Basics of AI/Life at MIT (人工智能基础知识/麻省理工学院的生活)
The branch of computer science that aims to make machines intelligent（旨在使机器变得智能的计算机科学分支)
artificial intelligence（人工智能）
Language models, like ChatGPT and Deepseek, that are trained on massive amounts of text data and have many parameters（像 ChatGPT 和 Deepseek 这样在海量文本数据上训练、拥有大量参数的语言模型）
LLMs（大型语言模型）
A specific kind of machine learning task in which a model has to take new datapoints and then sort them into pre-defined groups（一种特定的机器学习任务，模型需要将新的数据点分到预先定义的类别中)
classification（分类）
Another name for a sequence of 2 words（由两个词组成的序列的另一名称）
2-gram（二元语法）
An MIT study showed that using LLMs to help you write essays has ___ effects on your brain（一项 MIT 研究表明，使用大型语言模型帮助写作文对大脑产生 ___ 影响）
negative（负面）
This field aims to help computers learn to sense, see, and recognize things like humans do（该领域旨在帮助计算机像人类一样感知、看见并识别事物）
computer vision（计算机视觉）
A special neural network that is trained on a lot of data, can pay “attention” to every word you give it, and generates text (hint: three letters!)（一种在大量数据上训练、能对输入的每个词“关注”并生成文本的特殊神经网络〔提示：三个字母！〕
GPT（生成式预训练变换模型）
This is what tells your model how wrong it is（这告诉模型它有多“错”）
Loss function（损失函数）
A model that predicts the probabilities of n-grams（预测 n 元语法概率的模型）
probabilistic model（概率模型）
What it’s called when an AI like Deepseek or ChatGPT makes up information（当 Deepseek 或 ChatGPT 之类的 AI 编造信息时称为）
hallucination（幻觉）
Every building, class, and major at MIT has its own （麻省理工学院的每栋建筑、课程和专业都有自己的）
number（编号）
The higher a language model’s ______, the more random the words it uses to respond to you will be（语言模型的 ______ 越高，它用来回应你的词就越随机）
temperature（温度）
The ____ you choose determines how complicated your mapping can be（你选择的 ____ 决定了映射能有多复杂）
model（模型）
Created by researchers at Google, this kind of model can understand long pieces of texts better than previous models by paying ATTENTION to how different words and sentences are related, even if they are far apart（由谷歌研究人员提出，这类模型通过“注意”词句间的关系，即使它们相隔很远，也能比以往模型更好理解长文本）
Transformer Model（变换器模型）
This is a term used by AI researchers and computer scientists to refer to the challenge of keeping biases out of machine learning models and making sure that AI systems align with the goals of their creators（AI 研究人员和计算机科学家用来描述消除机器学习模型偏见并确保 AI 系统与其创造者目标一致的挑战的术语）
the alignment problem（对齐问题）
“Artificial ____ intelligence” is a kind of AI that would be able to not only do many things but also do them better than humans; some people think it won’t ever exist, others think it just doesn't exist yet（“人工____智能”是一种可以执行多种任务并且比人类做得更好的 AI；有人认为它永远不会出现，也有人认为它只是尚未出现)
general（通用）
An online IDE that uses AI to help you create apps and games（一个利用 AI 帮助你创建应用和游戏的在线 IDE）
Replit（Replit 平台）
These are the specific numbers your model uses for multiplication and addition（模型在乘法和加法中使用的具体数值)
parameters（参数）
After tokenization splits text data up into words (or tokens), _______ assigns each word a vector value that allows a model to understand how the different words relate to one another（在分词把文本拆成词〔或 token〕后，_______ 会给每个词分配一个向量值，使模型能够理解词与词之间的关系）
embedding（嵌入表示）
Kai-Fu Lee predicts that AI will take __% of jobs in the next 15 years（李开复预测在未来 15 年内 AI 将取代 __% 的工作）
50
In your first semester at MIT, you don’t get letter grades (A, B, C, etc); instead, you have a system called（在麻省理工学院的第一学期，你不会得到字母成绩 A、B、C 等，而是使用一种称为
Pass/No Record (P/NR)（通过/不记录）
IDE stands for…（IDE 的全称是……)
Integrated Development Environment（集成开发环境）
These are the three parts of a learning algorithm（这是一个学习算法的三部分）
Loss function, optimizer (usually gradient descent), updater (usually backpropagation)（损失函数、优化器〔通常为梯度下降〕、更新器〔通常为反向传播〕）
This is a fancy name for the task of teaching computers to tell what feelings different sentences have（一种高级说法，用来描述教计算机判断句子情感的任务）
sentiment analysis（情感分析）
When faulty data causes AI to identify false patterns and make bad decisions, the AI is said to have ____（当错误的数据导致 AI 识别出虚假模式并做出错误决策时，称该 AI 具有 ____）
Bias（偏见）