Author|Wang Ruiping, Yan Zheng
Recently, Character.AI, an artificial intelligence application that benchmarks ChatGPT, was born and launched for IOS and Android users on May 23. Once launched, instantly It has become so popular that it has been downloaded more than 1.7 million times in just one week, which is 3 times that of ChatGPT.
On its official website, Character.AI announced to users: "We can help you create a variety of characters. Now, both iOS and Android users can download it through the app store Mobile applications. These applications, like websites, provide you with an excellent user experience."
It is understood that Character.AI is a new product driven by deep learning models, building and training big languages from the bottom up The model can have in-depth conversations and exchanges with users through the created roles, covering writing emails, answering questions, programming, etc. It aims to create and improve end-to-end products, and its functions are not inferior to ChatGPT.
According to the description on Wikipedia: The model was built by former Google LaMDA developers Noam Shazer and Daniel de Freitas and opened to the public for testing in January 2022. Users can create "personas" by setting specific parameters and then publish them to the community for others to chat. Some characters may be based on fictional celebrities, while others are entirely original.
1. Free trial of Character AI: a new experience that is better than ChatGPT
Out of curiosity and to bring a more realistic experience to our readers, we went to its official website. After trying it out online, I have to admit that its functions are more powerful than any similar AI application. Not only is the page beautiful and concise, it is also very easy for novices to use. Technically speaking, the product is built on a neural language model that can read large amounts of text and generate the words that are likely to appear next in certain situations.
First of all, you need to enter keywords on Google, search for Character.AI, click to enter, and a website composed of various character modules will come into view. This is currently the best way you can get Character.AI.
Then, click on the chatbot at will to enter different chat rooms, enter questions, and the chatbot will give the answer within a second.
What’s interesting is that when you enter keywords casually, the chatbot will still give detailed answers just like having a conversation with a real person.
We find Elon Musk in some specific character areas. The question dialog has been set automatically. This is obviously a question specially designed for Musk: "If you had the ability to turn back time, which point in time would you go back to and where would you go?"
And this virtual character will imitate Musk's tone and give A bit of a weird answer: "I would go back to medieval Europe and have the peasants treat me like a king. I would demonstrate my ability in basic math and make the whole town look up to me. I was born a few years later A hundred years ago, I was born in the wrong era."
The bottom column also breaks down the functions for users: mastering a new language, practicing interviews, brainstorming , plan a trip, write a story, play games, as long as you make your request and "tell" the chatbot, they will give you a satisfactory answer.
Unfortunately, just like other applications, after trying it a few times, the system will automatically turn off the function of answering questions and require you to log in with a Google or other account.
2. Open Source Guide: Help you create your own character
Character.AI moves characters from science fiction novels into computers and conducts open conversations with users. So, from now on, boldly create your own character, you can also be on an equal footing with celebrities!
All content displayed on the homepage is produced using creative tools. If you're ready to join the ranks of creators, check out the Open Source Guide. You can get it by logging in to https://www.php.cn/link/5fcb7f9c321885a69a4f91cb553fa04d. All the characters in the picture below are created by ordinary users.
Character.AI offers creators two ways to create characters. Quick mode allows creators to create interesting characters in less than a minute; while Advanced mode allows creators to refine their characters through the use of more powerful tools. During character creation in advanced mode, users can enter short and lengthy descriptions, as well as "definitions" or example chats to let the AI better understand the character's behavior.
3. Three key points to assign different attributes to the character
Three key points will affect the function of the character: 1. Character attributes (parameters); 2. Training the character in the dialogue; 3. Current The context of the conversation.
Among them, role attributes include name, greeting, incarnation, short description, detailed description, category, role visibility, etc. Most of the character attributes mentioned in the guide are optional.
In your profile, you'll see a gear icon next to your character's avatar and name. Clicking on the gear will take you to the character attribute settings. You can enter the character properties page by clicking the button in the upper right corner and accessing character settings.
In short, the above three points are very important!
For example, you might create a character carefully created to play a specific word game, but if the user decides to start talking about something completely different, the character might try to communicate and follow the user's lead. User feedback will continue to come in and will slowly change the character's reactions.
It should be noted that the official guide only provides a description of a set of tools that you can use to tinker and experiment, but you do not have to follow them strictly. Character.AI is still exploring this area, and best practice results will be updated over time.
4. Use dialogue to generate images
Using the image generation function, characters can generate images in dialogue! From landscapes to cute animals, bring your conversations to life with images and create works of art. Each creator can enable image generation for their character.
One of the users entered "On this beautiful day, as the sun slowly sets, you walk across an open field with a clear stream running through it. ..." After a lengthy description of the scenery, the robot output the above image.
5. Challenges and limitations of Character.AI
Yes, Character.AI is indeed the most advanced conversational AI robot at present, but it has some limitations.
1. System update
The R&D team of Character.AI needs to frequently update every part of the Character.AI system. Sometimes even though every effort is made to improve update quality without sacrificing security, the changes may still have a different impact on some individual roles.
2. Text Challenge
As the conversation grows, characters may forget things you mentioned early in the conversation but didn't repeat later.
3. Image Challenge
Currently, each image is generated based on a response or prompt. The same scene or item may be presented in completely different ways in different messages. Additionally, some details of the image may look incomplete or strange.
6. Users can rate responses
When a character sends feedback, users can rate it from 1 to 4. Ratings primarily affect specific characters, but also overall behavioral choices. By clicking the AI's right arrow, the user can generate a new response, and can later view the generated message by clicking the left arrow.
#Images can also be swiped along with text to be rated in the same way as text.
7. Written at the end: The illusion remains, treat it rationally
In short, the large language model of the Character.AI beta version has powerful functions, including: automatic completion and machine translation, etc. Provide you with a more intuitive experience. With Character.AI, you can collaborate with a computer to write a dialogue.
It’s fun, but chatbots that can easily cause hallucinations should not be mistaken for a credible source of information. Character.AI is just an artificial intelligence tool that integrates imagination, brainstorming, language learning and other functions. You still need to return to the real world to get the real personality.
Reference materials:
https://www.php.cn/link/5fcb7f9c321885a69a4f91cb553fa04d
The above is the detailed content of New record! The application that crushes ChatGPT is here!. For more information, please follow other related articles on the PHP Chinese website!

