Can AI make money?

Various large models have been released intensively recently, and views such as "catching up with GPT4" and "becoming China's OpenAI" are scattered in various articles. Let’s get back to the most fundamental question: This is really a soul torture for people in this industry. Without the foreshadowing of the past ten years, always asking this question seems to be a bit eager for quick success, but with the foreshadowing of ten years of losses, whether to make money or not becomes a question that combines technology and business: that is, technology The touchstone is also proof of commercialization capabilities.

Before answering this question, we need to summarize the potential business model behind AI.

1. Potential business model behind AI

If AI develops to a very mature stage, what are the potential monetization methods? There are not many business models that have been explored and proven in the past, and there are only a few that match AI:

1. Subscribe

This is almost the most typical monetization path for AI now. In fact, it is a type of cloud service. Major cloud vendors naturally put self-developed AI functions into their own cloud product matrix.

2. New value-added services

The "Her" in the movie is actually a new type of value-added service, and electronic pets that can communicate in the future are also in this category. Compared with one, the difference is that this type is the final dish, while one is the raw materials. There will be a lot of overlap between the two, but it is roughly the same as the difference between PaaS and SaaS. In the past, the SaaS we always talked about was closer to new value-added services, so we don’t list them in a separate category, such as various enterprise-level assistants.

3. Hardware product sales

This kind of final operation is similar to Lenovo selling computers. Large models in the multi-modal direction need this support. Without the success of large-scale new smart products such as robots, smart speakers, and AR glasses, the success of large multi-modal models is unlikely. In the case of industrial division of labor, this model will be superimposed on the previous two and become the driving force for the first two.

4. New advertising

Some people said before that large models will make search ads difficult to display. I don’t think so at all. The screen is so big that it can even be used as a recommendation: If you want to buy it, please see…. The key is to increase the frequency and accuracy.

5. Solution Sales

Products like Watson are unlikely to become completely standard products. They always have to be connected with various surrounding specific situations, and solutions are inevitably needed to connect them. From a technical point of view, it seems similar to 2 and 3, but from a business model point of view, the difference is so big that it needs to be listed separately. New value-added services and hardware products are still sold as standard products, and the price per customer is the upper limit of iPhone or Vision Pro.

But the solution is not. The unit price here must be very large, such as tens of millions, to be valuable, otherwise it will not be able to support the long-term investment in the early and late stages. To a certain extent, the AI ​​of existing products will actually become solutions, such as e-commerce, short videos, etc. This is especially true for large industry models. AI here is not a disruptive force, but will appear as a solution to enhance existing products.

6. Games and Metaverse

This looks like a product, but the big difference from the first to the fifth is that only this product supports the virtual central bank model. The virtual central bank model means that you can directly issue your own token (not necessarily a digital currency). Only such products support a separate ecological and monetary system.

If you cut these business models vertically, there will be two obvious common characteristics:

1. As mentioned in the collection of lies, no one can escape,

   AI is actually a deep-well model, which is reflected in the update of existing models (including people). As a result, creating new models is actually not as good as Internet, but the impact on existing models will be greater than that of the Internet.

2. The economic value of technology will reflect more of a kind of personification, doing what people are doing in the economic system and making transcendence (various assistants, etc.).

These two points are very critical because they directly affect who will make AI money and the potential endgame.

2. Who will make money from AI in the end?

The above characteristics of AI determine that it is actually a link in the supply chain. In this way, in addition to 1, if the corresponding company wants to overcome its own model, it must become the kind of company it is not yet.

For example, even for relatively light games and the metaverse, it means that large model companies need to force themselves to become companies that understand games and the metaverse.

This leads to two further questions:

1. If it is in the form of cloud service 1, will AI be a separate cloud service or a part of existing cloud services?
2. If it needs to be combined with an existing field, will it be dominated by pure AI companies or will companies native to the field evolve and dominate?

For the first question, I think the answer is relatively clear. Things like cloud cannot exist as a separate category and will definitely be merged. This is determined by the inherent scale effect in asset-heavy industries.

As for the second question, the answer is actually relatively clear. In different fields, the weight of the field and the weight of the technology are different. For example, the weight of the game field is low, and the weight of the tax and medical field is high. The higher the weight of the field, the less likely it is for a single technical AI company to dominate. The reality is that most of the time the domain weight is high. It's hard to say which company is specific, but the ratio of this kind of technology and domain knowledge will be more critical.

3. Business challenges of making money with AI

The previous article about pondering things looked at the commercialization process more from the attributes of technology. This time, I looked at it from the perspective of pure business model. The conclusion is not complicated:

The business challenges of pure AI companies are very obvious. If they stop in the supply chain, the path to monetization will be too narrow. If you want to get through the last section by yourself, you not only need to get the model, but also the product (the product represents the integration of domain knowledge and technology).

