Collection of generative artificial intelligence models/services
Haebom
10
1
This list is completely based on my personal taste. I have personally used it and collected things that I found meaningful. If there is anything I have missed or explained incorrectly, please let me know. The order of the list has no meaning and I just wrote it down as I thought of it. I have added a thumbs up emoji to the recommended services.
( +8/3 added If you click on each service name, you will see a link that takes you directly to that service. If it is well written in Korean, a Taegeukgi flag is attached. - Services or domestic companies that officially support Korean)
( +8/17 I was asked to pick out the services I personally use, so I've gathered the recommended services at the top. It's mainly services that I use daily or frequently.)
Shares are welcome, tips are welcome, comments are welcome, subscriptions are welcome, feedback is welcome.
Typeset.io is an AI chatbot service that can talk to research papers and paper PDFs. This tool helps you understand research papers better and explains complex academic texts in simple language. Users can highlight text, math, and tables in PDFs, get summaries and explanations for confusing parts, and dig deeper into deep topics.
Captions is an AI-powered creative studio that lets users create studio-quality videos with just a few taps. With AI script writer and AI voice dubbing, you can transcribe videos recorded in, say, Korean into English, French, Italian, etc.
DeepL is an instant translation service for text and document files, providing accurate translations for individuals and teams. DeepL supports 31 languages and can translate PDF, DOCX, and PPTX file formats. It also has an AI-powered editing feature called DeepL Write that corrects grammar and punctuation errors, restructures sentences, and expresses the nuances of speech.
Speak is an English speaking practice service that allows you to practice real conversations without a real tutor. Based on GPT, you can set up situations and roles and have conversations with an AI tutor at any time and on any topic you want.
ChatGPT is a service that provides instant answers, creative inspiration, and learning new things. chatGPT is free to use and available on iOS and Android. It provides users with instant answers, creative inspiration, and the opportunity to learn new things.
DALL-E 3 understands much more nuanced and detailed than previous systems. This allows you to easily translate your ideas into highly accurate images. DALL-E 3 is currently in research preview and will be available via API and Labs to ChatGPT Plus and Enterprise customers in October 23.
Grammarly is an AI-powered writing assistant that helps users instantly produce clear and persuasive writing while maintaining their unique voice. It comprehensively corrects grammar, spelling, tone, and more to enable effective communication.
GPT-3 is a large-scale language model developed by OpenAI. It was released in June 2020 and has 175 billion parameters. This is more than 100 times that of GPT-2. GPT-3 can perform various language tasks and can generate text that is difficult to distinguish from human writing. Recently, it is said that an updated version of GPT-3, GPT-4, is under development.
🇰🇷 👍 GPT-4 is a large-scale language model developed by OpenAI, and is the successor to GPT-3. It was released in March 2023 and has 17.5 billion parameters. This is more than 10 times that of GPT-3. GPT-4 shows improved performance than GPT-3, can generate longer texts, and can perform more complex tasks. In addition, GPT-4 is a multimodal model and can process images and speech.
LaMDA (Language Model for Dialogue Applications) is a large-scale language model developed by Google AI. It was released in January 2021 and has 137B parameters, which is one-tenth of GPT-3. LaMDA can perform various language tasks and generate text that is difficult to distinguish from human writing. In addition, LaMDA can understand and express high-level human characteristics such as consciousness, emotion, and creativity.
Google DeepMind's Gopher is a large-scale language model with 288 billion parameters. Released in March 2023, it is larger than OpenAI's GPT-3 (175 billion) and smaller than Microsoft's Megatron-Turing (530 billion), but DeepMind explains that it outperforms existing models, and aims to build a slightly more specialized and secure language model.
LLaMA is a large-scale language model developed by Meta AI. It was released in February 2023 and has 6.5 billion parameters, which is a quarter of GPT-3. LLaMA can perform various language tasks and generate text that is difficult to distinguish from human-written text. In addition, LLaMA is a multimodal model and can process images and speech.
👍 LLaMA 2 is the second version of a large-scale language model developed by Meta AI. It was released in July 2023 and has 7 billion parameters, which is more than 4 times that of LLaMA. LLaMA 2 shows improved performance than LLaMA, can generate longer texts, and can perform more complex tasks. In addition, LLaMA 2 is a multimodal model, which can process images and speech.
🇰🇷 The Llama-2-70b-instruct-v2 model was developed by Upstage and uses the backbone model of LLaMA-2. This model performs text generation in English and uses the HuggingFace Transformers library. Unlike LLaMA 2, which has little Korean data, this model is significant in that it is specialized for Korean.
