Home Business Those false opportunities in AI implementation

Those false opportunities in AI implementation

Those false opportunities in AI implementation

What is the first priority of an enterprise? In fact, it is extremely numerical economic rationality.

If you take away all the decorations, you will definitely come back here. If we compare an enterprise to a living body, these are the physiological and safety needs that Maslow talked about. Further advancement above this can certainly give it some kind of mission and meaning. From a primary perspective, the definition of entrepreneurship is to realize the leap from use value to financial value as imagined by ordinary people .

It’s just that the gap in the middle is so deep and wide that no matter how many companies fall in, there is no sign of it being filled. In many cases, failure is not due to insufficient efforts, but because the imagined use value is inherently incapable of crossing over to the other side. An opportunity that does not have value beyond the possible is a false opportunity.

Fake opportunities can be further classified. For example, they are useful for some large companies, but have no entrepreneurial value. In this article we will talk about a few typical fake opportunities.

01 Supporting value

From the perspective of use value, each subcategory, from acoustic microphone array algorithms to perceptual speech recognition to cognitive large models, creates value. However, such algorithms are often only usable in special scenarios and have only supporting value. , is a link in the supply chain. This type of scenario is usually not suitable for starting a business.

When a company becomes a pure supply chain link, what needs to be managed is not only its own technology and products, but also its position in the entire industry chain. Your position and voice determine your living space. At this time, Porter’s five forces model is most suitable.

It can also be simplified further: use value x pricing power = financial value.

In the past, this path was not without its successes.

For example, Dolby is a de facto industry standard and an independent brand. Its pricing power is obviously greater than its pure technical value.

For example, the often mentioned LiDAR, Knowles’ microphone with algorithm, etc., they polish scarcity and barriers at the point of fusion of software and hardware (scarcity supports pricing power). To a certain extent, the hard part is always more important than the soft part.

This style is really not that suitable for AI, because the core of AI is still a software algorithm.

In the value chain, the weight of algorithms that AI personnel are good at is really low. This does not mean that this kind of thing cannot be done, but it means that it is not a good choice for AI entrepreneurs.

Algorithms that simply have supporting value have two fates: one is to become part of the public cloud and become a completely undifferentiated commodity, and the other is that the most important customers will become particularly critical or the cost is too high. Choose the company to conduct self-research. (Still the five forces model)

We can summarize this model like this: the value of the algorithm may be large or small, but the pricing power is very poor, so it is not a good entrepreneurial opportunity. It is easier for purely technical personnel to find some work and make some money.

02 The upper limit of water temperature is 90 degrees

Putting a lot of technology into C-end products is another situation.

It can indeed serve as a support for a core new experience, but different scenarios have different so-called usable lines. Different availability levels correspond to different business values . This is the fundamental difference between orange cats and little tigers.

It is easiest to see the situation clearly with the smart speakers that have been relatively popular in the past ten years.

Have smart speakers crossed the usable line?

In fact, it has been crossed, but its upper limit is lower than imagined in the past. It is neither a general computing platform nor a new entrance. It’s like a gas stove with or without a screen.

Before the emergence of big models, the limitations of technology meant that it was still speakers rather than personal assistants and robots. This ceiling determines the development trajectory of the most representative AI product in the past decade. (Unlike the above example, this one has problems with the use value itself )

The wonderful thing about this type of product is: assuming the technology is mature enough, this large category will become a market with fierce competition among giants; if the technology is immature, it will still be a traditional category.

Toynbee’s summary of civilization has an interesting point: the surrounding areas are not effective if they are too intense or too comfortable, but should be moderately stimulating. This is key to establishing the product in this direction: speakers are too cool and difficult to live; computers are too comfortable and have no activity and new opportunities. On the contrary, new categories that seemed marginal in the early days, such as drones and sweepers, have been given the opportunity to develop and find their own broad space as the market expands.

In addition, if you really want to make this type of product, you need to make it clear from the start that you are positioning yourself as a C-end consumer product company such as Xiaomi and DJI, and you must establish your own brand through opportunities in new categories. We need to focus on brand channels, not technology.

This is a hard application without network effects. It just needs to do things well with traditional consumer goods. It is very taboo to always think about back-end monetization. Because the early team has so much energy and is always worried about back-end monetization, it can easily lead to poor performance of your front-end products.

Under this logic, although technology is critical, it is still a part. But when AI personnel create such companies, it is easy for technicians to manage the whole. This is very troublesome, and it is equivalent to the part being greater than the whole. If it actually needs to be transferred, the weight of the product is greater than the technical point. It is the product that drives the research and development rather than the transfer. (Just look at Apple for this)

To sum up, for such a complex direction that requires heavy investment, if the upper limit of water temperature is 90 degrees, then the efforts will be in vain and the results will be very unlikely to be good.

03 Management and operation are the most important

It can also be called project system.

Because the project system is the easiest to capture user needs, it is the easiest to run out of cash flow.

If you look carefully, you will find that Kazuo Inamori’s Kyocera actually did a lot of customized project work in the early days.

