2023, isAI Industrya watershed moment.

This year, we have witnessed the leapfrog development of AI technology.

From deep learning to natural language processing, from image generation to video generation, from voice cloning to digital human cloning... many AI tools and products are like rising stars, not only reshaping people's lifestyles, but also redefining the future business landscape.

As John Culkin said: “We shape our tools, and then our tools shape us.”

"Top AI Player" sorts out and reviews the popular tools in the AI field in the past year, hoping to help you better review the breakthroughs and achievements of AI technology.up to dateprogress and foresee how they will continue to impact our world.

We selected some representative AI products in several common fields based on valuation, influence, user evaluation and other dimensions, and sorted out their iteration history,up to datePerformance, etc. Each category has its own unique technical characteristics and application scenarios.

AI Chatbot

AI chatbots are one of the hottest and most representative development trends in the field of AI, representing a change in the way people obtain information, make decisions and communicate.

Currently, AI chatbots are available on the market in many forms, including standalone mobile applications, messaging applications integrated into social networks or search engines, etc.

ChatGPT

SurfaceStrongestAI chatbot, developed by OpenAI and released on November 30, 2022.

The emergence of ChatGPT not only promoted the development of natural language processing technology, but also promoted the popularization of AI technology and increased the social awareness and influence of AI.

Users can access GPT-4 (OpenAI) by subscribing to ChatGPT PlusFirstAdvanced language model), faster response, more features, more stable service and more flexible usage, the subscription fee is US$20 per month.

On November 6, 2023, OpenAI held its first developer conference (OpenAI DevDay) and announced a series of updates to GPT, including the launch of GPT-4Turbo (GPT-4’ssuperversions) and multi-mode APIs, etc.

It is worth mentioning that OpenAI will officially launch the GPT store this week, where users can create custom GPTs and profit from them.

Claude

The AI chatbot of the American AI startup Anthropic (founded by former members of OpenAI) was officially released on March 15, 2023.

In July 2023, Anthropic released Claude2. The upgraded Claude2 has improved performance in coding, mathematics, and reasoning, and its processing capacity has been increased to 100K tokens. It can process hundreds of pages of technical documents or even entire books.

On November 22, 2023, Anthropic released Claude 2.1. The context window reached 200,000 tokens, which is twice the previous processing capacity of Claude and significantly higher than the upper limit of 32,000 tokens of the GPT-4 Enterprise Edition.

Anthropic also said that Claude2.1's frequency of "hallucinations" or lies was half of what it used to be.

As of now, Anthropic’s valuation is close to $5 billion, with total financing of nearly $1.5 billion.

Bard

On February 6, 2023, Google launched Bard, an AI chatbot driven by the LaMDA large model.

On April 10, 2023, Bard switched to the more powerful PaLM large language model, and its computing power was enhanced.

On May 10, 2023, PaLM was further updated to PaLM2, with enhanced multilingual translation and logical reasoning capabilities.

Bing Chat

On February 7, 2023, Microsoft officially integrated GPT-4 into the new version of Bing and Microsoft Edge browser. The integrated chatbot is called Bing Chat.

On March 4, 2023, Microsoft introduced "Precise", "Balanced" and "Creative" modes for Bing Chat. Users can switch between these three modes to experience different chat tones.

On March 22, 2023, Bing Chat integrated the Bing Image Creator feature. This feature is based on OpenAI's DALL-E and can automatically generate images based on the text content entered by the user.

Bing Chat is considered an alternative to ChatGPT Plus’ $20 monthly subscription due to its free and easy use.

Character.ai

It was co-founded in 2021 by former Google LaMDA team members Noam Shazeer and Daniel De Freitas, and launched its beta version in September 2022.

Character.ai has built an AI role-playing community where users can communicate and converse with anime characters, celebrities, and various customized characters.

On May 23, 2023, Character.ai mobile app officially landed on iOS and Android systems worldwide. According to official data released by Character.ai, since its release in May 2023, its Android app market downloads have exceeded 3 million times.

In September 2023, Character.ai's valuation was revealed to be over US$5 billion.

Pi

Pi is an AI chatbot launched by the American AI startup Inflection AI in May 2023. Unlike ChatGPT and other products that are positioned as productivity tools, Pi focuses on companionship and emotional intelligence.

Inflection AI was founded in 2022 by former DeepMind executive Mustafa Suleyman. It has received investments from companies such as Microsoft and Nvidia and is currently valued at US$4 billion.

Perplexity.ai

Perplexity.ai is a free AI chatbot that supports online search. Click "Popular Now" below the text box to viewMost popularTips and news.

