Observation丨What AI really changes is the way humans work and think

Source: BundBoard

Author: Bund Board of Directors

Original title: "We will eventually meet humanity in AI丨The Bund scene"

From June 28th to 30th, the CEOs of the Bund Board of Directors went to Beijing and had three key conversations related to artificial intelligence.

Facing artificial intelligence, many founders are both excited and nervous.

The excitement is that the rapid growth of artificial intelligence will bring many new opportunities and create new innovation models and organizational structures.

The tension lies in what unknown conflicts and challenges will arise from artificial intelligence and current organizations and businesses.

Fear cannot hinder the pace of entrepreneurs going to artificial intelligence.

Compared with specific technologies, costs and functions, entrepreneurs are more concerned about the essence behind it, whether artificial intelligence will shake the existing business system and existing organizational values.

Example card description: The following is the content of the learning experience after the dialogue, not the original text shared by the dialogue guest himself.

01. Principles and values are the most important things for a large model

At the first stop, we came to Zhipu AI, and discussed with COO Zhang Fan how enterprises embrace large-scale models.

Many people have some misunderstandings about the large model. Some people think that the large model is a wishing well, and as long as they make a wish, it can be realized. When they find that the large model cannot answer every question and respond to every request, their belief will collapse.

Some people think that the big model is just a big intelligent database, just feeding data crazily, expecting the big model to generate thinking ability and judgment ability, but in the end they will find that the big model is still stupid.

To create a handy large model, it must be taught the corresponding judgment criteria and principles.

This is like when employees join a company, they need to know what they can say and what they cannot say, what they can do and what they cannot do. The framework of these judgments and reasoning is much more important than the data itself.

The same is true for data feeding. What you feed the big model are the behavior patterns and judgment patterns of sales champions, and it will gradually learn to become a sales champion; if you feed it ordinary sales data, it will become an ordinary sales champion. Sale.

It is equivalent to that the large model is a blank sheet of paper, with learning ability and judgment ability, which needs to be adjusted by the enterprise to be what it wants.

** Therefore, the clearer the requirements, the clearer the values and judgment criteria are, the lower the trial-and-error cost of the large-scale model is to create a large-scale model exclusive to the enterprise. Whether it is other technologies or costs, it has gradually matured and is not difficult to realize. **

02. The large model is not a tool, but a partner to grow together

The second stop came to Baichuan Intelligence, where we discussed the nature of artificial intelligence and how we should treat artificial intelligence with CEO Wang Xiaochuan.

We asked Wang Xiaochuan a question, which occupations or industries will be replaced by AI?

His answer is that most people sitting in front of computers will be replaced by AI. **

This answer can't help but make us think that what the big model really threatens is not the blue-collar class, but the white-collar class.

The tasks that the Internet can reach can basically be achieved through learning. Perhaps at the moment, the large model is a good helper for white-collar workers and can help us improve efficiency. But the ability of the large model is not limited to tools, it is more like a living body, it will adjust and correct its own judgment.

For example, if we want to recruit a person, I will ask the big model what conditions are needed to recruit such a person, and it will tell me, and it can also help me publish recruitment information and screen resumes that meet the basic requirements.

The judgment of the big model is based on big data, not the limited experience of each of us, and the conclusions drawn will be more objective and rational.

The shape of the organization is reversed again. During the industrial revolution, we accelerated development through fine division of labor, but now we have to go back and do not need division of labor. We let large models do it. There are fewer and fewer large companies, and many small companies with individuals + AI will emerge.

With the help of GPT, everyone is not an employee in their original position, but can use GPT to help realize their personal dreams. The society will become flatter, and everyone can live more self-consciously and find the meaning of their true existence.

03. What is a good large model depends on our judging criteria

At the third stop, we came to Baidu Smart Cloud, and had an in-depth exchange with Shen Dou, Executive Vice President of Baidu Group and President of Baidu Smart Cloud Business Group.

Baidu Wenxinyiyan entered the field of large-scale models in China earlier, and they pay more attention to the application of large-scale models in various business scenarios and industries. They believe that the most important indicator of the best large model is that it is easy to use, rather than blindly pursuing the ultimate in various technical indicators.

In large business scenarios, one of the biggest concerns is whether the large model technology will bring uncontrollable information security during the application process.

If the content generated by the large model is directly transferred to C, it will bring a lot of uncertainty about information security. Therefore, the model of people + big model is an inevitable development process.

The big model helps people to provide efficiency, and people to ensure the information security of the big model. If problems are encountered, specific people can also be held accountable and followed up.

The development speed of the large model is based on the development speed of the hardware. Now chips are being innovated from generation to generation, and computing power is getting faster and faster, which has laid the foundation for the development of large models.

In the future, the cost of hardware usage will become lower and lower, and the era of large models will reach a certain level, just like the era of mobile Internet, with various innovations and explosive development.

At present, any problem that can be solved with language and text, AI can greatly improve the efficiency.

For example, when writing an English news report, or planning a travel route, AI can do better than humans.

Therefore, it is worthwhile for us to use AI to redo every industry. In this process, the most important reconstruction is not the reconstruction of business and process, but the reconstruction of people's thinking. **

What AI will really change is the way we work and think. And at that time, the era of big models will come.

Write at the end

This time, the CEOs met with people who know the big model best, and discussed the most practical enterprise landing scenarios. The most clear message they got was that **AI can help people, but AI cannot completely replace people. **

The real demand of CEOs is that the tasks can be completed faster and better, and the cost can be lowered.

It is not very important whether this requirement is completed by manpower or by AI.

The current positioning of artificial intelligence is more like a reserve of intelligent talents, requiring companies to invest time in training him, telling him what he can do and what he can't do, so that he can achieve the desired effect.

The KPI of each middle and senior executive should be from cultivating a reserve cadre to cultivating a reserve cadre and an artificial intelligence cadre.

Compared with the simple and rude all-staff AI office, how to assign suitable jobs to artificial intelligence may be the most important ability of CEOs in the era of large-scale models.