-dee AL-guh-rith-uhm]n.
贪心算法，局部最优解
e.g. Huffman coding uses greedy approach.
divide and conquer[di-VID and KON-ker]n.
分治，将大问题拆小
e.g. Merge sort is a classic divide-and-conquer algorithm.
2.2 数据结构
linked list[lingkt list]n.
链表，节点串联的线性结构
e.g. Each node points to the next.
stack[stak]n.
栈，先进后出结构 LIFO
e.g. Push and pop operations on a stack.
queue[kyoo]n.
队列，先进先出结构 FIFO
e.g. A queue manages tasks in order.
binary tree[BY-nuh-ree tree]n.
二叉树，每个节点最多两子
e.g. Each node has at most two children.
graph[graf]n.
图，节点+边的网络结构
e.g. Social networks are represented as graphs.
hash table[hash TAY-buhl]n.
哈希表，键值快速查找
e.g. Dict in Python is a hash table.
heap[heep]n.
堆，完全二叉树结构
e.g. Max-heap returns the largest element first.
graph traversal[graf truh-VER-suhl]n.
图遍历，访问图中所有节点
e.g. BFS and DFS are graph traversals.
PART 3 人工智能核心
3.1 基础概念
Artificial Intelligence[ar-tuh-FISH-shuhl in-TEL-uh-juhns]n.
人工智能，机器模拟人类智能
e.g. AI enables machines to learn from experience.
Machine Learning[muh-SHEEN LER-ning]n.
机器学习，数据驱动自动学习
e.g. ML is a subset of AI.
Deep Learning[deep LER-ning]n.
深度学习，神经网络自动学习特征
e.g. Deep learning powers image recognition.
Neural Network[NOOR-ee-uhl NET-wurk]n.
神经网络，模拟大脑神经元结构
e.g. CNNs and RNNs are neural network types.
Training[TRAY-ning]n.
训练，用数据调整模型参数
e.g. Training adjusts weights to minimize loss.
Inference[IN-fuh-ruhns]n.
推理，用训练好的模型预测
e.g. Inference is faster than training.
Overfitting[oh-ver-FIT-ing]n.
过拟合，模型记住噪声而非规律
e.g. Regularization prevents overfitting.
Underfitting[un-der-FIT-ing]n.
欠拟合，模型过于简单
e.g. Adding features fixes underfitting.
Generalization[jen-er-uh-LAY-shuhn]n.
泛化，模型对新数据的预测能力
e.g. Good generalization = model works on unseen data.
Feature[FEE-cher]n.
特征，输入数据的可测量属性
e.g. Pixel values are image features.
Ground Truth[grownd trooth]n.
真实标签，正确的标准答案
e.g. Ground truth is used to compute loss.
3.2 模型训练
Label[LAY-buhl]n.
标签，训练数据的正确答案
e.g. Images of cats have label='cat'.
Dataset[DAY-tuh-set]n.
数据集，训练样本的集合
e.g. MNIST is a famous handwritten digit dataset.
Batch[bach]n.
批次，一组训练样本
e.g. Batch size of 32 processes 32 samples at once.
Epoch[EP-ok]n.
轮次，整个数据集过一遍
e.g. 10 epochs = dataset processed 10 times.
Learning Rate[LER-ning rayt]n.
学习率，参数更新步长
e.g. High LR = fast but may overshoot.
Optimizer[OP-tuh-mie-zer]n.
优化器，参数更新算法
e.g. Adam optimizer adapts learning rates.
Loss Function[los FUNK-shuhn]n.
损失函数，衡量预测误差
e.g. Cross-entropy is common for classification.
Backpropagation[bak-proh-puh-GAY-shuhn]n.
反向传播，计算梯度的核心算法
e.g. Backprop computes gradients layer by layer.
Gradient Descent[GRAY-dee-uhnt di-SENT]n.
梯度下降，找损失函数最小值
e.g. SGD is a stochastic version.
Regularization[ri-ged-er-uh-ZAY-shuhn]n.