Claude 2 is a large-scale language model developed by Anthropic AI. It was released in May 2023 and has 10 billion parameters, which is one-sixth of GPT-3. Claude 2 can perform a variety of language tasks and generate text that is difficult to distinguish from human-written text. Claude 2 is also a multimodal model, and can process images and speech.
Vicuna-13B is an open source chatbot trained on LLaMA by collecting user-shared conversations. In early evaluations compared to GPT-4, Vicuna-13B achieves over 90% quality of OpenAI ChatGPT and Google Bard, and outperforms other models in over 90% of cases. Training costs are around $300, and the code and weights are publicly available for non-commercial use. But in practice, it doesn’t have that much power 😉
BLOOM is the world's largest open multilingual language model, with 176 billion parameters and capable of generating text in 46 natural languages and 13 programming languages. It is the first language model with over 100 billion parameters for languages such as Spanish, French, and Arabic, and is the result of a research project involving over 1,000 researchers from over 70 countries and 250 institutions.
Alpaca is a 7B parameter language model developed by the Stanford Center for Research on Foundation Models (CRFM), fine-tuned on 52K instruction-following demonstrations based on the LLaMA 7B model. It is surprisingly small, reproducible, and inexpensive (<$600), while achieving similar performance to OpenAI's text-davinci-003. For research purposes only, commercial use is prohibited.
🇰🇷 KoAlpaca is an open source language model that understands Korean. It is a project that trained an Alpaca model that understands Korean based on the Stanford Alpaca model. This model can be used in chat-type web pages, KakaoTalk bots, Telegram bots, etc., and provides various Korean-based and English-Korean models.
🇰🇷 Polyglot is a large-scale language model developed to improve non-English language performance compared to various multilingual models. This project attempts to model various languages, including Korean data.
🇰🇷 Kakao Brain KoGPT API uses KoGPT, a GPT-3-based language model, to understand Korean lexically and contextually and to generate sentences that match the user's intent. It can perform all tasks related to Korean, such as judging whether a given sentence is positive or negative, summarizing the content, predicting the conclusion, answering questions, and writing the next sentence. It is used to solve high-level language tasks such as machine reading comprehension, machine translation, writing, and sentiment analysis.
🇰🇷 HyperCLOVA X is Naver's large-scale AI, an upgraded version of the existing HyperCLOVA. By combining the customer's own data with HyperCLOVA X, it can provide immediate responses that meet the user's needs, and it provides functions that improve productivity by becoming a powerful backbone in various areas such as reading, writing, coding, searching, summarizing, consulting, recommending, and planning.
Character.ai is a service that allows users to create their own characters and have conversations with them. Characters can have various characteristics such as gender, appearance, personality, and hobbies, and users can define them themselves. When talking to a character, the character can generate various texts based on the user's questions and requests and respond appropriately to the context.
🇰🇷 Nutty Messenger is a messenger that you can use with your AI friend Ruda, and you can enjoy various activities such as daily conversations, games, and sharing your honest feelings. You can build intimacy through conversations with Ruda, and you can enjoy games such as word games and number guessing games, providing a unique experience of building memories with Ruda. (You can talk to speakers such as Iruda and Kang Da-on.)
🇰🇷 DearMate is a service where users and AI mates share small moments of happiness in their daily lives, and you can create intimate relationships with various AI chatbots. You can share your daily life and emotions by exchanging DMs with characters such as Coco, Mars, and Bluny, and each character has its own unique personality and characteristics, providing a variety of conversation experiences.
ChatSonic is a conversational AI chatbot powered by WriteSonic, with GPT-4-based features. It can have real-time conversations about current events and trending topics, and offers various features such as digital AI artwork generation, conversations with personalized avatars, and content suggestions via Chrome extension.
👍 ChatPDF is a service that lets you chat with PDF files. It is free and does not require membership. You can upload any PDF file, such as a book, research paper, manual, essay, or legal contract, and ask a question to receive an answer about the content of the PDF.
🇰🇷 CLOVA X is a conversational AI service built on HyperClova X, Naver's large-scale artificial intelligence. It was released on August 23, 2023, and can be used in synergy with Naver Shopping, Travel, etc. through skills (plug-ins).
🇰🇷 👍 Google's Bard is an AI service based on a large-scale language model (LaMDA). It is trained on a huge data set of text and code and can perform various tasks. It is currently available in beta form and provides simple Google WorkSpace integration.
