Article directory
AIGC
AIGC, the full name of Artificial Intelligence Generated Content, is a new type of artificial intelligence technology, mainly used to solve the problems of large-scale data processing and analysis. AIGC technology introduces the graph computing model into the field of artificial intelligence, can efficiently process graph-structured data, and has strong data mining and deep learning capabilities. Compared with traditional artificial intelligence technology, AIGC technology performs better when dealing with highly connected data. At the same time, AIGC technology can realize distributed computing and fast storage, and can realize high-performance computing and analysis on large-scale data sets, and has broad application prospects.
Compared with traditional artificial intelligence technology, AIGC technology has the following advantages:
- Efficient processing of graph-structured data: AIGC technology can efficiently process large-scale graph-structured data, and is especially suitable for data scenarios that need to analyze highly connected relationships.
- Realize distributed computing: AIGC technology supports distributed computing, which can realize multi-machine collaborative computing and improve computing performance.
- Fast storage and retrieval: AIGC technology supports fast storage and retrieval, and can quickly obtain the required data and calculation results.
- Wide application: AIGC technology has broad application prospects and can be applied to social network analysis, financial risk control, medical image analysis, intelligent transportation and other fields.
GAI
Generative artificial intelligence (GAI)
AGI
Artificial General Intelligence (AGI), also known as "Strong AI" and "Full AI", is a machine intelligence that has general human intelligence and can perform any intellectual tasks that humans can perform. Artificial general intelligence is a major goal of some artificial intelligence research and a common topic in science fiction and futuristic studies. In contrast to weak AI (weak AI), general artificial intelligence can attempt to perform the full range of human cognitive capabilities.
Many definitions of general intelligence (such as being able to pass the Turing test) have been proposed, but no one definition can be agreed upon by everyone. However, artificial intelligence researchers generally agree that the following characteristics are necessary for a general artificial intelligence:
- Automatic reasoning, using strategies to solve problems and make decisions in an environment of uncertainty.
- Background knowledge, including common sense knowledge base.
- Automatic planning.
- transfer learning.
- Communicate using natural language.
- And, integrate the above means to achieve the same goal.
There are also important capabilities, including machine perception (such as computer vision), and the ability to act in a world of intelligent behavior (such as the ability of a robot to move itself and other objects). It may include the ability to detect and avoid danger. Many intersectional fields of study of intelligence (such as cognitive science, machine intelligence, and decision-making) try to emphasize additional characteristics, such as imagination (the ability to construct mental images and concepts without relying on presuppositions) and autonomy. Indeed, many of these capabilities already exist in computer-based systems, such as computational creativity, automated reasoning, decision support systems, robotics, evolutionary computing, intelligent agents, yet these systems have not reached human levels.
Again, Weak General Intelligence (Weak AI), also known as "Narrow AI (Artificial Narrow Intelligence, ANI)", is application- or task-specific artificial intelligence.
It is an artificial intelligence that is good at performing a single task. Speech and image recognition are still narrow AIs, although their advances look tantalizing. Even Baidu's translation engine, despite its complexity, is a narrowly defined form of artificial intelligence.
application
Common ones include producing text, generating pictures, generating videos, generating audio, and generating modeling models
1. Computer field
(1) Application development, AI intelligent design program or algorithm, assisting manual application development. Reduce labor costs and improve application development efficiency
(2) Code writing, AI intelligently writes codes, reduces code errors, improves code operation efficiency, and reduces labor costs. For example, Ghostwriter of GitHub Copilot Replit is an artificial intelligence-driven programming assistant that can make suggestions for code and assist programming. OpenAI large model + evolutionary algorithm training AI to write code and modify code
(3) Hardware design, assisting in computer hardware design such as chip design, reducing labor costs and improving design efficiency. Google published an article in Nature saying that AI can generate chip design drawings within 6 hours, and design better than humans
2. Financial field
(1) Financial information, which automatically compiles information based on algorithms, and automates part of the editing work. Improve the speed of information generation, improve the ability to capture hot spots, and increase the timeliness of information
(2) Virtual customer service, providing virtual customer service consultation for financial customers, reducing manpower time costs, and providing customers with personalized services
(3) Data report, assisting analysts to capture data, conduct data analysis, and generate preliminary reports with workload advantages. Improve the timeliness, comprehensiveness, and accuracy of data analysis, reduce labor costs, and improve analysis efficiency
3. E-commerce field
(1) 3D appearance, based on product images from different angles, the 3D geometric model and texture of the product are automatically generated with the help of visual generation algorithms. The 3D shooting and generation of products can be completed within minutes, and the accuracy can reach millimeter level. Display the theme appearance of products in an all-round way, greatly reduce the communication time for users to select products, improve user experience, and quickly close transactions. For example, Alibaba’s per-square-foot-per-house business uses AIGC technology to realize the simulated display effect of “putting goods at home” online
(2) Online try-on, the 3D product model generated by the AI algorithm can be used for online try-on, highly restoring the experience of product or service trial. It highly restores the experience of product or service trial. The average conversion rate of 3D shopping is 70%, which is 9 times higher than the industry average level. Such as Uniqlo virtual fitting, Adidas virtual shoes, Chow Tai Fook virtual jewelry, Gucci virtual watches and glasses, IKEA virtual furniture matching, Porsche virtual test drive, etc.