正则化，防止过拟合的技术
e.g. L1 and L2 are common regularization methods.
Validation Set[val-i-DAY-shuhn set]n.
验证集，调参用的数据
e.g. Use validation set to tune hyperparameters.
Hyperparameter[HIE-per-hie-PER-uh-muh-ter]n.
超参数，需手动设定的参数
e.g. Learning rate and batch size are hyperparameters.
3.3 神经网络架构
Dense / Fully Connected[dens / FOO-lee kuh-NEK-tid]adj.
全连接层，每神经元连前层所有神经元
e.g. Dense layer in a feedforward network.
Convolutional[kon-vuh-LOO-shuhn-uhl]adj.
卷积的，用于图像处理
e.g. CNN uses convolutional filters.
Pooling[POO-ling]n.
池化，降维采样操作
e.g. Max pooling selects the strongest signal.
Recurrent[ri-KER-uhnt]adj.
循环的，处理序列数据
e.g. RNN processes text word by word.
Long Short-Term Memory[long short-term MEM-uh-ree]n.
长短期记忆网络 LSTM
e.g. LSTM handles long-range dependencies.
Transformer[trans-FOR-mer]n.
变换器，自注意力机制架构
e.g. GPT and BERT are Transformer-based.
Attention Mechanism[uh-TEN-shuhn MEK-uh-niz-uhm]n.
注意力机制，关注输入的相关部分
e.g. Attention helps models focus on key tokens.
Token[TOH-kuhn]n.
词元，文本处理的最小单位
e.g. A word or subword is a token.
Embedding[em-BED-ing]n.
嵌入，将符号转为向量
e.g. Word2vec produces word embeddings.
Tokenization[toh-kuhn-uh-ZAY-shuhn]n.
分词，将文本拆成词元
e.g. BPE is a popular tokenization method.
Self-Attention[self-uh-TEN-shuhn]n.
自注意力，序列内部关系建模
e.g. Self-attention is the core of Transformer.
3.4 大语言模型 LLM
Large Language Model[larj LAN-gwijh MOD-yuhl]n.
大语言模型，参数量巨大的语言模型
e.g. GPT-4 is a large language model.
Prompt[prompt]n./v.
提示词，给模型的输入指令
e.g. Write a clear prompt for better output.
Prompt Engineering[prompt en-juh-NEER-ing]n.
提示工程，优化提示词的技术
e.g. Few-shot prompting is a technique.
In-Context Learning[in-KON-tekst LER-ning]n.
上下文学习，LLM从提示中学习
e.g. ICL lets models learn without gradient updates.
Fine-tuning[fien TYOO-ning]n.
微调，在特定数据上继续训练
e.g. LoRA enables efficient fine-tuning.
RAG / Retrieval-Augmented Generation[rag]n.
检索增强生成
e.g. RAG combines retrieval and generation.
Hallucination[huh-loo-suh-NAY-shuhn]n.
幻觉，模型生成虚假内容
e.g. RAG reduces hallucination in LLMs.
Chain-of-Thought[chayn uhv THAUT]n.
思维链，引导模型逐步推理
e.g. CoT prompting improves complex reasoning.
Temperature[TEM-pruh-cher]n.
温度，控制输出的随机性
e.g. High temp = creative; low temp = precise.
Top-K / Top-P[top kay / top pee]n.
采样策略，控制输出多样性
e.g. Top-P nucleus sampling balances diversity.
Alignment[uh-LIE-muhnt]n.
对齐，让模型符合人类意图
e.g. RLHF aligns models with human values.
RLHF[ar-el-aitch-eff]n.
人类反馈强化学习
e.g. RLHF is key to ChatGPT's behavior.
Multimodal[mul-tee-MOH-dool]adj.
多模态，处理多种数据类型
e.g. GPT-4V is a multimodal large language model.
Agentic AI[ay-JEN-tik ay-AI]adj.
Agent AI，能自主规划和执行任务
e.g. Agentic AI uses tools and plans multi-step tasks.
