WeChat*** Silicon-based letter (ID: GH F908794E917C), author: silicon-based letter, original title: "Deep Review AI Explosion, CNY Dumb Fire", header picture from: Visual China As the most important marketing vane of the year, this year's CNY is quiet.
In the era of fragmented communication, it is difficult to create a large marketing event with national memory points. One of the few bright spots in the opening marketing of the Year of the Dragon is brought by AI.
If AI marketing in 2023 is just a self-sprout for some brands, then 2024 will be a year of profound marketing changes. Standing at the moment of marketing in the Year of the Dragon, let's take a look at who is catching up and who is leaving the scene at this moment through typical AI marketing cases.
1. Brand: Elixir or spectacle.
Whenever there is a major infrastructure or customer migration, we always hear "xx is worth doing all over again". For example, the migration of PCs to smartphones gave birth to the mobile Internet, and the post-90s took over the baton from the post-80s to achieve a new consumer market, and every major switch is a market reshuffle. The brand's focus on the AI track is a bet on the next round of growth.
This Spring Festival,The food and beverage industry is the first to start CNY's AI marketing campaign. Coca-Cola's "Dragon Connect You and Me" series of activities, in addition to the traditional packaging change and New Year's TVC, the most important activity is to use AI to lead millions of people to impact the Guinness Challenge; Master Kong will also upgrade the New Year's IP "Jiakangjia New Year's Flavor" for three years, and launched a large-scale user AI co-creation of "Ten Thousand Miles of Mountains and Rivers"; Ambrosia handed over the first AIGC answer sheet in China's yogurt industry, and won the Douyin Challenge based on AI drawing 2500 million**.
When traditional marketing giants collectively bet on AIGC at the most important dynamic sales nodes, what kind of marketing leap does this reflect?
1.AIGC, rewriting the marketing ecosystem.
(1) Bring the most thorough customized experience to the mass consumer market
From the moment mass communication fails, brands have embarked on the path of personalized communication.
For a long time, continuous communication with users has been the ultimate goal of brand personalized communication. But the inherent contradiction is:Fixed, standard, and limited ideas cannot serve a fluid, disordered, and infinite user group. To serve a wider range of people, creativity can only take the largest set of the crowd, resulting in the scattering of creativity; To be creative accurate, we can only focus and focus again, resulting in the loss of efficiency.
AIGC is the latest answer to this question. With the powerful semantic understanding ability and smooth user interaction of the AI model, each user can get their own customized experience in the shortest possible time (this experience used to only exist in luxury!). And for much longer). The essence of advertising and marketing is to create a dream that is higher than the user class, and substituting the customized experience of luxury goods into the mass consumer market is the reason why AIGC is sought after by FMCG giants. More mass consumer markets such as daily chemical clothing and IT3C are facing the same situation, and may usher in a new wave of AIGC marketing demand.
(2) Use the most abundant form to meet the needs of the whole industry.
In the early morning of February 16, OpenAI released its first AI** generative model, SORA, which has achieved a similar breakthrough to large language models in the field of vision. Prior to this, similar breakthroughs were only in AI biotexts, and no one expected the "Aha moment" of AI ** to come so quickly.
Thanks to the development of multimodal large models, from text to sound, from ** to **, the generation ability of AI can be applied to all human mainstream communication media, which almost determines that there is no upper limit to the product application of AIGC. Considering the maturity of different modal technologies and the matching of industry characteristics, the current gameplay between industries is not the same.
AI writing and AI writing technology are the most mature and have the widest range of applications. Brands that emphasize personalized user communication can try, especially for brands with large user scale and serious homogeneous competition in the industry. Combined with social fission tools such as sharing, it has the potential to create phenomenal marketing events.
At present, AI voice has begun to be widely used in the customer service fields of local life, finance and insurance. Combined with multi-round dialogue and AI speech synthesis capabilities, virtual social networking has also gained a lot of growth. For example, in the case of "AI Yanzi Sun", which exploded last year, customers in transportation, liquor, and IT3C with similar marketing demands can try it.