(3) Virtual anchor, based on vision, voice, and text generation technologies to create a virtual anchor, to provide viewers with 24-hour uninterrupted product recommendation and introduction. Fill in the gaps in the live broadcast of real-life anchors and broadcast continuously 24 hours a day; accelerate the process of rejuvenating stores and brands, and shorten the distance with new consumers; people are more stable and controllable, and are not afraid of "collapsing houses". Brands such as L'Oreal, Philips, Kazilan, and Perfect Diary have launched their own brand virtual anchors
(4) Customer service consultation, AI virtual customer service, intelligent question and answer, automatic reply, timely solution to consumer problems, all year round, no upper limit of reception, saving about 80% of the cost compared with manual labor. For example, JD Yanxi 2.0 can provide 10 million intelligent services per day and 2 million hours of voice calls per month
(5) Creative material generation, automatically generating marketing copy, promotional pictures, improving personalization and pertinence. For example, Alibaba's AI designer "Luban" can generate advertising materials for merchants
(6) Construction of the virtual freight yard scene. By reconstructing the three-dimensional geometric structure of the scene from the two-dimensional image, an interactive three-dimensional shopping environment can be generated quickly, at low cost, and in large quantities, so as to enhance the sense of immersion and consumption experience. For example, Alibaba's virtual reality program "Buy+" provides a 360° virtual shopping experience on-site open shopping
4. C-end entertainment
(1) AI face-changing application, generating AI face-changing images or videos. It is mostly used for secondary creation and sharing on social platforms, which can largely meet the needs of users for novelty hunting and become a sharp weapon for breaking circles. As soon as FaceAPP, ZAO, Avatarify and other image and video synthesis applications were launched, they immediately went viral and caused a boom on the Internet, ranking first in the free download list of the App Store
(2) Photo animation, provide original photos, input keywords and parameters, select styles and templates, and then automatically generate animation pictures, which are easy to operate and meet the needs of users for novelty hunting. Products that can convert photos into anime styles such as Befunky, ToonApp, Colorinch, SocialBook
(3) Smart dubbing, adding smart dubbing to user-produced content. You only need to enter text content and select the type of human voice to generate dubbing, further reducing the threshold for video production. Lower the production threshold, more realistic and emotional than ordinary mechanical dubbing. Products that can be dubbed intelligently, such as Syntesys, Murf, Listnr, Lovo, Play.ht
(4) The voice changing function supports users to experience various voices such as uncle and loli, increasing interactive entertainment and enriching user experience. Voicemod can turn the user's voice into eight different roles such as Morgan Freeman, a pilot, and an astronaut
(5) The generation of interesting text content should be used to provide inspiration for the continuation of the novel and the construction of the world view of the characters, lower the writing threshold, and provide convenience for the novel author. Fast and easy to operate. For example, the Caiyun Xiaomeng AI continuation function provides three different story directions for users to choose. Users can start text and voice interaction with the characters they create, and AI will automatically generate chat content according to the character settings
(6) Virtual digital human, AI-constructed virtual person virtual image and definable virtual image and other "digital avatars" can interact through virtual images, such as trying on clothes, virtual socializing, etc. Lower the threshold for avatar customization and enrich user experience. Douyin, a subsidiary of Byte, has launched a business called "pheagee (pheagee)", and its platform is positioned as a "digital fashion creative platform". In addition, there are AI Studios, DeepBrain AI, Character Creator and other products
5. Gaming field
(1) NPC generation, learn human body movements on a large number of sports data sets, and generate walking, running, jumping and other actions. Integrate text-to-speech research, and affect facial changes such as mouth shape and expression according to voice synchronization. Generate faces, clothing, voices, and personality traits with the help of AI. In terms of application, NetEase AI Lab researchers designed whole-body action sequences for speech-text matching. The "Peace Elite" team of Tencent Games Photon S Studio joined hands with Tencent AI Lab and Tencent Games CROS GVoice (Tencent Game Voice) team to apply AI Codec to the "Peace Elite" game based on continuous breakthroughs in deep learning in voice codecs Among them, it is the first in the industry to realize AI Codec's lower bit rate and higher quality voice coding, thus becoming the first product to fully apply AI Codec technology to the field of game voice
(2) Environment generation, using AI algorithms to generate a larger open world environment, reducing production costs, increasing the scale of the map, and increasing the richness of the map
(3) Logic generation, including level generation, plot generation, battle mode and balance test. For example, Tencent AI Lab can generate 1000+ levels within a few hours for self-developed 2D mini games. Tencent's "Juewu" AI imitates real players of Glory of Kings through reinforcement learning, including indicators such as development, operation, and cooperation, as well as specific parameters such as hand speed, skill release frequency, hit rate, and number of kills per minute, making AI more accurate. Close to the real performance of players on the official server, increasing the overall accuracy of the test to 95%