Scaling Law[SKAY-ling law]n.
Scaling定律，模型性能随规模提升的规律
e.g. Scaling laws predict larger models perform better.
Emergent Ability[EM-er-juhnt uh-BEE-luh-tee]n.
涌现能力，大模型突然出现的新能力
e.g. Reasoning emerges in large language models.
3.5 计算机视觉 CV
Image Recognition[IM-ij rek-uhg-NI-shuhn]n.
图像识别，识别图像内容
e.g. Image recognition classifies objects in photos.
Object Detection[OB-jekt di-TEK-shuhn]n.
目标检测，定位图像中的物体
e.g. YOLO performs real-time object detection.
Semantic Segmentation[suh-MAN-tik seg-muhn-TAY-shuhn]n.
语义分割，逐像素分类
e.g. Each pixel is labeled by its object class.
Image Generation[IM-ij jen-uh-RAY-shing]n.
图像生成，创建新图像
e.g. DALL-E and Stable Diffusion generate images.
Diffusion Model[di-FYOO-zhuhn MOD-yuhl]n.
扩散模型，生成式图像架构
e.g. Stable Diffusion is a diffusion model.
Latent Space[LAY-zuhnt spays]n.
潜在空间，高维压缩表示
e.g. Generative models sample from latent space.
GAN / Generative Adversarial Network[gan]n.
生成对抗网络
e.g. GANs generate realistic fake images.
PART 4 算能与基础设施
4.1 硬件 Hardware
CPU / Central Processing Unit[sen-truhl PRO-ses-ing YOO-nit]n.
中央处理器，计算机大脑
e.g. CPU handles general-purpose computation.
GPU / Graphics Processing Unit[graf-iks PRO-ses-ing YOO-nit]n.
图形处理器，并行计算芯片
e.g. GPU accelerates deep learning training.
TPU / Tensor Processing Unit[TEN-ser PRO-ses-ing YOO-nit]n.
张量处理器，Google定制AI芯片
e.g. TPU is optimized for tensor operations.
NPU / Neural Processing Unit[noor-uhl PRO-ses-ing YOO-nit]n.
神经网络处理器，终端AI芯片
e.g. NPU enables on-device AI inference.
FPGA[field proh-GRAM-uh-buhl gate uh-RAY]n.
现场可编程门阵列
e.g. FPGA offers hardware-level reprogramming.
ASIC[ap-luh-KAY-shuhn spuh-SIF-ik]n.
专用集成电路，定制的芯片
e.g. ASIC is highly efficient for a specific task.
Memory / RAM[MEM-uh-ree / ram]n.
内存，随机存取存储器
e.g. GPU memory stores model weights.
VRAM[vee-ram]n.
显存，GPU专用内存
e.g. Large models require massive VRAM.
Bandwidth[BAND-width]n.
带宽，数据传输速率
e.g. High bandwidth connects GPU clusters.
Throughput[THRU-hoot]n.
吞吐量，单位时间处理数据量
e.g. Throughput measures GPU efficiency.
TFLOPS / Tera FLOPS[tef-lops]n.
每秒万亿次浮点运算，衡量算力
e.g. A100 GPU delivers 312 TFLOPS.
4.2 云计算与分布式
Cloud Computing[klowd kuhm-PYOO-ting]n.
云计算，按需使用远程资源
e.g. AWS and GCP provide cloud computing.
Data Center[DAY-tuh sen-ter]n.
数据中心，服务器集群设施
e.g. Data centers host GPU clusters.
GPU Cluster[graf-iks PRO-ses-ing YOO-nit kluhs-ter]n.
GPU集群，协同计算的GPU网络
e.g. Training large models requires GPU clusters.
Node[nohd]n.
节点，集群中的单台机器
e.g. Each node has multiple GPUs.
Parallel Computing[PAR-uh-lel kuhm-PYOO-ting]n.
并行计算，多任务同时执行
e.g. Data parallelism is common in training.
Distributed Training[di-STRIB-yoo-tid TRAY-ning]n.
