Explore WAIC | The Fourth Paradigm's 'Style Theory' focuses on toB large models and lays out generative AI to reconstruct enterprise software

The 2023 World Artificial Intelligence Conference has kicked off. The large model of the Fourth Paradigm "Shishuo" was displayed on site at WAIC.

In addition, the Fourth Paradigm "City of Big Models" brings together the most representative application practices of big models in finance, retail, real estate, aviation, manufacturing, justice and other industries.

Focus on toB large model, lay out generative AI to reconstruct enterprise software

Different from the AIGC that everyone knows, such as generating pictures, generating posters, and generating copywriting, Fourth Paradigm focuses on large model technology in the field of enterprise software and proposes the technical strategy of AIGS (AI-Generated Software): using generative AI restructures enterprise software.

According to reports, in the enterprise software transformed by "Shishuo", no matter what function is called, the result can be obtained through a simple dialogue, without the need for layers of clicks. In addition, by adding or modifying data/interface rules, the software function upgrade efficiency is shortened from one month to one day.

According to the official introduction, "Shishuo" is positioned as a multi-modal large model in the ToB (enterprise-level service) field, with core technologies such as input and output multi-modality, knowledge base, Copilot, and thinking chain.

At the same time, "Shishuo" provides enterprise-level guarantees to ensure content is credible, costs are controllable, and data security is guaranteed. Under the guidance of "Shishuo", enterprises can privatize and deploy large-scale models to obtain the required computing power at a relatively controllable cost. Secondly, "Shishuo" generates more professional content after learning industry knowledge, and locates the original source when outputting information, so that all information is "well-documented." Finally, "Shishuo" can learn and correct mistakes based on user feedback.

"Big Model City" reveals the results of industrial implementation

Fourth Paradigm said that since its release, the "Shishuo" large model has been exploring the implementation of large models with hundreds of domestic companies, and has accumulated dozens of early domestic AIGC applications. "City of Big Models" showcases the real practice of combining large models into different industries.

For example, in the field of retail store management, store management usually involves employee management, product procurement, inventory control, handling customer complaints, etc. Store managers need to obtain data and complete management tasks in scattered systems, which is extremely time-consuming time. Store management after the transformation of generative AI can directly ask questions to the store management assistant, such as "check product sales, duty personnel status, illegal operations", and the assistant can call relevant functions from different systems and directly provide Required information.

In the field of real estate intermediaries, real estate agents require specialization and standardization. For example, in second-hand housing transactions, it involves more than 20 links such as house purchase qualification review, property rights verification, down payment and fund supervision. The training of agents has always been is a difficult problem in this field. The real estate agency model uses "Shishuo" as the base of capabilities, and uses the knowledge base, standardized operating rules and communication skills accumulated by I Love My Home over the past 23 years as the corpus to train vertical models for specific industries and fields.

In the field of industrial design, a commercial aircraft manufacturing company is using generative AI to transform industrial design software. The operation of traditional industrial software is extremely complex. Even basic functions such as digital analog query and assembly require remembering specific parameters to complete professional tasks. operate. With the "Industrial Design Software Assistant" based on generative AI, you only need to ask the assistant voice questions "Help me find similar parts" and "Give me the assembly plan for these two parts." AI can automatically find it from a large number of parts. Reasonable assembly plan, what you ask is what you get.

In the field of production operations, the MOM system (Manufacturing Operations Management System) is known as the "brain" of smart workshops and digital factories. The semantic understanding capabilities of generative AI enable people to efficiently communicate with computers on the core business of manufacturing operations, providing A series of information and functions such as knowledge answering, data and graphic query, and report generation. In addition, it can also be connected with the factory's terminal equipment (monitoring, cameras, access control, etc.), and perform "factory administrator" functions such as workshop safety, illegal operations, and equipment status maintenance.

In the judicial field, judicial personnel must not only master relevant laws, regulations, and judicial interpretations, but also flexibly apply these in specific practices such as legal analysis, legal retrieval, legal writing, and legal argumentation. Generative AI can quickly locate and retrieve information in numerous legal documents, and can also write and proofread corresponding legal documents. In addition, for common cases and quick-trial cases, we can assist in extracting the key elements of the trial.

More importantly, through high-frequency communication with employees, these AI assistants have recorded a large amount of process data such as behaviors and thoughts that cannot be captured in traditional systems. Through continuous thinking chain learning, each professional role has been gradually formed. "Best Practice Guide".

The above is the detailed content of Explore WAIC | The Fourth Paradigm's 'Style Theory' focuses on toB large models and lays out generative AI to reconstruct enterprise software. For more information, please follow other related articles on the PHP Chinese website!