The topic of AI has been on fire for a year and a half, some people use AI to hype financing, some people cheat baby moms to learn AI to earn more than 10,000 per month, some people put the traditional scripting software to set a AI shell to sell directly to join, and some people are using AI to reduce costs and increase efficiency.
At present, the domestic AI market can be said to be three religions and nine streams, mixed, disorganized and strange.
Cause brothers have been wondering:AI this industry in the end to earn money? Which companies are making money? How much money can earn? How can ordinary people use AI to make money? What are the pitfalls?
Today I've asked Business Circle member @Brother Leyang to talk to you thoroughly about the AI industry.
Leyang brothers since last February when ChatGPT hot into the industry, this year and a half he has developed tools, selling resources, training, have done projects, can be said to be related to AI anything to do.
This article is the most real status quo of the AI industry, understand AI, improve AI industry awareness, see this one is enough, the other is redundant, do not believe that you finish reading.
How do companies that do AI make money? How much do they make?
existingAI CompaniesIt can be roughly divided into 6 categories and I'll show you how they all make money?
1. Pure AI companies - not making money and posting money
Companies that develop big AI models and underlying technologies, like domestic companies like Smart Spectrum, Dark Side of the Moon, and so on, are not making money right now, but instead are still investing heavily in iterative technologies because most of them have capital and national teams standing behind them.
These companies are going to be platforms and infrastructures in the future, underlying platforms like WeChat and Jitterbug, and this is not a project that we ordinary people can get involved in.
In some recent industry news, the current state of the AI industry's six tigers is said to be as follows.
(AI Six Tigers, referring to the six largest modeling companies in China, namely Wisdom Spectrum AI, Baichuan Intelligence, Zero-One Everything, Dark Side of the Moon, Minimax, and Step Star)
- Tiger Cub One:Headcount down to 500 after Wide Entry program, expansion over, lighten up for IPO push.
- Tiger Cub II:Pre-training is out of the question, C-end is out of the question, B-end isn't making money, and I'm lost for the time being.
- Tiger Cub #3:Focus on key business and lock HC up.
- Tiger Cub IV:Continue to remain silent.
- Tiger Cub #5:From the bustling world into silence.
- Tiger Cub #6:Continued expansion, bucked the trend, and at least two friends got offers from them.
2, AI tools of the company - mainly by selling tools to earn money
This kind of company is mainly with the ability of those AI big models above, develop application tools on it to sell money, collect recharge, or free to use to sell advertising space, go the familiar software logic.
For example, AIDigital HumanAI+Lawyer, AI+Medical, AI+Office, AI+Writing Copywriter, ChatGPT Shell Station, etc., a bit similar to the time when the small program came out, we developed an application tool on it, such as "watching the typhoon small program", free to use, but selling disaster relief materials.
The essence of AI is still dependent on the birth of technology, which is like when the mobile Internet came out, everyone is a swarm in the development of APP.
Now there is also a combination of AI hardware and software, packaged together to sell, similar to the live broadcast era of live all-in-one machine logic.
There are companies that make money in this, and there are companies that lose money, after all, R&D costs are high. This kind of loss of R & D costs are generally not recovered, and earn about ten w to a thousand w a year ranging.
3. AI training companies - rely on charging for courses
AI is a basket, anything can be loaded into it, I think you have also found that the direction of AI is particularly large, and pan-AI is very difficult to land on a specific point, only to rely on training to make money.
We should all know that the xx boat teacher who was killed off, last year at the beginning also made a lot of money, but with the control of the platform, more and more people enter, it is more and more difficult to make money.
I remember Mr. Kuang wrote an article "Lao Kuang:The sky is falling for knowledge payment! The head of each track is collapsing!!!!". , it is true that now the general environment of knowledge payment is relatively poor, and the overall is not very good to sell, but if it is done in the long term and formalized, I think there is still some room for imagination.
For many people who have not been in contact with the first step is to not know how to use AI, will not use the feeling of fear of not keeping up with the times, learning AI feels like mastering a technology. A bit like the year of Xinhua Wenda computer training school, in word, ppt, ps and other software out of the time, first teach you how to use computer software, and then give you a certificate, as for you to learn to operate to do whatever you.
This ability to deliver reliable content, move toward skills training, and add in state certification exams at a reasonable price, still has a long round of opportunities to make money, just not profiteering.
If the things taught are of little use, especially the kind of profiteering pure leek-cutting training that tells you that baby moms can earn over 10,000 a month by learning AI, short-term waves of fast money will still be scolded to death, and can't run away.
Very few AI training companies lose money, and the ones that do make anywhere from 10w to a couple hundredw a year.
4. AI Infrastructure Supporting Companies - Making Money for People Who Want to Make Money
As the saying goes others pan for gold and I sell shovels, it's a solid business.
A lot of people made a killing selling ChatGPT accounts in the earliest days.
This industry includes AI consumption of key, AI data labeling and other peripheral hardware and software support. A variety of training AI with graphics hardware, some can even sell millions of dollars, even with the 4090 graphics card to play the game we can not buy, and now can only rely on scalpers.
Very few of these lose money, and the ones that make it range from a few w to a few hundred w a year.
5, AI application (use) type of company - with a project that can make money + AI empowered to go up to make money
There are many ways to combine in different industries, a traditional project overlaid with AI can give birth to a new form.
This year, there are also a lot of vertical industries that are gradually starting to look for opportunities to bring AI+ into their business.
Some of the very best programs, the better combinations, are currently muffled, and these people don't like to show their faces or advertise outside.
Some of the small and medium-sized programs that have been advertised on the market and are still good are probably the following.