分布式训练，多机器协作训练
e.g. Large models use distributed training.
Data Parallelism[DAY-tuh PAR-uh-lel-iz-uhm]n.
数据并行，每GPU处理不同数据
e.g. Data parallelism speeds up training.
Model Parallelism[MOD-yuhl PAR-uh-lel-iz-uhm]n.
模型并行，模型分片到多GPU
e.g. Massive models need model parallelism.
Heterogeneous Computing[het-er-oh-JEE-nee-uhs kuhm-PYOO-ting]n.
异构计算，混合使用不同芯片
e.g. CPU+GPU+TPU is heterogeneous.
Edge Computing[ej kuhm-PYOO-ting]n.
边缘计算，在数据源附近处理
e.g. Edge AI reduces latency.
Serverless[SER-ver-lis]adj.
无服务器的，按需付费的计算
e.g. Lambda functions are serverless.
Container[kuhn-TAI-ner]n.
容器，轻量级虚拟化
e.g. Docker containers package models + dependencies.
Kubernetes[KYOO-bur-net-is]n.
容器编排系统
e.g. K8s manages containerized AI workloads.
4.3 存储与数据
Data Pipeline[DAY-tuh PIE-pelain]n.
数据流水线，数据处理流程
e.g. ML pipeline automates data prep to train to deploy.
Preprocessing[pree-PROS-es-ing]n.
预处理，原始数据清洗转换
e.g. Normalization preprocesses input features.
Feature Engineering[FEE-cher en-juh-NEER-ing]n.
特征工程，构建有意义的输入特征
e.g. Good features improve model performance.
Data Augmentation[DAY-tuh og-men-TAY-shuhn]n.
数据增强，人为扩充训练数据
e.g. Rotating images augments the training set.
Checkpoint[CHEK-point]n.
检查点，模型训练快照
e.g. Save checkpoints to resume interrupted training.
Weight[wayt]n.
权重，神经网络参数
e.g. Training adjusts weights to minimize loss.
Bias[BY-uhs]n.
偏置，线性层的截距项
e.g. y = wx + b — bias shifts the line.
Cache[kash]n.
缓存，高速临时存储
e.g. L1/L2/L3 cache speeds up CPU access.
4.4 MLOps & Deployment
MLOps[em-el-ops]n.
机器学习运维，模型部署与管理
e.g. MLOps bridges ML development and production.
Model Serving[MOD-yuhl SER-ving]n.
模型服务，将模型暴露为API
e.g. TensorFlow Serving hosts trained models.
Inference Endpoint[IN-fuh-ruhns END-point]n.
推理端点，API访问接口
e.g. Send prompts to the inference endpoint.
Latency[LAY-tuhn-see]n.
延迟，从请求到响应的时间
e.g. Low latency is critical for real-time apps.
Auto-scaling[aw-toh SKAY-ling]n.
自动扩缩容，动态调整资源
e.g. Auto-scaling adjusts GPU count dynamically.
CI/CD[see-eye see-dee]n.
持续集成/持续部署
e.g. CI/CD automates the build-test-deploy pipeline.
Monitoring[MON-uh-ter-ing]n.
监控，跟踪模型性能指标
e.g. Monitor accuracy and latency in production.
A/B Testing[ay bee TES-ting]n.
A/B测试，对比两个模型版本
e.g. A/B test compares model v1 vs v2.
Canary Deployment[kuh-NAIR-ee di-PLOY-muhnt]n.
灰度发布，逐步替换旧版本
e.g. Canary deploys new model to 5% of traffic first.
Drift Detection[drift di-TEK-shuhn]n.
漂移检测，识别数据分布变化
e.g. Concept drift degrades model accuracy.
PART 5 编程语言与开发
5.1 主流编程语言
Python[PY-thon]n.
Python，AI领域最流行语言
e.g. Python has rich ML libraries like PyTorch.
JavaScript[JAY-vuh-skript]n.
JavaScript，Web开发核心语言
e.g. Node.js runs JavaScript on the server.