Perplexity.ai is an AI-driven search engine. Unlike traditional search engines, Perplexity.ai has a chatbot-like interface that allows users to ask questions in natural language and directly provide answers when answering search queries rather than website links. Perplexity calls this product an "answer engine."

On January 4, 2024, Perplexity completed a $73.6 million Series B financing round, with a valuation of $520 million, led by Institutional Venture Partners. This is also the largest round raised by Internet search startups in recent years.maximumA sum of money.

Prior to this round of financing, Perplexity.ai's monthly active users had increased to 10 million.

Grok

Grok is launched by Musk's xAIThe firstThe AI big model product will be launched in November 2023. The big model behind it also has the same name. The current version is Grok-1. Its prototype Grok-0 began training after xAI was announced.

Compared with large models such as ChatGPT, which have specific knowledge base deadlines, Grok can obtain knowledge from the 𝕏 platform in real time.up to dateInformation, providing users with more timely news retrieval and opinion acquisition services.

In addition, unlike the rigid responses of common AI assistants, Grok's responses are humorous and rebellious.

Gemini

In the early morning of December 6, 2023, Google released the multimodal large model Gemini.

Gemini is available in three versions: Gemini Ultra for highly complex tasks, GeminioptimalModel Gemini Pro and Gemini Nano for end devices (mobile phones, PCs).

Currently, Bard has integrated a fine-tuned version of Gemini Pro. In the future, Gemini will be gradually integrated into multiple products and services such as Google Search, advertising, Chrome browser and Duet AI to enhance the intelligence level of the Google ecosystem and provide users with a more accurate and personalized experience.

Janitor AI

Janitor AI is a role-playing AI chatbot platform whose core function is to allow users to create fictional chatbot characters and interact with them in natural language.

Users can choose different role templates, including personality, language style, hobbies and other settings, to inject diverse personalities into their chatbot roles. In addition, Janitor AI provides a wealth of APIs and SDKs to facilitate developers to integrate it into their own applications.

A Word from the Heart

On March 16, 2023, Baidu’s large language model product “Wenxin Yiyan” was officially released. This is the first Chinese language model product after OpenAI released ChatGPT.FirstA large generative language model product.

Wenxinyiyan possesses five major abilities: literary creation, commercial copywriting, mathematical logic reasoning, Chinese comprehension, and multimodal generation.

In October 2023, Wenxin Big Model 4.0 was launched, bringing more than ten AI native applications such as the fully reconstructed new search. As of the end of December, the number of Wenxin Yiyan users has exceeded 100 million.

iFlytek Spark

On May 6, 2023, iFLYTEK officially released the "iFLYTEK Spark Cognitive Big Model", which surpassed ChatGPT in three major capabilities: text generation, knowledge question and answer, and mathematical ability.

In June 2023, iFlytek Spark Cognitive Big Model passed the domesticThe firstThe official and trusted AIGC large model basic capability (function) evaluation has been completed, and all functional items have been certified.

In October 2023, iFlytek Spark Cognitive Big Model V3.0 was released, with seven major capabilities continuously improved, surpassing ChatGPT overall, and the six core capabilities of medical care surpassing GPT-4.

Thousand Questions on Tongyi

The conversational AI model launched by Alibaba began internal testing on April 7, 2023.

In September 2023, Tongyi Qianwen became one of the first large models in China to pass registration. After the launch of Tongyi Qianwen APP, its functions continued to be upgraded. It currently provides dozens of functions such as text dialogue, voice dialogue, translation, PPT outline assistant, Xiaohongshu copywriting, video generation, etc.

At the same time, Alibaba Cloud has successively open-sourced Qwen-7B, Qwen-14B, Qwen-1.8B, the visual understanding model Qwen-VL, and the audio understanding model Qwen-Audio. In early December, the 72 billion-parameter large language model Tongyi Qianwen Qwen-72B was officially open-sourced, which is known as the "industry's most powerful language model".StrongestChinese open source model".

Bean curd

Doubao is an AI conversational product developed by ByteDance based on the Skylark model. It officially started external testing on August 17.

Doubao provides functions such as chatbot, writing assistant and English learning assistant. It can answer various questions and conduct conversations. It supports web, iOS and Android platforms, but iOS needs to be installed using TestFlight.

Kimi Chat

Kimi Chat is a large-scale model product developed by Beijing Moonshot AI Technology Co., Ltd. (Moonshot AI) and was officially launched on October 9, 2023.