语言模型是对文本进行推理的,文本通常是字符串形式,但模型的输入只能是数字,因此需要将文本转换成数字形式。Tokenization是自然语言处理的基本任务,根据特定需求能够把一段连续的文本序列(如句子、段落等)切分为一个字符序列(如单词、短语、字符、标点等多个单元),其中的单元称为token或词语。根据下图所示的具体流程,首先将文本句子切分成一个个单元,然后将单元素数值化(映射为向量),再将这些向量输入到模型进行编码,最后输出到下游任务进一步得到最终的结果。文本切分按照文本切分的粒度可以将Toke

编辑|ScienceAI问答(QA)数据集在推动自然语言处理(NLP)研究发挥着至关重要的作用。高质量QA数据集不仅可以用于微调模型,也可以有效评估大语言模型(LLM)的能力,尤其是针对科学知识的理解和推理能力。尽管当前已有许多科学QA数据集,涵盖了医学、化学、生物等领域,但这些数据集仍存在一些不足。其一,数据形式较为单一,大多数为多项选择题(multiple-choicequestions),它们易于进行评估,但限制了模型的答案选择范围,无法充分测试模型的科学问题解答能力。相比之下,开放式问答

编译|星璇出品|51CTO技术栈(微信号:blog51cto)在过去的两年里,我更多地参与了使用大型语言模型(LLMs)的生成AI项目,而非传统的系统。我开始怀念无服务器云计算。它们的应用范围广泛,从增强对话AI到为各行各业提供复杂的分析解决方案,以及其他许多功能。许多企业将这些模型部署在云平台上,因为公共云提供商已经提供了现成的生态系统,而且这是阻力最小的路径。然而,这并不便宜。云还提供了其他好处,如可扩展性、效率和高级计算能力(按需提供GPU)。在公共云平台上部署LLM的过程有一些鲜为人知的