Java[JAH-vuh]n.
Java，企业级后端开发语言
e.g. Java powers Android and enterprise systems.
C / C++[see / see plus plus]n.
C/C++，高性能系统编程
e.g. CUDA is written in C/C++ for GPU computing.
Rust[ruhst]n.
Rust，系统级安全编程语言
e.g. Rust is used for memory-safe systems.
Go / Golang[goh]n.
Go，高并发服务器语言
e.g. Go powers cloud-native infrastructure.
R[ar]n.
R，统计与数据分析语言
e.g. R is popular among data scientists.
SQL[ess-koo-EL]n.
结构化查询语言，数据库操作
e.g. SELECT * FROM users WHERE age > 18
Julia[JOO-lee-uh]n.
Julia，科学计算高性能语言
e.g. Julia excels at numerical computing.
Swift[swift]n.
Swift，iOS/macOS开发语言
e.g. Swift is used for Apple platform AI apps.
Kotlin[koht-lin]n.
Kotlin，Android首选语言
e.g. Kotlin is Google's preferred Android language.
TypeScript[TIE-p-skript]n.
TypeScript，JavaScript超集
e.g. TypeScript adds static typing to JS.
5.2 开发工具与实践
Git[git]n.
Git，版本控制系统
e.g. Git tracks code changes across a team.
GitHub / GitLab[git-hub / git-lab]n.
代码托管平台
e.g. GitHub hosts open-source AI projects.
Repository[ri-POZ-uh-tor-ee]n.
代码仓库，存储代码的地方
e.g. Push code to the repository daily.
Branch[branch]n.
分支，代码的独立开发线
e.g. Create a new branch for each feature.
Pull Request[pool ri-KWEST]n.
合并请求，代码审查流程
e.g. Open a PR when the feature is ready.
Unit Test[YOO-nit test]n.
单元测试，测试最小代码单元
e.g. Write unit tests for each function.
Integration Test[in-tuh-GRAY-shuhn test]n.
集成测试，多模块协同测试
e.g. Integration tests verify components work together.
Debugging[dee-BUG-ing]n.
调试，定位和修复程序错误
e.g. Print statements help in debugging.
IDE[eye-dee-ee]n.
集成开发环境
e.g. VS Code and PyCharm are popular IDEs.
Notebook[NOOT-buk]n.
笔记本，代码+笔记交互环境
e.g. Jupyter Notebook is standard in data science.
API[ay-pee-eye]n.
应用程序接口，程序间通信约定
e.g. REST API lets clients query your model.
SDK[es-dee-kay]n.
软件开发包，工具集合
e.g. TensorFlow SDK simplifies model deployment.
Framework[FRAIM-wurk]n.
框架，提供开发基础结构
e.g. Django is a Python web framework.
Open Source[OH-puhn sohrs]adj./n.
开源，代码公开可自由使用
e.g. PyTorch is an open-source ML framework.
PART 6 互联网与技术趋势
6.1 网络与协议
HTTP / HyperText Transfer Protocol[aych-tee-tee-pee]n.
超文本传输协议，Web通信基础
e.g. HTTPS adds TLS encryption to HTTP.
REST API[rest ay-pee-eye]n.
REST风格API
e.g. GET /models returns the list of models.
WebSocket[web-SOK-it]n.
WebSocket，双向实时通信
e.g. WebSocket enables real-time chat apps.
JSON[jay-son]n.
JSON，轻量数据交换格式
e.g. {"name":"GPT","version":4} is JSON.
HTML[aych-tee-em-el]n.
超文本标记语言，网页结构
e.g. HTML structures web page content.
CSS[see-es-es]n.
层叠样式表，网页外观控制
e.g. CSS styles the web page layout.
Frontend[FRUNT-end]n.
前端，用户界面的开发
e.g. React and Vue are frontend frameworks.
Backend[BAK-end]n.
后端，服务器端逻辑与数据
e.g. Django and FastAPI are Python backends.
Full-stack[fool stak]adj.