(4) Battle training AI sparring, AI and players communicate and cooperate in a real battle environment, and teach players professional-level strategies and operating techniques in the process, helping players quickly familiarize themselves with the gameplay. After the King of Glory introduced the teaching of King's Absolute Enlightenment, the number of players' active communication in a single game has increased significantly, which has improved the playability of PVE gameplay
(5) Simulate opponents. By imitating professional players and mastering typical personal styles, players feel like they are fighting against real professional players. Increase game playability and enrich player experience. For example, Tencent AI Lab cooperated with the mobile version of "Cross Fire" to create a "star game" - challenging professional players. The AI mimics the pros, mastering their typical personal style, and the player feels like they are playing against real pros. This gameplay has been very popular since it was launched, and the number of games has increased by 3-4 times compared with the usual average
(6) Highlight generation, realize automatic editing, automatic soundtrack, automatic copywriting according to the game video, increase the generation speed and increase the generation scale
(7) Event commentary, realize automatic commentary according to the game situation, save labor costs, and realize 24-hour uninterrupted commentary
6. Education
(1) Convert 2D textbooks to 3D, and convert 2D textbooks into 3D textbooks to enrich textbook content and teaching content, improve students' learning interest, and make teaching content more specific and intuitive
(2) Synthetic virtual teachers. Synthetic virtual teachers have teaching abilities similar to traditional teachers to interact with students and can be customized. Improve teaching fun and reduce teaching costs
7. Industrial field
(1) Assist CAD design, automate repetitive, time-consuming and low-level tasks in engineering design, and shorten the engineering design cycle. SketchGraphs can assist architects, engineers and other users to design 2D and 3D prototypes using AutoCAD and SOLIDWORKS
8. Medical field
(1) Personalized rehabilitation, synthesizing speech audio for the voiceless, synthesizing limb projections for the disabled, and synthesizing non-aggressive medical companionship for patients with mental illness. Provide personalized rehabilitation services for patients. Rulingxin Intelligence is a platform that provides digital diagnosis and treatment services for mental health
(2) Intelligent diagnosis, assisting doctors in preliminary diagnosis, and intelligently generating diagnostic reports, improving medical efficiency, reducing doctors' workload, and helping accurate assessment and timely comparison of patients' conditions. Baidu's biological computing platform "Propeller PaddleHelix" provides a complete set of open source tool sets and computing platforms to support the construction of technical solutions for new drug research and development, vaccine design, and precision medical scenarios
9. Legal field
(1) Retrieval of legal provisions, quickly search relevant legal provisions according to the characteristics of the case, and assist in sorting out the applicable legal conditions. Convenient and fast, save time, reduce labor costs and improve accuracy
(2) Case analysis, analyzing the facts of the case, and giving the legal identification factors that should be considered in different situations, which play an auxiliary role in judgment for manual work
(3) Writing legal documents, AI generates legal documents according to the format and requirements
(4) Auxiliary judicial judgment, providing comparison of existing legal materials, especially legal provisions and judicial judgments, can realize auxiliary generation of judgment documents, automatic backfilling of case information and other functions
In 2023, the Colombian court used the text generation function in ChatGPT to increase the reasoning basis in the judgment. In the judgment document, ChatGPT gave specific legal terms, applicable situations, and legislative purposes. ChatGPT gave specific legal terms and applicable situations. , the purpose of legislation and the precedents of the Constitutional Court, etc., can effectively improve the handling of litigation cases
10. Agriculture/food field
(1) Auxiliary quality inspection, AI-assisted agricultural product quality inspection, and generate quality inspection reports, improve quality inspection efficiency and accuracy, and reduce labor costs. For example, the Pepsi potato chip factory applies a machine vision system, and the camera is used as a sensor to calculate the size, weight and quantity of potatoes, and judge whether the quality of potatoes is qualified
(2) Suggestions on distribution: AI generates a recommendation on the distribution ratio of agricultural products based on historical data analysis. Improve the sales efficiency of agricultural products and lower the threshold. For example, Coca-Cola tried to feed the full-link data of 60 vending machines to the AI algorithm to generate distribution suggestions. Its transaction volume increased by 15%, and the number of replenishment times decreased by 18%.
11. Art/design field
(1) 3D model generation, AI generates 3D design models according to design requirements, lowers design thresholds, improves personalization, reduces labor costs, and improves design efficiency. For example, Nvidia Magic3D AI can generate 3D models in one sentence
(2) AI art creation, AI assisted art creation, provides inspiration for art creation. Improve creative efficiency and provide various art forms
source
Generative AI and Strong Artificial Intelligence: Explore the future of AI technology
AIGC represented by ChatGpt, that is, what exactly is generative AI?
Generative AI - AIGC landing application field combing