Perplexity.ai offers a personal search assistant called Copilot. Upload a text or PDF file (up to 10MB) and it will find answers to a variety of questions, including "History of Argentina", "Unique flowers in Colorado", and "Checkout time at W Hotel CDMX". They also offer a Pro version that allows you to upgrade to GPT-4 to upload more files and improve your experience with Copilot.
Komo.ai is a platform where users can ask questions, discuss, and explore topics the community is talking about. The "Ask" function allows you to ask or discuss anything, the "Explore" function allows you to see topics the community is talking about, and the "Search" function allows you to get quick answers or links to resources.
🇰🇷 👍 Bing Chat is a chat-based search service that provides real-time search by integrating LLM with Microsoft's Bing search engine. Unlike the initial enthusiastic response, the popularity is cooling off due to restrictions such as the fact that it can only be used on the Edge browser.
You.com is an AI-powered search engine that keeps your data 100% private while providing a personalized search experience for you. Personalize your search with over 150 apps, including StackOverflow, Medium, Twitter, and more. Recently, we have enhanced our search with GPT-4 and Stable Diffusion XL.
🇰🇷 We provide insights and action items that will maximize your marketing performance based on 3 petabytes of search data secured from Google and Naver. It shows customers' search intents and search paths in high-resolution maps and includes the "GPT Analysis" function.
Writing assistance
Model/Service Name
Explanation
Jasper
Jasper is an enterprise AI and AI marketing tool that helps teams quickly create blog posts, marketing copy, AI-generated images, and more. It uses the best models, including OpenAI’s GPT-4, Anthropic, and Google’s, and combines them with recent search data, brand voice, SEO, and grammar optimization tools.
🇰🇷 CLOVA Studio is a no-code AI tool based on the ultra-large AI HyperCLOVA provided by NAVER CLOUD PLATFORM. With this tool, you can easily perform various tasks such as sentence generation, summarizing long articles, classifying sentences or emotions, creating conversational interfaces, and sentence conversion.
🇰🇷 Wrtn.ai is an AI portal for everyone, providing various generative AI services. It also aims to be a platform that can use language models such as GPT-3.5, GPT-4, and PaLM2 to chat, create images, and create your own AI.
Compose.ai is a Chrome extension that reduces your writing time by 40% with AI-based autocomplete and text generation. It works with Google Docs and Gmail and offers autocomplete, sentence restructuring, email composition, and response generation. It also learns your writing style to provide personalized suggestions and integrates with a variety of tools including email, Slack, Notion, Coda, and more. It is free to use, with a premium version offering advanced features including personalization.
Rytr is an AI-powered writing assistant that helps you create high-quality content in seconds, including blog posts, emails, and advertising copy. Choose from over 40 use cases and templates, 30+ languages, 20+ voice tones, and a rich text editor to quickly transform your raw ideas into finished work.
🇰🇷 Wordtune is an AI-powered writing tool that helps you write more clearly, persuasively, and authentically. Wordtune understands the sentences you type and suggests better words and expressions based on the context. Wordtune can also check grammar and spelling, and suggest ways to make your sentences more concise and fluent.
HyperWrite is an AI-based writing tool that helps users write high-quality content faster and easier. HyperWrite uses cutting-edge AI technology to understand the sentences you type and suggest better words and expressions based on the context. HyperWrite can also check grammar and spelling and suggest ways to make your sentences more concise and fluent.
Copy.ai is an AI-powered writing tool that helps users write better marketing copy and content. Copy.ai is built on the GPT-3 language model and helps users write a variety of content types.
Hypotenuse AI is an AI-based writing tool that helps users write high-quality content faster and easier. Hypotenuse AI uses cutting-edge AI technology to understand the sentences users type and suggest better words and expressions based on the context. Hypotenuse AI can also check grammar and spelling and suggest ways to make sentences more concise and fluent.
🇰🇷 👍 Microsoft 365 Copilot is an AI-powered productivity tool that helps you get things done faster and easier. Copilot understands what you type and can automate tasks or provide suggestions based on that.
🇰🇷 Google Workspace AI Solutions are solutions that help you optimize your business and increase productivity using the AI capabilities of Google Workspace. These solutions provide the following capabilities: Already being applied to document creation, schedule management, customer support, and more.
👍 Mem.ai is an AI-based note-taking app that helps users organize their ideas and get new ideas faster and easier. Mem.ai offers the following features: You can create 100 notes per day with a free account. Paid accounts offer more features and usage.
🇰🇷 Notion AI is an artificial intelligence feature that can be used within Notion to help you get things done faster, improve your writing, and think big. It automates complex tasks, summarizes important content, analyzes meeting minutes, corrects grammar and spelling, translates to multiple languages, and edits tone.