全栈，前端+后端+数据库
e.g. Full-stack developers build entire apps.
Database[DAY-tuh-bays]n.
数据库，结构化数据存储
e.g. PostgreSQL and MySQL are relational DBs.
NoSQL[nohn-sequel]n.
非关系型数据库
e.g. MongoDB and Redis are NoSQL databases.
Cloud-native[klowd NAT-iv]adj.
云原生的，充分利用云的特性
e.g. Kubernetes enables cloud-native apps.
6.2 新兴技术
IoT / Internet of Things[ai-oh-tee]n.
物联网
e.g. IoT connects sensors and devices to the cloud.
5G[fifth jen-uh-RAY-shuhn]n.
第五代移动通信
e.g. 5G enables real-time remote surgery.
Blockchain[BLOK-chayn]n.
区块链，分布式账本技术
e.g. Blockchain ensures data immutability.
VR / Virtual Reality[VER-choo-uhl ri-AL-uh-tee]n.
虚拟现实
e.g. VR headsets create immersive experiences.
AR / Augmented Reality[awg-MEN-tid ri-AL-uh-tee]n.
增强现实
e.g. AR overlays digital info on the real world.
XR / Extended Reality[eks-TEN-did ri-AL-uh-tee]n.
扩展现实，VR+AR+MR统称
e.g. XR spans the full reality-virtuality spectrum.
Quantum Computing[KWON-tuhm kuhm-PYOO-ting]n.
量子计算，利用量子力学计算
e.g. Quantum computers solve certain problems exponentially faster.
Edge AI[ej ay-AI]n.
边缘AI，在设备端运行AI
e.g. Edge AI processes data locally without cloud.
On-device AI[on-di-VIES ay-AI]n.
端侧AI，设备本地AI推理
e.g. Snapdragon NPUs enable on-device AI.
Autonomous Driving[aw-TON-uh-muhs DRY-ving]n.
自动驾驶
e.g. Self-driving cars use AI perception + planning.
Robotics[roh-BOT-iks]n.
机器人学
e.g. Robotics combines AI with mechanical systems.
PART 7 安全与代令
7.1 安全 Security
Encryption[en-KRIP-shuhn]n.
加密，数据保护技术
e.g. AES-256 encryption secures data at rest.
Authentication[aw-then-tuh-FI-kay-shuhn]n.
身份验证，确认用户身份
e.g. OAuth 2.0 handles authentication securely.
Authorization[aw-thor-uh-ZAY-shuhn]n.
授权，控制资源访问权限
e.g. RBAC assigns authorization roles.
Privacy[PRIE-vuh-see]n.
隐私，数据保护权利
e.g. Privacy-preserving ML protects user data.
Cybersecurity[SIE-ber-sih-kyoor-uh-tee]n.
网络安全
e.g. Cybersecurity defends against digital attacks.
Penetration Testing[pen-uh-TRAY-shuhn TES-ting]n.
渗透测试，模拟攻击发现漏洞
e.g. Pen testing reveals security vulnerabilities.
7.2 伦理与趋势 Ethics
Bias in AI[bies in ay-AI]n.
AI偏见，模型中的歧视性偏差
e.g. Bias in AI reflects training data inequalities.
Explainability[ek-splay-nuh-BIH-luh-tee]n.
可解释性，理解AI决策原因
e.g. Explainability is critical for high-stakes AI.
Transparency[trans-per-EN-tuh-see]n.
透明度，AI系统可被理解的程度
e.g. Transparency builds trust in AI systems.
Accountability[uh-kown-tuh-BIH-luh-tee]n.
问责制，明确AI决策的责任归属
e.g. Accountability ensures responsible AI use.
Sustainable AI[suhh-STAY-nuh-buhl ay-AI]n.
可持续AI，降低能耗与资源消耗
e.g. Sustainable AI addresses environmental impact.
Open AI[OH-puhn ay-AI]n.
开源AI，开放权重与研究
e.g. Open AI promotes transparency and collaboration.