AI is suitable for games, online services, e-commerce, self-sufficiency, etcAmong them, the development of AI digital humans is even more outstanding. Last year, the same explosion of mold and mildew speaking Chinese ** came from Shenzhen-based AI company Heygen, whose translation, voice cloning and mouth swapping technology is very mature. At present, live broadcast rooms, including Liu Run, are also trying a lot of digital human live broadcasts, and the potential here will continue to be released in the future.
2.AIGC Marketer**: TPMF
The unfamiliarity with AI has stopped most brands that want to try it.
In the world of product marketing, there is a concept called PMF (product-market fit). If a product finds its own PMF, that is to say, it has found its own market foothold and begins to have user stickiness. In the field of AI large-scale model products, Wang Xiaochuan, the founder of Baichuan Intelligence, proposed that large-scale models should find their own TPF (Technology Product Fit), emphasizing the important impact of technology on AI products. Then the brand owners who are further downstream in the AIGC marketing end are choosing AI marketing methodsIt is recommended to comprehensively consider the three perspectives of technology, product and market, and choose the matching degree of AI capability-commercial product-user trafficThat is, TPMF (Technology Product Market Fit).
(1) T - Technology AI generation capabilities
An AI product may be the result of a combination of multiple models, and different models often integrate multiple capabilities. According to the technical maturity of different models, the models can be simply divided into the following categories from the perspective of capabilities:
Language Models:It is currently the most mature to be able to understand and generate natural language, and many models at home and abroad have been able to generate coherent and grammatically correct long texts, such as GPT, BERT, etc.
Visual Model:Ability to understand and generate images such as SORA, etc.
Multimodal model:Combine multiple modalities such as language and vision, and have the ability to understand and generate multimodal content. In the rapid development, there is some progress but it is not yet mature.
Conversation Model:Able to interact with natural language and answer questions. I have a basic ability to chat, but I still lack common sense and logical reasoning skills.
Inference Model:Able to reason logically, answer questions, such as comet, etc., but still nowhere near human level.
Knowledge Graph Model:Understand entity relationships and build knowledge graphs. With the initial framework in place, quality and scale still need to be improved.
Machine Translation Model:The quality of automatic translation between different languages still needs to be improved.
Speech Recognition Synthesis Model:For speech recognition and speech synthesis, the quality still needs to be further improved.
Recommender System:Personalized recommendations based on user interests, but the level of personalization can still be improved.
Language models are one of the first areas to receive attention, and they are also one of the most technologically mature. To put it simply, "Wensheng Wen" has the best effect. The technology in the fields of "Wensheng Diagram" and "Wensheng **" is also advancing by leaps and bounds, especially with the release of SORA, with the longest generation of 60S**, the consistency of the lens, and the characteristics of ultra-realistic, it may open a trillion-dollar dynamic and personalized advertising era. However, there is still a lot of room for improvement in technologies such as model dialogue, model reasoning, and knowledge graphs, and there may be more uncontrollable factors in practical application, so it is necessary to pay attention to the constraints.
(2) P-product commercial products
It is an indisputable fact that AI products are difficult to land, and the commercial AI products that can be used are even rarer. The current implementation paths in the field of marketing are:
Privatization Model Deployment and Vertical Model Training:This option is suitable for powerful advertisers or marketing groups. It requires a complete top-level design and large investment, and the advantage is that under the condition of ensuring data security, the best model generation effect can be customized and trained according to the actual needs of business scenarios.
For example, last year, Leo Digital and Guiyi Intelligence created a large marketing model. **With years of exclusive data accumulation and actual advertising application scenarios, you can land the large model capabilities on products such as intelligent execution and strategy optimization of advertising delivery.
