Nowadays, everyone is very interested in the technical strength of large models. However, Robin Li holds a different view, which is really thought-provoking.

Many people focus on technical strength, but often ignore whether enterprises can improve key business data with the help of large-scale models. This point is easily overlooked, but it is actually crucial.

At present, the technical capabilities of large models are talked about by everyone.

A lot of manpower and material resources were used in the research and development of technology, and various giant models came out one after another.

Often, both companies and ordinary people seem to be more inclined to focus on the pros and cons of technology and whether it is advanced enough.

However, there are drawbacks to this excessive focus on technology.

From the perspective of social resources, my country's computing power is limited, and many companies have devoted themselves to the research of basic large models. Phenomenons such as the "Battle of 100 Models" are common. However, this phenomenon not only consumes a large amount of social resources, but also most projects do not bring actual benefits.

In reality, many companies simply pursue technological leadership, but rarely consider whether this technology is really suitable for their own business needs.

Many large domestic AI model companies participate in scoring competitions. They feel that high scores are enough to demonstrate their capabilities.

This third-party benchmarking result is not trustworthy. Its main purpose is to attract investment rather than truly meeting the actual business needs of the enterprise. Therefore, its real value to the development of the large model industry is limited.

This type of enterprise is of moderate size and has a technological background, and they show a high degree of maturity in using large models for application development.

As Robin Li mentioned, the maturity of these companies in this regard is admirable.

But most businesses?

Many companies have not really made in-depth use of large models.

They have never thought about the match between the large model and the core of their business, nor have they studied how to use the large model to bring benefits to the business.

This has caused large-scale models to remain at the theoretical level, just a means for companies to show off, and have not truly become a driving force for corporate development.

Robin Li often emphasizes that having millions of AI native application models based on basic models is of extremely high value.

Once the basic technology reaches a high and leading level, large models can quickly produce valuable results in multiple scenarios.

Enterprises need to deeply integrate large-scale models with specific business scenarios to create application models that meet their own business needs.

Focusing only on basic large-scale models and ignoring the construction of application models is like having a powerful engine but lacking a matching body. The result is that it cannot operate properly.

But in the country, the current emphasis on these native application models is far from enough.

Chinese companies have deficiencies in the application of enterprise-level models or models in small languages.

Chinese companies currently do not have the energy to do specialized optimization work.

In the context of global competition, this may cause Chinese companies to fall behind in this regard.

Many foreign companies have performed well in using enterprise-level models. They deeply understand their business and make targeted adjustments to the model, which not only improves efficiency but also reduces costs and enhances the competitiveness of the company. We still need to work hard to catch up in this regard.

Robin Li emphasized that if the core indicators of the business cannot be substantially improved through large models this year, it cannot be regarded as truly accepting and applying large models.

Enterprises need to pay attention to how large models directly affect their business.

Taking Baidu as an example, Robin Li pointed out that employees have questions about "how does this technology relate to my business?" This exposed misunderstandings about large-scale models within the company.

The large model is not independent of the enterprise's business, it should serve the efficient growth of the business.

For example, the model application layer can reduce inference costs while maintaining the same effect, which is the key advantage of future large-scale models in the competition in the enterprise field.

Robin Li pointed out that in the era of large-scale models, large enterprises will obtain most of the benefits. It should be noted that the “big manufacturers” here are not limited to companies in the Internet industry.

In this era, companies with strong strength and abundant resources do have more development opportunities. They can quickly carry out large model-related business and integrate large model technology into actual business more efficiently.

Even a large-scale enterprise like Baidu has certain limitations in the field of large models. No one company can monopolize all the opportunities in the era of large models.

This shows that other companies can also flexibly determine their own positioning and use large models to develop businesses or applications that suit their own needs.

Generally speaking, Robin Li's view that "Baidu will build a large Wenxin basic model, and other companies can develop applications on this basis" does have some truth.

Now we have to pay attention to the value realization of large models in practical applications, and we cannot just limit ourselves to technical discussions.

Dear readers, does your company, or companies you are familiar with, have any special methods or experience in using large model technology?