In the age of AI, model size is not a decisive factor

Three quarters have passed this year. For the technology industry, unless the feasibility of "normal temperature superconductivity" is officially confirmed, the most popular technology vocabulary this year is not "generative" "Artificial Intelligence" is none other than

In the current technology industry, if a technology company does not dabble in large-scale models and generative artificial intelligence, it seems to have fallen behind in the technology competition

According to a recent release A survey report by the IBM Institute for Business Value shows that three-quarters of CEOs surveyed believe that deploying advanced generative artificial intelligence will bring competitive advantages to enterprises, and users have a positive attitude towards this

In People While IBM is enthusiastic about large-scale models, it has chosen a different new path

"Model" is not bigger, the better

In the past 8 months, various large-scale models Emerging quickly, like new shoots in spring. Whether at the consumer or enterprise level, all walks of life are actively embracing generative artificial intelligence and enjoying the dividends and conveniences it brings. However, for enterprises, is a larger and more complete large model really the best choice?

Rewritten content: In recent years, companies in various industries have been seeking a way to "reduce costs and increase efficiency". Although large-scale models can significantly improve efficiency in the initial stage, in the long run, the consumption of computing power by large models and the cost and time of subsequent expansion are important issues that enterprises need to consider. Therefore, I believe that enterprise-level large models need to be "small and precise". Only data related to the vertical field of AI enterprises must be provided, so that AI can truly achieve "specialization in the industry" and achieve the greatest value at the lowest cost

IBM and Xie Dong, Chief Technology Officer and General Manager of IBM Greater China R&D Center, expressed the same view at the recent IBM Watsonx Greater China press conference. Xie Dong said that for enterprises, the goal of application models is to solve specific problems at a lower cost. He pointed out that for enterprise-level applications, the smaller the model, the better, because small models are more flexible and lower-cost

In order to meet the urgent needs of enterprise-level users for generative artificial intelligence, IBM hopes to use the enterprise's own business needs and data, tailoring generative AI solutions and models to them. Recently, IBM officially released the IBM WatsonX platform in Greater China

From “AI” to “AI”

AI was born as early as the 1950s, but until this year, the industry More applications are "AI", which uses AI technology to empower a certain technology or certain fields. Starting this year, the future will enter the "AI" era with AI first. Xie Dong said that companies now hope to apply AI technology to core businesses to enhance actual productivity. The entire industry will also move from the data-first "AI" era to the AI-first "AI" era

IBM is an important player in the global AI field and has been in this field since the early days of AI. Deep cultivation. As early as 1956, IBM used AI technology to successfully realize a human-machine battle in checkers. Then, between 1996 and 1997, IBM's Deep Blue computer successfully defeated the top chess players. In 2011 and 2019, IBM achieved a qualitative leap from AI knowledge accumulation to AI debater

As a pioneer in the industry, IBM launched the Watsonx platform at this moment. Although it seems a bit "backward" in the industry, after carefully analyzing the capabilities of the Watsonx platform, we will find that it leads generative artificial intelligence from C-side users to B-side, redefining the application of generative artificial intelligence at the enterprise level. The role in

I believe that the true core value of a digital technology lies not only in its application in the consumer field, but also in its promotion in enterprise-level applications. Only in this way can the true value of the digital technology be realized . At the current time point, the launch of the WatsonX platform has undoubtedly opened up a new path for generative artificial intelligence in enterprise-level applications. With years of deep experience in the field of artificial intelligence, and taking hybrid cloud and artificial intelligence as the concept of future development, in this hybrid cloud competition, the WatsonX platform is the core, and artificial intelligence will become the core driving force of IBM in the future.

Not just a "model"

It is worth mentioning that compared with the large models of various generative AIs currently on the market, watsonx is not just a model, but A partially open source and open platform. Miao Keyan, general manager of IBM's Greater China Technology Division and general manager of China, said that the IBM watsonx system is a new generation AI and data platform based on an open hybrid cloud architecture, based on basic models and generative AI

Miao Keyan emphasized that watsonx is a collection of innovative technologies from IBM Research Institute, integrating the most advanced enterprise technology and open technology - OpenShift, and is supported by the powerful open ecological community - Hugging Face. Miao Keyan said, "'x' represents unlimited possibilities and also represents IBM's expectations for watsonx."

The IBM watsonx platform is divided into three product sets - watsonx.ai, watsonx.data , watsonx.governance, currently, watsonx.ai and watsonx.data have been launched, the premise version of watsonx.data is now available to Chinese customers, and watsonx.governance is planned to be launched in the fourth quarter of this year.

Rewritten content: Among them, watsonx.ai can help AI builders use models from IBM and Hugging Face to complete various AI development tasks. These models are pre-trained to support a variety of natural language processing (NLP) type tasks, including question answering, content generation and summarization, text classification and extraction. Future releases will provide access to more proprietary base models trained by IBM to improve efficiency and task specialization in related areas

In the current era of "data is king", the IBM Watsonx product family Watsonx.data can help enterprises deal with the data challenges of large data volume, high complexity, and difficulty in governance that are commonly faced when AI workloads expand. It is worth mentioning that Watsonx.data also allows users to access data across cloud and local environments through a unified portal

Xie Dong said that later this year, watsonx.data will leverage the foundation of watsonx.ai Models to simplify and accelerate how users interact with data. In this way, users can discover, enhance, optimize and visualize their data and metadata through natural language conversations

Xie Dong said that watsonx.governance will be launched later this year. This product is an automated data and model lifecycle solution for developing strategies, assigning decision rights and ensuring organizations are accountable for risk and investment decisions

watsonx.ai, watsonx.data and watsonx.governance, can Known as the "Troika" of IBM Watsonx. Through these three aspects, enterprises can achieve digital transformation from multiple perspectives such as artificial intelligence creation, data management and enterprise management. In my opinion, IBM watsonx is more than just a "model". As a generative artificial intelligence platform, it has the potential to become the blueprint and dream for future enterprise-level artificial intelligence applications

The above is the detailed content of In the age of AI, model size is not a decisive factor. For more information, please follow other related articles on the PHP Chinese website!