Call the API API of the open-source model:Suitable for advertisers with development capabilities. At present, the open source ecology of large models is already very rich, and mainstream large model manufacturers basically have open source capabilities. Advertisers can quickly build AI products and solutions by calling APIs. The ability of Chinese large models has its own merits, and it can be evaluated after multiple tests before actual use. It is generally believed that such as Wenxin, Ali Qianwen, and Zhipu GLM belong to hexagonal warriors with strong comprehensive strength, Minmax voice model has a better effect, and Moonshoot's Kimi model is more advanced in the processing of ultra-long text.
For example, the "coaxing simulator" that exploded before the Spring Festival is the landing undertaking of the developer's direct call to Kimi's API.
B-end commercial products + C-end application scenario combination:At present, most of the AI products on the market are C-end applications, and C-end applications are mainly based on efficiency tools, ** generation, virtual social networking and other scenarios. B-end AI products for direct use by advertisers are more integrated in the integration solutions provided by the ** platform. It is difficult for B-end efficiency improvement tools to make C-end users have direct perception, and it is even more difficult to form event marketing with communication memory points. At present, some advertisers have begun to try innovative marketing methods that combine B-end commercial products in C-end user scenarios and communicate directly with users.
(3) M - Market user traffic.
The core of user traffic is to solve the problem of traffic coming from the first.
A basic flow formula is:Flow potential energy = product potential energy Platform potential energy Timing potential energy.
The essence of the success of AI marketing events depends on whether the AI-generated effect meets expectations. Earlier, we have talked about the generation effect, AI Shengwen, AI Shengtu, AI**; At present, AIGC takes efficiency tools, generation and virtual social networking as its main product positions, and there are more products to choose from, so the advertising mainly combines the two actual industry status quo, carefully considers the entry angle, and selects the best effect advantage of the product.
A good event needs a good fermentation platform, and the upper limit of the platform's users determines the upper limit of the activity. The last one is to consider the timing of the event, the Spring Festival has always been the peak marketing season for food and beverages, and it is also the peak traffic throughout the year. CNY has its own traffic potential, so it's not surprising that brands choose to get together. Next, at important marketing nodes such as Women's Day, summer vacation, and National Day, it is expected to see more high-quality marketing cases.
2. Dachang: present but absent.
Throughout the past ten years of mobile Internet, the platform of large manufacturers has eaten up the dividends. From Internet finance, to local life, to short-term marketing, almost every marketing node change is started by large manufacturers. But why is this year's CNY, and why are major manufacturers collectively dumb in AI marketing, and what kind of information is revealed behind this?
1.Rodeo games are no more.
From the beginning of the battle between Didi and Kuaidi, the story that the Internet has always been familiar with is "burning money-enclosure-harvesting traffic".
This set of stories has burned many business myths that we are familiar with, such as Didi, Meituan, and Mobike. The reason why the myth persists is that the marginal cost of the Internet is so low – opening a city is nothing more than adding a server; And the benefits of network effects are so high that once users have set up a data flywheel here, there is a reason not to leave. The game of burning money and enclosure has been in full swing in this land for ten years, and similar stories are currently being replicated around the world through going to sea.
But AI is a whole new game. The high cost of reasoning in this game is an unavoidable fact for everyone. Every time it is generated, it is a new reasoning. The developer of the popular AI game "Coaxing Simulator" during the Spring Festival reported that it costs thousands of dollars per day. And this cost cannot be covered by advertising, a traditional way of monetizing traffic. In other words, the monetization model for AI developers is almost only to let users subscribe and pay.
Those who get traffic win the world, which is almost unquestionable in the era of mobile Internet. But in the age of big models, the rules will be rewritten. Burning money without thinking clearly about the business model (especially the monetization model) will only lead to inaccurate users, huge costs, and the dilemma of not being able to close the loop.
