AI术语表（由a16z提供）
Andreessen Horowitz（简称为a16z），它是一家知名的风险投资公司，成立于2009年，总部位于美国加利福尼亚州的门洛帕克。目前投资了81家AI相关公司，覆盖基础设施/工具、模型提供商、横向应用、纵向应用、自主性、生物科学、医疗保健等多个领域。最近推出的AI+a16z系列文章包括对AI监管、行业应用和未来发展趋势的思考。本篇AI术语表和AI经典作品集（原文链接见文末）适合对AI有兴趣，但没有深厚技术背景的读者作为入门学习材料。
· Accelerator加速器
A class of microprocessors designed to accelerate AI applications.
一种专为加速人工智能应用程序而设计的微处理器类别。
· Agents代理人
Software that can perform certain tasks independently and proactively without the need for human intervention, often utilizing a suite of tools like calculators or web browsing.
代理人是指能够独立主动执行特定任务的软件，无需人类干预，通常利用计算器或网络浏览器等工具套件。
· AGI (Artificial General Intelligence)人工通用智能
Though not widely agreed upon, Microsoft researchers have defined AGI as artificial intelligence that is as capable as a human at any intellectual task.
尽管没有得到广泛的共识，但微软的研究人员将AGI定义为在任何智力任务上与人类一样有能力的人工智能。
· Alignment对齐性
The task of ensuring that the goals of an AI system are in line with human values.
确保人工智能系统的目标与人类价值观一致的任务。
· ASI (Artificial Super Intelligence)人工超级智能
Though subject to debate, ASI is commonly defined as artificial intelligence that surpasses the capabilities of the human mind.
尽管存在争议，但ASI通常被定义为超越人类思维能力的人工智能。
· Attention注意力
In the context of neural networks, attention mechanisms help the model focus on relevant parts of the input when producing an output.
在神经网络的背景下，注意力机制帮助模型在生成输出时集中关注输入的相关部分。
· Back Propagation反向传播
An algorithm often used in training neural networks, refers to the method for computing the gradient of the loss function with respect to the weights in the network.
反向传播是一种经常用于训练神经网络的算法，它指的是计算损失函数对网络中权重的梯度的方法。
· Bias偏见
Assumptions made by an AI model about the data. A “bias-variance tradeoff” is the balance that must be achieved between assumptions a model makes about the data and the amount a model’s predictions change, given different training data. Inductive bias is the set of assumptions that a machine learning algorithm makes on the underlying distribution of the data.
AI 模型对数据所做的假设。"偏差方差权衡"是在模型对数据做出的假设和模型的预测在不同的训练数据下发生变化之间必须实现的平衡。归纳偏见是机器学习算法对数据的基础分布所做的一组假设。
· Chain of Thought思维链
In AI, this term is often used to describe an AI model's reasoning steps to arrive at a decision.
在人工智能领域，这个术语通常用来描述AI模型用于得出决策的推理步骤的顺序。
· Chatbot聊天机器人
A computer program designed to simulate human conversation through text or voice interactions. Chatbots often utilize natural language processing techniques to understand user input and provide relevant responses.
聊天机器人是一种通过文本或语音交互模拟人类对话的计算机程序。聊天机器人通常利用自然语言处理技术来理解用户的输入并提供相关的回应。
· ChatGPT
A large-scale AI language model developed by OpenAI generates human-like text.
由OpenAI开发的大规模AI语言模型，可以生成类似人类的文本。
· CLIP (Contrastive Language–Image Pretraining)对比式语言-图像预训练
An AI model developed by OpenAI that connects images and text, allowing it to understand and generate descriptions of images.
CLIP是OpenAI开发的人工智能模型，它将图像和文本连接起来，使其能够理解并生成图像的描述。
· Compute计算
The computational resources (like CPU or GPU time) used in training or running AI models.