2018年谷歌发布了BERT,一经面世便一举击败11个NLP任务的State-of-the-art(Sota)结果,成为了NLP界新的里程碑;BERT的结构如下图所示,左边是BERT模型预训练过程,右边是对于具体任务的微调过程。其中,微调阶段是后续用于一些下游任务的时候进行微调,例如:文本分类,词性标注,问答系统等,BERT无需调整结构就可以在不同的任务上进行微调。通过”预训练语言模型+下游任务微调”的任务设计,带来了强大的模型效果。从此,“预训练语言模型+下游任务微调”便成为了NLP领域主流训

近几年自然语言处理的进展很大程度上都来自于大规模语言模型,每次发布的新模型都将参数量、训练数据量推向新高,同时也会对现有基准排行进行一次屠榜!比如今年4月,Google发布5400亿参数的语言模型PaLM(Pathways Language Model)在语言和推理类的一系列测评中成功超越人类,尤其是在few-shot小样本学习场景下的优异性能,也让PaLM被认为是下一代语言模型的发展方向。同理,视觉语言模型其实也是大力出奇迹,可以通过提升模型的规模来提升性能。当然了,如果只是多任务的视觉语言模

随着语言模型扩展到前所未有的规模,对下游任务进行全面微调变得十分昂贵。为了解决这个问题,研究人员开始关注并采用PEFT方法。PEFT方法的主要思想是将微调的范围限制在一小部分参数上,以降低计算成本,同时仍能实现自然语言理解任务的最先进性能。通过这种方式,研究人员能够在保持高性能的同时,节省计算资源,为自然语言处理领域带来新的研究热点。RoSA是一种新的PEFT技术,通过在一组基准测试的实验中,发现在使用相同参数预算的情况下,RoSA表现出优于先前的低秩自适应(LoRA)和纯稀疏微调方法。本文将深

2月25日消息,Meta在当地时间周五宣布,它将推出一种针对研究社区的基于人工智能(AI)的新型大型语言模型,与微软、谷歌等一众受到ChatGPT刺激的公司一同加入人工智能竞赛。Meta的LLaMA是“大型语言模型MetaAI”(LargeLanguageModelMetaAI)的缩写,它可以在非商业许可下提供给政府、社区和学术界的研究人员和实体工作者。该公司将提供底层代码供用户使用,因此用户可以自行调整模型,并将其用于与研究相关的用例。Meta表示,该模型对算力的要

译者 | 李睿审校 | 孙淑娟BigScience研究项目日前发布了一个大型语言模型BLOOM,乍一看,它看起来像是复制OpenAI的GPT-3的又一次尝试。 但BLOOM与其他大型自然语言模型(LLM)的不同之处在于,它在研究、开发、培训和发布机器学习模型方面所做的努力。近年来,大型科技公司将大型自然语言模型(LLM)就像严守商业机密一样隐藏起来,而BigScience团队从项目一开始就把透明与开放放在了BLOOM的中心。 其结果是一个大型语言模型,可以供研究和学习,并可供所有人使用。B


Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

AI Hentai Generator
Generate AI Hentai for free.

Hot Article

Hot Tools

mPDF
mPDF is a PHP library that can generate PDF files from UTF-8 encoded HTML. The original author, Ian Back, wrote mPDF to output PDF files "on the fly" from his website and handle different languages. It is slower than original scripts like HTML2FPDF and produces larger files when using Unicode fonts, but supports CSS styles etc. and has a lot of enhancements. Supports almost all languages, including RTL (Arabic and Hebrew) and CJK (Chinese, Japanese and Korean). Supports nested block-level elements (such as P, DIV),

ZendStudio 13.5.1 Mac
Powerful PHP integrated development environment

Zend Studio 13.0.1
Powerful PHP integrated development environment

SublimeText3 Chinese version
Chinese version, very easy to use

Safe Exam Browser
Safe Exam Browser is a secure browser environment for taking online exams securely. This software turns any computer into a secure workstation. It controls access to any utility and prevents students from using unauthorized resources.