Kimi Chat's unique advantage is that it has super-long context support, supporting the input of 200,000 Chinese characters of text content. It can also handle multiple file formats, such as TXT, PDF, Word documents, PPT slides, Excel spreadsheets, etc., and has the ability to browse URLs and reply to users after reading relevant content.

AI Image Generation Tool

2023 is a year of rapid development in the field of AI image processing. However, at present, AI-generated images still have some limitations that need to be further overcome, such as insufficient details and precision, and in most cases, there are still image flaws and defects. The adjustment of light, shadow, and color tone relies more on post-processing.

Midjourney

Midjourney is a pioneer in the field of cultural graphics andLeaders, the quality of the pictures it generates has always been the industry benchmark.

An image generated with Midjourney won the Colorado State Fair Digital Art Competition in 2022.FirstIt has attracted public attention to AI painting and Midjourney.

Currently, Midjourney has been updated to version V6. The quality of the generated images has been gradually improved, and the functions have become more diverse and complete.

Initially, Midjourney was hosted on Discord and could only be accessed by messaging a Discord bot on its official Discord server.

On December 13, 2023, Midjourney launched a web version, but the threshold for use is to generate more than 10,000 pictures with Midjourney (you can enter "/info" in Discord to view the number of pictures generated). Compared with Discord, the web version of Midjourney is easier to operate, but has fewer functions.

Stable Diffusion

Stable Diffusion is an AI painting tool based on the diffusion model. It was developed by Stability AI and can complete tasks such as text-to-image and image-to-image. It was released on August 22, 2022.

Stable Diffusion is a completely open source project, including model code, training data, papers, etc., which enables it to quickly build a powerful and prosperous upstream and downstream ecosystem, such as the AI painting community Civitai, SD-based self-training models, and a wealth of auxiliary AI painting tools and plug-ins.

In June 2023, Stable Diffusion released an update to SDXL version 0.9, which upgraded the Stable Diffusion Vincent graph model.

On November 29, 2023, Stability AI released the next-generation text-based graph model SDXL Turbo, which reduces the number of steps required to generate an image from 50 to 1, and significantly improves the inference speed, enabling real-time image generation. On A100, SDXL Turbo can generate a 512x512 image in 207 milliseconds.

However, the installation and use of Stable Diffusion has high hardware requirements.

DALL E3

DALL·E3 is an image generation model released by OpenAI on September 21, 2023, and will be available to ChatGPT Plus and Enterprise customers in early October 2023.

DALL E3maximumThe feature is the integration with ChatGPT, which is natively built on ChatGPT to create, expand and optimize prompts. When a user enters an idea, ChatGPT will automatically generate a tailored and detailed prompt for DALL E3, and users can also use their own prompts.

This integration gives DALL·E3 a greater ability to comprehend and process abstract and lengthy prompts, making it easier for users to translate their ideas into accurate images.

Adobe Firefly

Adobe Firefly, also known as "Firefly", is a web application developed by Adobe. Its release marks an important breakthrough for Adobe in the field of artificial intelligence and AI drawing.

The main AI features include text-generated images and generative fill-in, where users can describe through simple text prompts, remove part of the image, add other content to the image, or replace it with generated content.

In addition, Adobe Firefly supports the use of simple text prompts in Creative Cloud applications, expanding the possibilities of combining application workflows with generative AI.

Leonardo AI

Leonardo is an AI painting community and also an AI painting tool.

Leonardo deeply integrates various plug-ins of Stable Diffusion, such as ControlNET's openpose posture reference, local redrawing, prompt prompt, etc., and even provides a fool-proof online model training function, which makes Leonardo more like a Stable Diffusion model sharing communityA combination of Civitai (civitai.com) and Stable Diffusion.

AI video generation tool

As the technology of visual graphics has become more sophisticated and mature, the visual video track has gradually become more lively, with visual video companies represented by Runway emerging one after another. Internet giants at home and abroad, such as Google, Meta, Microsoft, Alibaba, and ByteDance, have also invested personnel and energy to participate in it.

Runway Gen-2

Runway is an American AI startup founded in 2018. In February 2023, Runway released the text generation video models Gen-1 and Gen-2, which can be used by visiting the Runway official website through the web interface.

On November 2, 2023, Runway Gen-2 had a milestone update. The problems of flickering, incoherence, distortion and other issues that AI-generated videos were criticized for in the past have been greatly improved after this update.

Now, whether using Gen-2 text-generated video or image-generated video, the fidelity and consistency of the video have been greatly improved, and the resolution has been increased to 4K level.