🇰🇷 LINER is an AI-based workspace that understands content faster in the browser and creates new searches. LINER AI provides accurate answers quickly, stores important information in one place, and allows you to access it at any time. You can also enjoy LINER AI on your smartphone, and it provides various functions utilizing GPT-3.5 and GPT-4. You can choose to use the basic and pro versions, and it provides functions such as Copilot functions on YouTube, PDF, and web pages, and use of LINER AI on Google.
Tome is a new way to express and share ideas using AI. Instead of staring at a blank page, it creates one-page, presentation, mood board, etc., finds the right tone and words for your writing, searches the web for references, and more. It also deepens and clarifies your already written work, and automatically generates images.
👍 Gamma is a new way to generate ideas using AI, creating beautiful and engaging content without formatting or designing. Create documents, presentations, and web pages in seconds, style entire decks with a click, and embed GIFs, videos, charts, websites, and more. Plus, it’s readable on any device, measures engagement with built-in analytics, and streamlines collaboration with quick responses and comments. Gamma is a new way to present ideas that’s more visual than documents, more collaborative than slide decks, and more interactive than videos.
Loopin is an AI meeting assistant that helps you conduct meetings effectively. It converts meeting recordings into text and automatically generates meeting notes, and links related meetings and notes to make it easy to find important meetings and related notes. Loopin AI’s conversational chat also ensures you don’t miss important details, and automatically shares meeting notes via email, Slack, Notion, etc. Just log in with your Google Workspace account, join a meeting, start recording, and you’ll get human-quality meeting notes after the meeting.
🇰🇷 👍 Clova Note is a voice record management service that utilizes AI technology provided by Naver. It can be used in situations where you need to remember conversations such as meetings, conferences, and interviews, and it automatically converts recorded voices into text so that you can easily find and listen to the information you need. It provides various functions such as memo function, bookmark function, AI summary, search, and sharing, and you can organize voice records in an easy-to-view manner, check important conversation moments, automatically summarize, find and check only the necessary voice records, and easily share them with a link.
OpenAI Codex is an AI system that converts natural language into code, and is the model that GitHub Copilot is based on. It supports dozens of programming languages, including Python, and can interpret and execute natural language commands from users. As a successor to GPT-3, it can generate code that works with natural language understanding, allowing you to issue English commands to software and APIs. As a general programming model, it can be used for code transformation, code explanation, and code refactoring.
👍 GitHub Copilot is an AI-powered programming assistant that works within your editor to provide suggestions for entire lines or functions. It translates natural language prompts into code suggestions in dozens of languages, helping you spend less time writing boilerplate and repetitive code patterns and more time building software that matters. It integrates directly into editors like Neovim, JetBrains IDEs, Visual Studio, and Visual Studio Code, and helps you work with new languages and frameworks. It’s available for personal and business use, so you can focus on developing faster and doing more satisfying work.
Amazon CodeWhisperer is an AI coding assistant trained on billions of lines of code that generates code suggestions in real time, from snippets to entire functions. It supports 15 programming languages, including Python, Java, and JavaScript, and integrated development environments (IDEs) like VS Code, IntelliJ IDEA, and AWS Cloud9, and provides security scanning to help you find and fix security vulnerabilities immediately. It’s free for personal use, and helps you code faster, improve security, and use the tools you love.
Phind is a personal programming assistant and search engine designed to help users with their coding. It is currently available in alpha for VSCode and is built in San Francisco. Phind connects you with the developer community.
LLM, a technology-based AI, aims to enhance learning modes, focus on knowledge sharing, and find new ways for engineers to spend more time creating creative things. Through various projects such as experiments, research, and vision, we share ideas and opinions on emerging technologies combined with existing platforms and services.
The AI Assistant in JetBrains IDEs is a major new feature in all IntelliJ-based IDEs and .NET tools. It leverages generative AI and large-scale language models to build a tight integration of AI capabilities with your code. The AI Assistant offers AI chat, documentation generation, name suggestions, commit message generation, and more, and transparently connects to a variety of large-scale language models based on JetBrains AI services.
Code-LMs, commonly known as PolyCoder, provides guidance on how to use large-scale language models in source code. The project trains and publicly releases large-scale neural language models for programs, and describes a variety of models, including PolyCoder. It was made available on Huggingface in October 2022, and provides several models trained on large corpora covering a variety of programming languages.
📚 Welcome to Haebom's archives. --- I post articles related to IT 💻, economy 💰, and humanities 🎭. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? 🔔 Yes, that means subscribe. haebom@kakao.com
