How to calculate the valuation of artificial intelligence companies, here are several formulas

AsiaIndustrial NetNews: The Internet line will continue to develop, and what AI does is more inclined to the bottom – improving production efficiency on the production side.

How to calculate the valuation of artificial intelligence companies, here are several formulas

Before joining Fengrui Capital as an investment, I started a business in the field of artificial intelligence. Below, I will share with you some of my recent observations from the perspective of the intersection of investment and entrepreneurship, and you are welcome to communicate at any time.

Let me briefly introduce, Fengrui Capital is a new fund established in August 2015. We hope to become a research-oriented full-chain fund, which will be held for a long time regardless of the trend. We started to look at the field of artificial intelligence relatively early, and have invested in more than ten startups in the field of AI.

Cut to the chase below.

Technology and talent dividends are slowing down rapidly

When I started a business in the field of artificial intelligence in the past, people would ask me these questions:

These two graphs reflect my observations on these two aspects over the past period of time.

How to calculate the valuation of artificial intelligence companies, here are several formulas

The first graph, which I call “The Dividend of Technology”. This figure is the error rate of the algorithm that has won the first place in the ImageNet image classification task over the years. In 2013, the error rate of the algorithm that won the first place was 13%, in 2014 it was 7%, in 2015 it was 3.6%, and in 2016 it became 3.0%. As you can see, from 2013 to 2014, the error rate dropped by nearly half, from 2014 to 2015, it dropped by nearly half, and from 2015 to 2016, the decline became very small .

Anyone who does technology should know that under the framework of deep learning, with the ability to handle image classification tasks with existing technologies, there is not much room for the error rate to continue to decline. I was very surprised to see this conclusion. Technology is slowing down much faster than we ourselves thought.

The second picture, I call it the “talent bonus”.

The horizontal axis is time, and the vertical axis is salary. In the past, AI companies were expensive to hire. Recently, I randomly selected the recruitment list of a very well-known artificial intelligence company. In 2017, the salary of this company for recruiting image recognition processing engineers was 15-30K/m, which was already similar to that of ordinary IOS engineers.

At present, the general perception is that in the field of artificial intelligence, the framework of what to do has become more and more clear, but the talent gap is relatively large, and the supply of schools is not enough. Any artificial intelligence company is more inclined to recruit people who are skilled and can quickly put their ideas into practice, and tend to recruit talents with certain professional colleges and academic backgrounds.

From the above two pictures, what I want to say is: artificial intelligence technology, like other technologies, has reached a staged platform period, the speed of technological dividends is slowing down very fast, and the speed of talent supply development is very fast.

For investors, in the first stage of artificial intelligence entrepreneurship, the company’s valuation is “algorithm x talent”. The product of them is probably the value of your company in the market. At present, both ends of this multiplication are rapidly declining, which is our judgment on the first stage of artificial intelligence entrepreneurship.

Most startups are in the “narrowly segmented technology” stage. At this stage, the criterion for judging a company is the formula we just mentioned: “valuation = algorithm x talent”. We can see that the value of companies under this formula is quickly leveling off. In my opinion, this wave of opportunity dividends has basically ended.

In the wave of AI entrepreneurship at this stage, the entrepreneurial group of scientists who have made the most profit will not have such a big advantage in entrepreneurship now. Next, I believe the opportunity will still be reserved for product managers, engineers and business talent. The past valuation methods, past value judgment methods, and past technology and talent dividends have basically ended.

Now, many companies have entered the second stage. Whether it is an image company or a voice company, everyone has begun to enter the stage of providing solutions.

In the second stage, the way in which the market value of a company is judged also changes. I made a formula myself, which evolved from “algorithm x talent” to “valuation = algorithm + data x business value”. There is a plus sign after the algorithm. The main reason is that when everyone has no data and business value, it is compared to the algorithm, but when everyone has the data and business value, the importance of business value will quickly exceed the algorithm. Therefore, the proportion of the algorithm will be less and less.

AI is not the next generation of the Internet

How to calculate the valuation of artificial intelligence companies, here are several formulas

The most important thing the Internet did in the past was to liberate the channels and release the efficiency of the channels. So we see that the innovative models in the past are all about making products reach consumers directly. Whether it is e-commerce to destroy the channels of retailers and dealers in the middle, or Didi to destroy the channel of taxis, they are actually making a fuss about the channels.

I think artificial intelligence is not the next generation of the Internet, nor is it a substitute for the Internet. The two are parallel. Therefore, there are still opportunities for mobile Internet and the Internet, and they are still very large. The line of the Internet itself will continue to develop, and what artificial intelligence does is more inclined to the bottom-increasing production efficiency on the production side.

This is why we feel that the opportunity to C side is difficult now: the production side has not been transformed, so there will not be new products coming out; without new products coming out, the user experience will not be greatly improved.

This wave of artificial intelligence highlights the “data x business value”. We analyze it from two perspectives, to B and to C.

To the B side, my judgment is that artificial intelligence will develop to the deep end of the industry.From the perspective of production, whether it is service industry, agriculture orindustry, Everyone’s demand for efficiency improvement is very obvious: in the medical field, improve the efficiency of diagnosis; in the financial field, improve the efficiency of financial data services… The demand and commercial space are huge. So our challenge is no longer the leveling of technology, but the understanding of industry needs and product design.

Another point to mention: big data. This term is actually a cliché. From the point of view of the enterprise server. The path of the US enterprise service market is: traditional software → IT services → cloud computing → big data → artificial intelligence, and finally there is artificial intelligence. Therefore, the advantage of doing entrepreneurship in the field of artificial intelligence in the United States is that the infrastructure is very mature and perfect, but many things have been done by large companies, and startups need to find a living space in the cracks of large companies.

In China, the enterprise-side infrastructure is still very backward and blank. Therefore, China has a unique phenomenon: leapfrog development. In industries such as e-commerce, local services, and financial payment, there are many examples of leapfrog development.

To the C side, my point of view is: if you want to use artificial intelligence to improve the consumer experience, it is more difficult to break through a single point, and it is more a systematic project. For example, the interactive experience of car voice, smart home experience, etc. What it needs is not only how well the product itself is done, but also the overall home environment and vehicle environment. After these basic sensors and basic services and data are complete enough, we will have a good enough experience. On the contrary, doing this now will feel very tired or unable to pry, because the infrastructure and services are not up.

The Links:   SRDA-SDB71A01A-E   3HAB3772-75

Published on 09/19/2022