The trouble with the project system is that it requires running many business units and at the same time providing certain unified services. Some places are suitable for separation, and some places are suitable for integration. (Amoeba and Zhongtai deal with the issue of division and integration)

AI technology is generally used in parts that need to be integrated together.

At this time, the replicability of the output of the department responsible for AI technology is critical. If the supply here is insufficient, it means that no matter how fancy it is, the final thing it grabs is a CPU. A fixed CPU has such a large amount of calculation. Businesses inherently cannot scale.

Assuming that the bandwidth is enough, what will happen?

This will become a task of meticulous management and operation, which will be more troublesome.

(In real management and operations, there is no problem that needs to be solved with digitalization. It is essentially the same thing, and the success rate of digitalization is basically a narrow escape)

Management and operation is something that everyone thinks they are good at, but in fact the so-called skills are all below 60 points.

Ning Gaoning is actually right to go back to textbooks, but one can also understand with the most normal thinking: the core competitiveness in the early stages of entrepreneurship becomes management and operations, which is definitely not reliable for most scenarios. It’s a bit like Iron Man, who is physically weak, but his armor is too heavy, and he is crushed.

There are indeed several types of scenarios that will intensify the importance of management and operations to the point where it cannot be increased , such as project systems and long-chain operations. Long-chain operation means that pre-sales, products, software, hardware, supply chain, customer service, software scoring algorithms, cloud, front-end, client, testing, etc. need to work closely together to promote the continuous amplification of the business spiral.

If this kind of work is not paid well (mechanism construction), then no matter how many people are added, it will be in vain. The more people are added, the worse the effect will be.

There are three essential ways to solve this situation:

One is to avoid it. For example, if you make WeChat products, that’s fine. Of course, when WhatsApp was acquired for US$19 billion, it only had 30 to 40 people working on one product, so that was easy.

One way is to face it head-on. If you do this well, you will be truly invincible. Competing with other companies is a lot like Terminator versus zombies. When it comes to Huawei, people often talk about technology, but technology is the result. Huawei’s success is actually the success of the organization and mechanism. It’s just that most people really can’t compare themselves to Mr. Ren Zhengfei.

One way is to use carbon-based algorithms, but there is currently no general product that can really solve this problem. This is possible. If Douyin, MCN, and anchors are regarded as a company, this is undoubtedly the most successful realization of Amoeba, which is a high degree of unity of vitality and rules.

To sum up, under this model, the realization of use value can be realized but it is difficult to control the cost of realization. Coupled with the constraints of pricing power, it is like needing to wring out the water from a relatively dry towel, which is what most entrepreneurs are good at. It’s a departure.

04 Capital-heavy game

There are also some opportunities that are indeed new opportunities, but in fact the very large weight is a game of capital. This is not to say that technology is not important, but that it has become a complex of big capital and high-end technology. The most typical one is the large model. From a capital perspective, this field is actually no different from building railways and roads in the past, but it requires an extremely high density of talent. There are really only a very small number of people who are qualified to do this. (OpenAI has 200 people doing technical work. In the end, there will definitely be no need for 100 OpenAIs on the planet, so there won’t be 20,000 people in total. Most of them will just join in the fun)

In the past, there was a chicken head and a phoenix tail, but according to the digital space mentioned above, it will have the only characteristic of name and reality, so in this category there is actually no chicken, only phoenix. The chicken will soon be tortured to death. Furthermore, this type of product is a classic example of big wins and big losses. The average person should keep an understanding of its application characteristics and potential, but can actually stay away from this field. (The ones who are most likely to struggle are the R&D personnel who are close to the point. At this time, the wisdom and courage to retreat bravely are particularly tested)

To sum up, this type of field is not suitable for using the use value and pricing power mentioned above. The essence is that the threshold is too high and it cannot be done. The welt is suitable for self-dimensionality reduction.

05 The password in the AIGC name

So where do new opportunities come from? From another perspective, where do new categories come from?

Looking from the outside in, the first thing to do must be a new category, and I think the origin of the new category is in the name AIGC. This most primitive word actually has a code.

What needs to be considered is how to combine and superimpose the content and scenes of GC (Generate Content).

Ignoring the size of the problem, products with some success in the past were all centered around this, either directly as tools for content generation, or as packaging of generated content.

Any category that relies heavily on GC must be a new category, because this was not possible in the past.

On top of this is the integration with the scene. In this process, it is necessary to cut off the illusory branches of AI that are generated in the movie. (I’m trying it myself, but it takes some time. I’ll talk about it later when there is progress.)

06 Summary

Eiichi Shibusawa during the Meiji Restoration era wrote a book called “The Analects of Confucius and Abacus”, which talks about the ancient Confucian distinction between justice and benefit. Of course, we can combine justice and benefit to give enterprises a higher dimension. Value, the organization has a higher pursuit, not only to solve the economic reality but also to make it meaningful. This can indeed go higher and further, otherwise it will be difficult to continue to create products like chatGPT. But this article is not about that, but the most basic steps.


Please enter your comment!
Please enter your name here