作者：Tushar Gupta；Vaibhav Dua
ChatGPT 是 OpenAI 开发的一种语言模型，可用于自然语言处理任务，例如文本生成和语言翻译。 与其前身 InstructGPT 相比，ChatGPT 试图减少对用户的有害和欺骗性响应。 ChatGPT 基于 GPT-3.5（Generative Pretrained Transformer 3.5，生成式预训练转换器3.5）系列模型，该模型被认为是市场上最先进的语言模型之一，拥有 1750 亿个参数。 GPT-3.5 使用 GPT-3 作为基础模型，使用从互联网收集的 570GB 数据集进行训练。
一、ChatGPT 的特点：
ChatGPT 具有以下关键特性和功能，使其成为一种强大且有用的语言模型：
类人响应：ChatGPT 经过训练可以生成与人类在特定情况下如何响应类似的响应。这使它能够与用户进行自然的、类似人类的对话。 上下文感知：ChatGPT 能够维护上下文并跟踪对话的流程，即使在复杂或多轮对话中也能提供适当的响应。 大训练数据：ChatGPT 在大量文本数据上进行训练，这使其能够学习广泛的语言模式和风格。 处理不当请求：ChatGPT 被设计成拒绝不当请求，以防止其被滥用。二、ChatGPT 的应用
它可以解释复杂的主题：当使用 ChatGPT 为复杂主题寻找定义时，可以大大缩短研究时间。
它可以用来写博客文章：ChatGPT 可以创建信息丰富的博客文章，只需要一些结构上的改变就可以让它变得更好。
它可用于生成视频脚本：ChatGPT 是生成视频脚本的完美工具，否则创建视频脚本可能会花费大量时间。
它可用于创建产品描述：ChatGPT 可以创建产品描述，这对于需要为每个产品创建插图的电子商务网很有帮助。 ChatGPT 可以帮助他们节省金钱和时间。
三、ChatGPT的局限性
尽管有庞大的训练数据集、大量参数和其他技术特性，但与其他所有技术一样，ChatGPT 也有其局限性。 ChatGPT 的一些局限性包括：
• 它是一种通用目的的聊天机器人，在许多情况下可能不如专用定制聊天机器人准确和可靠。
• 有时，ChatGPT 可能会提供不正确或错误的答案，但这些答案由于模型的训练方式听起来可能是正确的。 在训练期间，没有真实来源的训练数据。
• 模型的响应取决于输入查询的措辞方式。 输入相同但一些措辞经过改写的查询可能会导致不同的答案或没有答案。
• 有时，不法分子可能会绕过该模型强加的内容审核，以收集非法和反社会信息。
四、OpenAI 的简史、投资和估值
ChatGPT 的开发商 OpenAI 由 Sam Altman、Elon Musk 等人于 2015 年创立。 OpenAI 收到了 10 亿美元的初始资金。一些最初的捐助者包括 AWS、Infosys、YC Research 和 Peter Thiel 等。2018 年，Elon Musk 辞去了 OpenAI 董事会的职务，以避免未来与特斯拉的 AI 发生任何冲突。2019 年，微软向 OpenAI 投资 10 亿美元，并同意共同开发微软的 Azure AI 超级计算技术。据报道微软正计划进一步投资，希望 OpenAI 能够将其 AI 技术与微软其他产品相结合。微软还计划将 OpenAI 的 ChatGPT 聊天机器人整合到其 Bing 搜索引擎中。ChatGPT 整合将允许 Bing 提供完整的句子来响应搜索查询而不是链接列表。它可以让 Microsoft Bing 成为游戏规则改变者，并使 Microsoft 能够在搜索引擎市场上与 Google 竞争。
2023年1 月的第一周，ChatGPT 开发商 OpenAI 正在考虑一项对该公司估值 290 亿美元的收购要约，使其成为美国最有价值的初创公司之一。OpenAI 现在正在考虑将 ChatGPT 货币化的选项，并宣布可能很快推出名为 ChatGPT Professional 的实验性付费版本。
还有其他几家公司也一直在开发类似的平台，最著名的是 Alphabet 和 Meta。 然而，与 OpenAI 不同的是，Alphabet 和 Meta 出于一些原因一直在犹豫是否要公开其 AI 模型。 尽管人们意识到 AI 的潜力，但ChatGPT 展示了 AI 引擎的当前能力，并在聊天机器人领域引起了轰动。 虽然 ChatGPT 在当前的炒作中高歌猛进，但从长远来看，ChatGPT 是否能够遏制其不道德的使用并实现盈利还有待观察。
【免责声明】本文没有明示或暗示保证内容的准确性、可靠性或完整性。本文仅供一般参考信息之用，不构成任何法律建议。如果要寻求任何具体的法律建议，请进一步联系作者或其他专业人员就具体问题进行咨询。
作者简介：
Tushar Gupta和Vaibhav Dua 是印度Intricateresearch公司的创始人。印度Intricate Research公司是一家知识产权（例如专利）检索、分析和研究的公司，该公司和文章作者的更多信息可通过其官方网站(www.intricateresearch.com)了解。
以下是英文原文：
ChatGPT: A Conversational AI model
Tushar Gupta；Vaibhav Dua
ChatGPT is a language model developed by OpenAI that can be used for natural language processing tasks, such as, text generation and language translation. In comparison to its predecessor, InstructGPT, ChatGPT attempts to reduce harmful and deceitful responses to the users. ChatGPT is based on the GPT-3.5 (Generative Pretrained Transformer 3.5) series model, which is considered as one of the most advanced language models available in the market with 175 billion parameters. GPT-3.5 uses GPT-3 as the base model which was trained using 570GB of dataset gathered from the internet.