So far, Runway has released more than 30 AI creation tools, including audio, image, video, 3D and generation, covering almost all audio and video content generation and processing tools. Its products have been used in many Hollywood blockbusters.Special EffectsProduction.

In July 2023, Runway raised approximately $100 million in a Series D round of financing led by Google, and its valuation has now reached $1.5 billion.

Pika Labs

Pika Labs is called Runway Gen-2StrongestAs a competitor, its emergence has expanded the investment circle's imagination of the field of AI video entrepreneurship.

On November 29, 2023, Pika Labs releasedFirstThe first product, Pika 1.0, quickly became popular due to its amazing video generation effect. On December 26, Pika 1.0 started free public beta.

The founders of Pika Labs are two Chinese, Guo Wenjing (CEO) and Meng Chenlin (CTO), both of whom are doctoral students from Stanford AI Laboratory.

On November 29, Pika labs announced the completion of a US$55 million Series A financing round, and its current valuation is nearly US$200 million.

Stable Video Diffusion

On November 21, Stability AI launched the video generation model "Stable Video Diffusion". This model is based on Stable Diffusion's existing text-to-image model and can generate videos by animating existing images.

Stable Video Diffusion provides two models, SVD and SVD-XT. SVD converts still images into 14-frame 576x1024 videos, while SVD-XT increases the frame rate to 24 frames under the same architecture. Both can generate videos at 3 to 30 frames per second.

Currently, Stable Video Diffusion has opened user waiting list registration.

Morph Studio

Morph Studio is the dark horse in the field of video.FirstA team that launched a text-to-video product that the public can test freely, even before Runway opened Gen2 to public beta.

Unlike some similar products that only provide 720P free services, Morph Studio has provided free services with a default 1080P and a maximum generation time of 7 seconds from the beginning. You can experience it for free by registering on Discord.

Animate Anyone

Animate Anyone is a software that can turn static images into animated videos. It was developed by Alibaba Intelligent Computing Research Institute. It can be applied to different types of characters such as humans, anime, and cartoons. It only needs to provide a character image and some preset action sequences to generate realistic animated videos.

Another tool similar to Animate Anyone is Magic Animate, a "human image animation generation tool" jointly launched by the National University of Singapore and ByteDance. It can also generate corresponding animated videos based on user-specified character images and action sequences.

AI Audio Tools

After experiencing the visual shock brought by AI painting tools such as Midjourney and SD, a revolution is also taking place in the field of AI-generated audio.

From the AI singer Stefanie Sun who shocked the Chinese music scene to the viral video of Taylor Swift speaking Mandarin, AI audio generation products have made significant breakthroughs in music creation, speech synthesis, and sound effect design.

ElevenLabs

ElevenLabs is a Text to Speech software that can convert input text into speech with realistic emotions and intonation.

The ElevenLabs behind it is a software company that specializes in using artificial intelligence and deep learning to develop natural speech synthesis and text-to-speech software.

In June 2023, ElevenLabs raised $19 million in Series A funding at a valuation of approximately $100 million.

In October 2023, ElevenLabs launched "AI Dubbing", an AI tool that can translate speech into more than 20 languages while retaining the speaker's original voice, emotion, and intonation.

Suno AI

Suno AI is a generative music model that can generate audio, including speech, music, and sound effects, from short text prompts.

Among them, Suno AI's speech generation model BaRK can generate various voices according to user needs and is suitable for the advertising, animation and gaming industries.

Suno AI's music generation model Chirp can generate music clips of about 30 seconds including instruments, lyrics and vocals, covering a variety of music styles including pop, classical, and electronic.

Suno AI’s sound generation model can generate various types of sound effects to add expression, atmosphere, and emotion to audio and video projects.

Mubert

Mubert is an AI music generation platform where users can generate music of specific length, style, genre and mood in real time, and support customization. It is mainly aimed at music producers, creators and brands, enabling them to create royalty-free music with the help of artificial intelligence.

Google MusicLM

Google MusicLM is a text-to-music generative model developed by Google as part of the "AI Test Kitchen" program.

MusicLM can create high-fidelity music from simple text descriptions similar to natural language prompts. It generates music at a high sampling rate of 24kHz, which means the generated audio quality is very high. In addition, MusicLM's music generation speed is very fast, almost instantaneous.

AI digital human generation tool

With the breakthrough progress of artificial intelligence technology, AI digital humans have become a hot field in 2023 with their realistic appearance, intelligent conversation capabilities and personalized services.

However, at the technical level, AI digital human products still need further breakthroughs in the future.Image synthesis, speech synthesis and emotion simulationIn terms of business, as competition intensifies, product differentiation and user experience may become key factors in determining market competitiveness.