训练或运行人工智能模型所使用的计算资源（如CPU或GPU时间）。
· Convolutional Neural Network (CNN)卷积神经网络
A type of deep learning model that processes data with a grid-like topology (e.g., an image) by applying a series of filters. Such models are often used for image recognition tasks.
一种深度学习模型，通过应用一系列滤波器处理具有网格状拓扑结构（例如图像）的数据。这种模型通常用于图像识别任务。
· Data Augmentation数据增强
The process of increasing the amount and diversity of data used for training a model by adding slightly modified copies of existing data.
通过添加现有数据的略微修改的副本，增加用于训练模型的数据量和多样性的过程。
· Deep Learning深度学习
A subfield of machine learning that focuses on training neural networks with many layers, enabling the learning of complex patterns.
机器学习的一个子领域，专注于训练具有多层的神经网络，使其能够学习复杂的模式。
· Diffusion扩散
In AI and machine learning, a technique is used for generating new data by starting with a piece of real data and adding random noise. A diffusion model is a type of generative model in which a neural network is trained to predict the reverse process when random noise is added to data. Diffusion models are used to generate new samples of data that are similar to the training data.
在人工智能和机器学习中，一种通过从一片真实数据开始并添加随机噪声来生成新数据的技术。扩散模型是一种生成模型，其中神经网络被训练以预测在向数据添加随机噪声时的反向过程。扩散模型被用来生成与训练数据相似的新数据样本。
· Double Descent双重下降
A phenomenon in machine learning in which model performance improves with increased complexity then worsens, then improves again.
机器学习中的一种现象，在该现象中，模型性能随着复杂性的增加而改善，然后恶化，然后再次改善。
· Embedding嵌入
The representation of data in a new form, often a vector space. Similar data points have more similar embeddings.
数据在新形式中的表示，通常是一个向量空间。相似的数据点具有更相似的嵌入。
· Emergence/Emergent Behavior (“sharp left turns,” intelligence explosions)紧急性/突发性行为（“急转弯”，智能爆炸）
In AI, emergence refers to complex behavior arising from simple rules or interactions. “Sharp left turns” and “intelligence explosions” are speculative scenarios where AI development takes sudden and drastic shifts, often associated with the arrival of AGI.
在人工智能中，紧急性指的是由简单规则或交互产生的复杂行为。“急转弯”和“智能爆炸”是人工智能发展出现突然和剧烈变化的推测性场景，通常与AGI的到来有关。
· End-to-End Learning端到端学习
A type of machine learning model that does not require hand-engineered features. The model is simply fed raw data and expected to learn from these inputs.
一种不需要手工工程特征的机器学习模型。该模型只需输入原始数据，并期望从这些输入中学习。
· Expert Systems专家系统
An application of artificial intelligence technologies that provides solutions to complex problems within a specific domain.
一种应用人工智能技术的应用，为特定领域内的复杂问题提供解决方案。
· Explainable AI (XAI)可解释的人工智能
A subfield of AI focused on creating transparent models that provide clear and understandable explanations of their decisions.
一个专注于创建透明模型的人工智能子领域，这些模型提供了对其决策的清晰且易于理解的解释。
· Fine-tuning微调
The process of taking a pre-trained machine learning model that has already been trained on a large dataset and adapting it for a slightly different task or a specific domain. During fine-tuning, the model’s parameters are further adjusted using a smaller, task-specific dataset, allowing it to learn task-specific patterns and improve performance on the new task.
将已经在大型数据集上进行过训练的预训练机器学习模型的过程，并将其适应于略有不同的任务或特定领域。在微调期间，模型的参数会使用较小的、任务特定的数据集进行进一步调整，使其能够学习任务特定的模式并改善在新任务上的性能。
· Forward Propagation正向传播
In a neural network, forward propagation is the process where input data is fed into the network and passed through each layer (from the input layer to the hidden layers and finally to the output layer) to produce the output. The network applies weights and biases to the inputs and uses activation functions to generate the final output.