Features of ChatGPT:
ChatGPT has below-mentioned key features and capabilities that make it a powerful and useful language model:
Human-like responses: ChatGPT has been trained to generate responses that are similar to how a human would respond in a given situation. This allows it to engage in natural, human-like conversations with users.
Contextual awareness: ChatGPT is able to maintain context and track the flow of a conversation, allowing it to provide appropriate responses even in complex or multi-turn conversations.
Large training data: ChatGPT has been trained on a large amount of text data, which has allowed it to learn a wide range of language patterns and styles.
Handle inappropriate requests: ChatGPT has been designed to refuse inappropriate requests with an intention to prevent its misuse.
Applications of ChatGPT
It can explain complex subjects: One can cut its research time by a substantial amount while using ChatGPT to find definitions for complex subjects.
It can be used to write blog posts: ChatGPT can create informative blog post which only needs some structural changes to make it better.
It can be used to generate video scripts: ChatGPT is a perfect tool for generating video scripts, which can otherwise take a lot of time to create.
It can be used to create product descriptions: ChatGPT can create product descriptions which is helpful for e-commerce websites that need to create illustrations for every product. ChatGPT can help them save money and time.
Despite the large training dataset, enormous number of parameters, and other technical features, just like every other technology, ChatGPT has its limitations. Some of the limitations of ChatGPT are:
It is a general-purpose chatbot which, in many cases, may not be as accurate and authentic as a purpose-built chatbot. Sometimes, ChatGPT may provide incorrect or false answers which may sound correct due the way the model has been trained. During training, there is no source of truth in the training data.The response by the model depends on the way input query is phrased. Inputting the same query with some rephrasing may result in a different answer or no answer.The content moderation imposed by the model may, sometimes, be bypassed by miscreants to gather unlawful and anti-social information.OpenAI Brief History, Investments, and Valuation
ChatGPT’s developer OpenAI was founded in 2015 by Sam Altman, Elon Musk, and others. OpenAI received an initial funding of USD 1 billion as donation. Some of the initial donors included AWS, Infosys, YC Research, and Peter Thiel among others. In 2018, Elon Musk resigned from the OpenAI board to avoid any future conflict with Tesla’s AI efforts. In 2019, Microsoft invested USD 1 billion in OpenAI with an agreement to jointly develop Microsoft’s Azure AI supercomputing technologies. Now, it has been reported that Microsoft is planning to add further investment with a vision that OpenAI can work on integrating its AI technology with other Microsoft products. Microsoft also has plans to incorporate OpenAI’s ChatGPT chatbot into its Bing search engine. ChatGPT integration will allow Bing to provide complete sentences in response to search queries rather than lists of links. It can be a gamechanger for Microsoft Bing and enable Microsoft to compete with Google in search engine market.
In first week of January, ChatGPT developer OpenAI was considering a tender offer valuing the company at USD 29bn, making it one of the most valuable US startups. OpenAI is now considering options to monetize ChatGPT and announced that it may soon launch an experimental paid version named ChatGPT Professional.
There are several other companies, most notably Alphabet and Meta, that have been working on similar platform. However, unlike OpenAI, Alphabet and Meta have been hesitant on making their AI models publicly available citing some reasons. While people were aware of AI’s potential, ChatGPT has shown the current capabilities of an AI engine and created a buzz around chatbots. While ChatGPT is riding high on the current hype, it is to be seen whether ChatGPT is able to contain its unethical usage and become profitable in a long term.
[Disclaimer] This article does not expressly or implicitly guarantee the accuracy, reliability or completeness of the content. This article is for general informational purposes only and does not constitute legal advice. For any specific legal advice, please further contact the author or translator for legal advice.
暂无评论内容