In addition, there is a need to strengthen supervision of data collection, storage and use to protect users' privacy rights and interests and ensure the legal, fair and transparent use of digital human technology.

Synthesia

Synthesia is an AI video creation platform that is mainly aimed at B-side customers such as large corporate clients and can generate virtual human videos, etc.

Synthesia CEO once revealed in a blog that 35% of the world's Fortune 100 companies are using Synthesia for training and marketing, and more than 50,000 teams are using this tool to produce videos on a large scale, saving 80% of the budget.

The company behind it, Synthesia, is a British AI startup founded in 2017. In June 2023, the company received approximately US$90 million in financing and its valuation reached US$1 billion.

HeyGen

At the end of October 2023, a video clip of Taylor Swift, a famous American female singer, speaking Mandarin went viral on the Internet, and the tool HeyGen used behind it also attracted widespread attention for a while.

HeyGen is a digital human generation platform that was launched on July 29, 2022. It took 178 days to reach $1 million in ARR (annual recurring revenue).

If Runway and Pika are mainly aimed at creative personnel and consumers, HeyGen focuses on solving the needs of B-side customer marketing, training and instructional videos.

On November 29, 2023, HeyGen announced that it had received $5.6 million in venture capital from Conviction Partners led by Sarah Guo. This round of investment brought HeyGen's valuation to $75 million.

D-ID

D-ID is a company that provides services and development of artificial intelligence human-like video products. Users only need to upload a photo of the person and input what they want to say (lines), and D-ID can use the AI voice robot to automatically convert the user's input into a video.

The main technology of D-ID is the facial de-identification technology service, which can create a virtual narrator who replaces the real person in the video and introduces the video content.

AI Efficiency Tools

Due to its huge user base, many work scenarios are also compatible with AIGC's capabilities, and office is one of the scenarios that is naturally suitable for the implementation of AI.

As more and more office software is using AI, AI can now directly meet our needs as long as we describe them in natural language. Writing meeting minutes, copywriting, drawing, developing applications, automatically generating PPT and Excel tables, etc. are all no problem.

QuillBot

QuillBot is an article summary writing and polishing tool based on NLP. It can automatically help users rewrite, summarize and expand articles through semantic analysis.

This type of article assistant has developed rapidly in the past year, but QuillBot has recently faced a certain loss of users. Some analysts say that this is mainly related to ChatGPT's powerful zero-sample learning ability. The latter can achieve unlimited topic creation through simple prompts, which is obviously more attractive.

But in terms of actual results, professional writing assistants such as QuillBot still have an advantage. They can provide richer grammar, logic and style guidance, and output more fluent and logical articles.

Novel AI

Novel AI is an AI tool designed for content creators. It is mainly used to assist writing. It can help writers and creators generate new ideas, provide writing inspiration, and even automatically complete or edit stories.

Jasper AI

Jasper AI is a popular AI writing assistant designed to help users create content faster and more efficiently, mainly targeting user groups such as advertising professionals, content marketers, entrepreneurs, etc.

Jasper AI also offers a variety of writing templates, including blog posts, social media posts, marketing emails, and web content.

Copy AI

Copy AI is an AI-driven content generation tool that can automatically generate creative copy, marketing text, and other types of writing content, especially for the marketing and advertising fields.

In addition, Copy.AI also provides a built-in document editor where users can enter instructions or questions on the left and edit and optimize the output results on the right.

Notion AI

Notion AI is an AI function integrated into the Notion product. Notion is a note-taking and project management tool. Its integrated AI functions include text generation, content organization, data analysis, etc., which are designed to help users manage notes, organize projects, automate routine tasks, etc., and improve work efficiency.

Looking back at 2023, we have witnessed vigorous development and innovation in the field of artificial intelligence.

In addition to the large-model and generative AI unicorn companies that have received much attention, emerging AI products with star founding teams and broad application prospects are also likely to attract the favor of various investors.

With the continuous advancement of AI technology, the continuous accumulation of data, and the further improvement of computing power, it is foreseeable that in the next few years, AI products and applications will become more and more abundant, and AI technology will continue to penetrate into a wider range of fields, including medical care, finance, manufacturing, etc. AI will bring more intelligent solutions to these fields, thereby improving efficiency, reducing costs, and promoting the transformation and upgrading of the industry.

At the same time, how to ensure the fairness, transparency and explainability of AI systems, how to balance the relationship between AI development and privacy protection, and how to avoid the abuse of AI technology or the occurrence of potential risks will also become important issues.