在神经网络中，正向传播是将输入数据输入到网络中并通过每一层（从输入层到隐藏层，最后到输出层）以产生输出的过程。网络对输入应用权重和偏差，并使用激活函数生成最终输出。
· Foundation Model基础模型
Large AI models trained on broad data, are meant to be adapted for specific tasks.
在广泛数据上训练的大型AI模型，意在用于适应特定任务。
· General Adversarial Network (GAN)对抗生成网络
A type of machine learning model used to generate new data similar to some existing data. It pits two neural networks against each other: a “generator,” which creates new data, and a “discriminator” which tries to distinguish that data from real data.
一种用于生成与一些现有数据相似的新数据的机器学习模型。它将两个神经网络对立起来：一个“生成器”，它创建新数据，和一个“鉴别器”，试图将这些数据与真实数据区分开来。
· Generative AI生成性AI
A branch of AI focused on creating models that can generate new and original content, such as images, music, or text, based on patterns and examples from existing data.
AI的一个分支，专注于创建能够基于现有数据的模式和示例生成新的和原始内容（如图像、音乐或文本）的模型。
· GPT (Generative Pretrained Transformer)生成预训练转换器
A large-scale AI language model developed by OpenAI generates human-like text.
由OpenAI开发的大规模AI语言模型，可生成类人的文本。
· GPU (Graphics Processing Unit)图形处理单元
A specialized type of microprocessor primarily designed to quickly render images for output to a display. GPUs are also highly efficient at performing the calculations needed to train and run neural networks.
一种专门设计用于快速呈现图像以输出到显示器的微处理器类型。在进行训练和运行神经网络所需的计算方面，GPU也非常高效。
· Gradient Descent梯度下降
In machine learning, gradient descent is an optimization method that gradually adjusts a model’s parameters based on the direction of the largest improvement in its loss function. In linear regression, for example, gradient descent helps find the best-fit line by repeatedly refining the line’s slope and intercept to minimize prediction errors.
在机器学习中，梯度下降是一种优化方法，根据其损失函数中最大改进的方向逐渐调整模型的参数。例如，在线性回归中，梯度下降通过反复调整线的斜率和截距以最小化预测错误，从而帮助找到最佳拟合线。
· Hallucinate/Hallucination幻觉
In the context of AI, hallucination refers to the phenomenon in which a model generates content that is not based on actual data or is significantly different from reality.
在AI的背景下，幻觉指的是模型生成的内容不基于实际数据或与现实大相径庭的现象。
· Hidden Layer隐藏层
Layers of artificial neurons in a neural network that are not directly connected to the input or output.
神经网络中的人工神经元层，这些层不直接连接到输入或输出。
· Hyperparameter Tuning超参数调整
The process of selecting the appropriate values for the hyperparameters (parameters that are not learned from the data) of a machine learning model.
选择机器学习模型的超参数（不从数据中学习的参数）的适当值的过程。
· Inference推断
The process of making predictions with a trained machine learning model.
使用经过训练的机器学习模型进行预测的过程。
· Instruction Tuning指令调优
A technique in machine learning where models are fine-tuned based on specific instructions given in the dataset.
一种机器学习技术，其中模型基于数据集中给出的特定指令进行微调。
· Large Language Model (LLM)大型语言模型
A type of AI model that can generate human-like text and is trained on a broad dataset.
一种AI模型，可以生成类人的文本，并在广泛的数据集上进行训练。
· Latent Space潜在空间
In machine learning, this term refers to the compressed representation of data that a model (like a neural network) creates. Similar data points are closer in latent space.
在机器学习中，这个术语指的是模型（如神经网络）创建的数据的压缩表示。相似的数据点在潜在空间中更接近。
· Loss Function (or Cost Function)损失函数（或成本函数）
A function that a machine learning model seeks to minimize during training. It quantifies how far the model’s predictions are from the true values.