It seems that the future will be like this: if the top large model companies cannot open up other monetization channels and are limited to the supply chain, they are more likely to be merged by large cloud companies. The middle domain model is more likely for companies with domain knowledge to gradually complete their own evolution and win. For example, a publishing company that makes content review products has a greater chance of success than a simple layman.

All business model analysis and judgments require a foundation: the technology itself must be sufficient to create value. So what is the technological maturity level? Is that enough?

Is the time coming? What is the technological maturity level?

Enough is really not enough. As long as you open up a real scenario to make products, you will find that the technology supply is still insufficient, and it will not be completely sufficient in the short term.

The journey of AI technology from 0 to 1 has actually never been completed.

This is a big difference between AI and previous technologies such as the Internet in terms of technology itself.

To a certain extent, many of the basic technologies of the Internet around 2000 are actually there, and the rest is a faster and larger scale improvement. (The basic Internet protocols we are still using now, such as TCP/IP and HTTP, are technologies of different ages)

But AI is not the case. Its foundation is constantly improving, and all corresponding applications need to be improved and applied at the same time.

Comparing the two, you will find that technically the Internet is one step at a time, and AI is almost constantly growing while deceiving itself. We announced today that we have solved this problem, and tomorrow we announced that we have solved that problem, but the progress until the big model is far from expected (and the success of the big model will come from a company whose original founder does not have an AI background, and there will be a bit of magic realism).

So if you look at it from the perspective of enough and not enough, it is really not enough. But even if AI is not enough, it is like water. The submerged part will be completely changed, and the corresponding functions will be completely changed. For example, when making pictures, no one will find someone to draw basic pictures anymore.

How to judge the technical maturity of a product if it is not enough? Or is it enough?

In fact, you can use the full scene coverage method. From a business perspective, it can only be the full scene coverage method. As mentioned earlier, AI business channels will always show a kind of anthropomorphism. If people live in a certain relationship and cannot handle the corresponding comprehensive relationship, it will not be anthropomorphic enough.

4. Full scene coverage method

It is easy for AI to use technical indicators to measure itself, but this will become an involution routine, and in the extreme it will deceive itself.

AI essentially tests general capabilities. If a specialized method is used, it can theoretically do extremely well on any test set, surpassing any of the best existing artificial intelligences. But this is of no use other than writing PPT. Because when AI is implemented, anthropomorphism causes any scene to be deeply intertwined with complex environments, which still requires general capabilities.

This kind of technology-specific evaluation method actually constitutes the fundamental reason for the hot and cold situation: on the one hand, AI seems to be able to do everything, which is already extremely miraculous; on the other hand, it is not easy to use, and if it is not easy to use, it will not make money. .

What is the full scene coverage method?

To put it simply, for example, recruitment is a scenario. Does the technology supply support directly create a digital employee to perform all the functions of recruiters in the past, such as recruiting people back based on a demand, without human intervention in the whole process?

If this thing cannot be realized, then except for the first mode, none of the subsequent high-value modes will work.

This is a real challenge.

5. Summary

If you think about it carefully, several new fields in the post-Internet era actually have their own setbacks. If we go back to around 2015, there are probably three new directions that have emerged: one is artificial intelligence, one is blockchain, and the other is SaaS. Many students who feel that the Internet has basically come to an end and are unwilling to be lonely go to these three directions. Then artificial intelligence and SaaS continued to be unprofitable for 10 years, while blockchain was profitable but almost disappeared due to other reasons.

Now the big model seems to be able to inject new power into these three simultaneously and is approaching the final moment. Every time I see a picture like the one below, I believe it more:

Many students will pay attention to which field will come first. This really cannot be concreted, but there can be a basic judgment model:

Determine how far the path from new value created by technology to commercial value is. Midjourney is actually the short one, and Watson is the long one. To really do things, funds and manpower must match the length.

Thinking about related articles:

AI big model has no business model?

Columnist

Thinking about things, WeChat public account: Thinking about things, everyone is a product manager columnist. Vice President of Sound Intelligence Technology. He is the author of books such as "Ultimate Copy: How Artificial Intelligence Will Promote Great Social Changes", "Perfect Software Development: Methods and Logic", and "7 Tipping Points in the Internet Era".

This article was originally published on Everyone is a Product Manager. Reprinting without permission is prohibited.

Title picture comes from Unsplash, based on CC0 protocol

The above is the detailed content of Can AI make money?. For more information, please follow other related articles on the PHP Chinese website!