- AI public number explosion (last year more fire, but now the public number of this piece of suppression, the scale should be more difficult to do, the head in the monthly scale of more than 100,000, poor in a few k or so)
- AI ghostwriting project (threshold in the middle, is expected to be a medium to long term project, the head has a monthly small hundreds of thousands of scale, poor in a few k or so)
- AI e-commerce modeling/photo portraits (high threshold, expected to be a medium- to long-term project, the head has a monthly small hundreds of thousands of scale, poor in a few k or so)
- AI old photo restoration animation (low threshold, short-term project, early July fire, as of the time of posting, the project has entered an unstable state, I do not know whether it can still earn money, the project is probably at the level of monthly income of 5k-1w)
- AI novel comic book tweets (which were hot last year, making a few k to a few w a month)
- AI Headlines (hot lately, making a few k to a few w a month)
- AI Beauty Dance (Shakeology was hot last year, not sure of the benefits)
- AI resurrects loved ones (supposedly there are good runs, as the track is niche and the returns are not clear)
- AI girlfriends and companion programs (reportedly there are good runs, but high returns many suspected of rubbing it in, Shenzhen side was caught in a wave, the specific returns are not clear)
- AI live naming (hot on Jitterbug mid-year, revenue unclear)
- AI Baby Walk (hot on the video number mid-year, proceeds unclear)
6. AI personal hobbyist - not profitable
A lot of individuals in the AI community, don't make money.
They're just bullish on the direction of the technology or a hobby, such as thinking that silicon-based life eventually replaces carbon-based life, purely for the love of power, and staying in it for now.
Why is that?
Unlike the Internet, which has been developing for so many years, actually the business model of AI, the realization model is still very immature.
This is a very normal phenomenon, like the wave of the mobile Internet, from 01 years of cell phones with Internet to 09 years of smartphone APP was born, it took 8 years. WeChat was born after 2 years, and Jitterbug was born after 6 years.
After the birth of WeChat, it took 2 years to get public numbers, 6 years to get small programs, and 10 years to get video numbers.
After the birth of Jitterbug, it was 2 years before live streaming, 3 years before e-commerce bandwagon, and 6 years before local life.
And AI has only been around for 1.5 years now.
How do small business owners and regular people, currently, get money with AI?
Actually, my epiphany is:AI can't mess with money, AI is just a tool.
How do you think computers made money when they first came out?
When there is abacus old accountant this industry, with computer + excel + financial software = an accountant can manage 200 companies accounts of the new agency.
At the time of the Industrial Revolution, the textile machine itself could not make money, it was the business form of the textile mill that realized profitability, but the textile machine realized more profit with less manpower to build the factory.
So the logic of AI realizing is to find a really profitable project and do it all over again with AI's ability, small teams with less manpower may be able to achieve the scale that only big companies can do before, when brain activity can be replaced by machines, this is actually a very friendly era for individuals and small teams.
We can't afford to be an AI-native business, but we can be a company that leverages AI capabilities.
You can find some on there that you feel good aboutAI ProjectsGo ahead and mess with it, or you can create a new program based on the capabilities of AI.
I've categorized the ways in which AI lands to get money into 5 categories.
Category 1:Finding an AI to do things for you
The nature of business is information asymmetry. When Taobao first came out there was a generation of operation, and when Shake Shack came out there was a generation of shooting and editing.
Right now there are a lot of people whose underlying logic is to make money with the help of AI making things for other people.
Category 2:Finding a way for AI to get traffic
As we said earlier, find an industry that makes money on its own and do it again with AI.
Category 3: Efficiency or Cost Disruption in Verticals
As we said earlier, find an industry that makes money on its own and do it again with AI cost effectively or efficiently.
Category 4: Model Innovation in the Pendant Industry
Starting a new business with AI, but exploring new models is slightly more difficult and not suitable for small and medium-sized players with capital.
Category 5: Marketing gimmicks or industrial cooperation
There are a lot of named budgets in the to G thread right now for people with resources above them to dish out.
What are the pitfalls of getting money with AI these days?
AI is a market that's a mixed bag of whatever you're doing.
Now there are a lot of companies to set a concept of AI, go out and sell franchise, sell tools. In fact, it is not AI at all, it may be a traditional running script software or the like.
Others may be in order to sell tools, bragging nonsense, but they do not understand the rules of the platform at all, or did not do the project, just fooled you to fall for it, so be careful to distinguish when in contact.
Many people should have seen all sorts of publicity selling software, such as: with digital people live system to earn tens of thousands of dollars a month.
I can tell you that most of them are all flimsy, look at those who sell digital people tools, they do not engage in digital people live streaming, they are all their own real people on the air. We have actually tested, take it to Jitterbug to start broadcasting, it will involve the problem of reporting white and often blocked.
And the video numbers have recently written outright rules against live streaming of digital people.
So the live streaming route is not going to work.
Digital people just came out that time, the jitterbug is not yet wind control, did broadcast a while, the data is good track is the local life, because this track is not much people broadcast, as long as you have a thing 24 hours a day to the stores to broadcast, more or less can earn a little.
But the e-commerce bandwagon track did not run out of particularly good data, because the jitterbug is still essentially a data horse racing mechanism, e-commerce bandwagon is too rolled up, like the big wolf dog, the small Yang brother those anchors of the performance is particularly strong, the digital human performance can only be counted as a flat podcast, so the race is simply not able to beat the professional anchors, and can't get the push flow.
Of course, in the niche live platforms, such as Taobao, Jingdong, etc., at present, you can still broadcast, can make a little bit of money, suitable for platform sellers to operate.
So are digital people useless? Not really.
It can really run not in the live broadcast, but in the short video track, the practice is to help individual IP lay matrix content, data is not enough to rely on the volume to make up, the production speed is fast and does not delay the founder IP to do other things.