机器学习模型在训练过程中寻求最小化的函数。它量化了模型预测与真实值的偏差有多大。
· Machine Learning机器学习
A type of artificial intelligence that provides systems the ability to automatically learn and improve from experience without being explicitly programmed.
一种人工智能，为系统提供了自动学习并从经验中改进的能力，而无需明确编程。
· Mixture of Experts专家混合
A machine learning technique where several specialized submodels (the “experts”) are trained and their predictions are combined in a way that depends on the input.
一种机器学习技术，其中训练了几个专业化的子模型（“专家”），并以取决于输入的方式结合它们的预测。
· Multimodal多模态
In AI, this refers to models that can understand and generate information across several types of data, such as text and images.
在AI中，这指的是可以理解和生成跨多种数据类型（如文本和图像）的信息的模型。
· Natural Language Processing (NLP)自然语言处理
A subfield of AI focused on the interaction between computers and humans through natural language. The ultimate objective of NLP is to read, decipher, understand, and make sense of human language in a valuable way.
AI的一个子领域，专注于通过自然语言进行计算机和人类之间的交互。NLP的最终目标是阅读、解读、理解并以有价值的方式理解人类语言。
· NeRF (Neural Radiance Fields)神经辐射场
A method for creating a 3D scene from 2D images using a neural network. It can be used for photorealistic rendering, view synthesis, and more.•
一种使用神经网络从2D图像创建3D场景的方法。它可用于真实感渲染，视图合成等。
· Neural Network神经网络
A type of AI model inspired by the human brain. It consists of connected units or nodes—called neurons—that are organized in layers. A neuron takes inputs, does some computation on them, and produces an output.
一种受人脑启发的AI模型。它由组织在层中的连接单元或节点（称为神经元）组成。神经元接收输入，对它们进行一些计算，并产生输出。
· Objective Function目标函数
A function that a machine learning model seeks to maximize or minimize during training.
机器学习模型在训练过程中寻求最大化或最小化的函数。
· Overfitting过度拟合
A modeling error occurs when a function is too closely fit to a limited set of data points, resulting in poor predictive performance when applied to unseen data.
当函数过于接近一组有限的数据点时发生的建模错误，导致在应用于未见数据时预测性能较差。
· Parameters参数
In machine learning, parameters are the internal variables that the model uses to make predictions. They are learned from the training data during the training process. For example, in a neural network, the weights and biases are parameters.
在机器学习中，参数是模型用于进行预测的内部变量。它们在训练过程中从训练数据中学习。例如，在神经网络中，权重和偏置就是参数。
· Pre-training预训练
The initial phase of training is a machine learning model where the model learns general features, patterns, and representations from the data without specific knowledge of the task it will later be applied to. This unsupervised or semi-supervised learning process enables the model to develop a foundational understanding of the underlying data distribution and extract meaningful features that can be leveraged for subsequent fine-tuning on specific tasks.
训练机器学习模型的初始阶段，模型在此阶段从数据中学习通用特征、模式和表示，而不需要了解它稍后将应用于的任务的具体知识。这个无监督或半监督的学习过程使模型能够开发对底层数据分布的基础理解，并提取可以用于后续对特定任务进行微调的有意义的特征。
· Prompt提示
The initial context or instruction that sets the task or query for the model.
为模型设置任务或查询的初始上下文或指令。
· Regularization正则化
In machine learning, regularization is a technique used to prevent overfitting by adding a penalty term to the model’s loss function. This penalty discourages the model from excessively relying on complex patterns in the training data, promoting more generalizable and less prone-to-overfitting models.
在机器学习中，正则化是一种通过在模型的损失函数中添加惩罚项以防止过拟合的技术。这种惩罚阻止模型过度依赖训练数据中的复杂模式，促进更具泛化能力且不易过拟合的模型。
· Reinforcement Learning强化学习
A type of machine learning where an agent learns to make decisions by taking actions in an environment to maximize some reward.
一种机器学习，其中代理通过在环境中采取行动来最大化某种奖励，从而学习做出决策。
· RLHF (Reinforcement Learning from Human Feedback)人类反馈的强化学习
A method to train an AI model by learning from feedback given by humans on model outputs.
一种通过学习人类对模型输出的反馈来训练AI模型的方法。
· Singularity奇点
In the context of AI, the singularity (also known as the technological singularity) refers to a hypothetical future point in time when technological growth becomes uncontrollable and irreversible, leading to unforeseeable changes to human civilization.
在AI的上下文中，奇点（也被称为技术奇点）指的是一个假设的未来时间点，那时技术增长将变得无法控制和不可逆转，导致对人类文明的无法预见的变化。
· Supervised Learning监督学习
A type of machine learning where the model is provided with labeled training data.
一种机器学习，模型提供标注的训练数据。
· Symbolic Artificial Intelligence符号人工智能
A type of AI that utilizes symbolic reasoning to solve problems and represent knowledge.
一种利用符号推理来解决问题和表示知识的AI。
· TensorFlow
An open-source machine learning platform developed by Google that is used to build and train machine learning models
一个由Google开发的开源机器学习平台，用于构建和训练机器学习模型。
· TPU (Tensor Processing Unit)张量处理单元
A type of microprocessor developed by Google specifically for accelerating machine learning workloads.
Google专门为加速机器学习工作负载而开发的一种微处理器。
· Training Data训练数据
The dataset is used to train a machine-learning model.
用于训练机器学习模型的数据集。
· Transfer Learning迁移学习
A method in machine learning where a pre-trained model is used on a new problem.
机器学习中的一种方法，其中预训练模型被用于新问题。
· Transformer
A specific type of neural network architecture is used primarily for processing sequential data such as natural language. Transformers are known for their ability to handle long-range dependencies in data, thanks to a mechanism called “attention,” which allows the model to weigh the importance of different inputs when producing an output.
Transformer 是一种特殊类型的神经网络架构，主要用于处理自然语言等顺序数据。由于拥有一种名为“注意力”的机制，Transformer 能处理数据的长范围依赖性，该机制允许模型在生成输出时权衡不同输入的重要性。
· Underfitting欠拟合
A modeling error in statistics and machine learning when a statistical model or machine learning algorithm cannot adequately capture the underlying structure of the data.
在统计和机器学习中，当统计模型或机器学习算法不能充分捕获数据的基础结构时发生的建模错误。
· Unsupervised Learning无监督学习
A type of machine learning where the model is not provided with labeled training data, and instead must identify patterns in the data on its own.
一种机器学习类型，其中模型未提供标签的训练数据，而必须自行识别数据中的模式。
· Validation Data验证数据
A subset of the dataset used in machine learning that is separate from the training and test datasets. It’s used to tune the hyperparameters (i.e., architecture, not weights) of a model.
机器学习中使用的数据集的一个子集，与训练和测试数据集分开。它用于调整模型的超参数（即，架构，而非权重）。
· XAI (Explainable AI)可解释的 AI
A subfield of AI focused on creating transparent models that provide clear and understandable explanations of their decisions.
一个关注创建透明模型的 AI 子领域，这些模型提供其决策的清晰和易于理解的解释。
· Zero-shot Learning零样本学习
A type of machine learning where the model makes predictions for conditions not seen during training, without any fine-tuning.
一种机器学习类型，其中模型在未进行任何微调的情况下，对训练期间未见过的条件进行预测。
原文链接
AI术语表：https://a16z.com/ai-glossary/
AI经典作品集：https://a16z.com/2023/05/25/ai-canon/
更多阅读：
如果您也在从事AI相关工作，欢迎扫码添加小编微信（pkcapital2023），和一群对创业和投资有热情的小伙伴一起交流