The other end of the enclosure game is the network effect, that is, the Internet has established one barrier after another based on the advantages of user data: the more users use the barrier, the more addicted they are, and the competitors outside the barrier are firmly blocked. But the surging of the open source model in the last year seems to be breaking down that barrier. Jensen Huang, the founder of Nvidia and a recent U.S. academician, recently said he believes the most important AI event of last year was the release of the LLAMA 2 open-source model. Throughout the domestic model, it is also radically open source, and large manufacturers have come up with very good models. If a year ago people thought that a generative effect like ChatGPT would be amazing, now people only think that this is how it should be.
When the low marginal cost of the Internet + traffic flywheel has become similar to the traditional manufacturing consumables + homogeneous product competition. The story after being disenchanted lost **, a little more bitter.
Burn perfect youth, but you can't get your wife back. How to find their own traffic passwords in the new era, CNY did not give an answer.
2.Reduce costs and increase efficiency, and increase the stock.
So what are the big factories doing? The answer points to a word that everyone in Dachang is familiar with:Reduce costs and increase efficiency.
Since the new path is uncertain, it's always good to walk in your comfort zone. In addition to hoarding cards, the most important measure is to use AI to empower the original business, and change the traditional advertising, search, game, cloud, ** to advertising + AI, search + AI, game + AI, cloud + AI, **AI. Increment from stock to improve business efficiency.
The main advantage of large manufacturers is that they have traffic entrances and user scenarios, and the disadvantage is that their businesses are fragmented and separated from each other. External business combined with AI,It can give AI practical use scenarios,Especially in efficiency tools、Virtual assistant、Virtual social networking and other recognized fields The efficiency benefits are still obvious,Here Wenxin is a unique file,The tool class represented by Jianying is also expected in the future。
Major manufacturers are also working on various tools to improve internal efficiency through AI. But unless there is a complete top-level thinking and determination to invest, at least a vertical model must be trained, otherwise it will basically become a departmental toy, but a short-lived shell.
The more significant benefit of large manufacturers combined with large models is the stock price. As one of the few bright spots in the capital market, all companies that were involved with AI could rise last year.
Boosting the original business is the cause and effect of the AI promoted by large manufacturers. The Spring Festival Gala is the vane of CNY, and we see that the major manufacturers that cooperate this year are Jingdong, Xiaohongshu and **. In the face of the pressure of Pinduoduo, JD.com hopes to use the Spring Festival Gala to gain another round of growth, and Xiaohongshu hopes to use the Spring Festival Gala to expand more low-tier people and become a new generation of national social stage. It's great that AI can help your business with the icing on the cake, but in the face of an existential battle, sometimes it may be more effective to "throw coins".
Banal gravity, shameful but useful. The big manufacturers, who should be the pioneers of marketing innovation, have generally played a conservative card in CNY, behind which is the ten-year intergenerational change of Internet logic, and also a cautious judgment of the current complex economic situation.
3. Start-up: rugged and starlight.
Last year was the worst year for startups. But last year was probably the best year for AI startups.
Benefiting from technological breakthroughs and the explosion of ChatGPT, AI, especially large language models, has become the biggest technology investment highlight in 2023, with the number of financing events in the AI field increasing by 145% over the previous year, and the financing disclosed by the four major and one small (Minimax, Zhipu AI, Baichuan Intelligence, Zero One Everything, and the Dark Side of the Moon) is about 10 billion.
Almost all of these companies have a unified and glamorous background: technology founders, former executives of large factories, capital pursuits, young and innovative teams, etc., but these companies are also facing almost the same embarrassment: how to implement technology, where is product differentiation, and where is China's AI super APP?
The problems faced by large manufacturers are also a severe test for start-ups, and the market gives them a lower tolerance rate for error. Throughout 2023, almost all domestic model companies are rolling models, but it seems that everyone has reached a consensus overnight2024 is a critical year for roll products. And this war has already quietly begun.
CNY, a peek at the leopard.
During the Spring Festival, Zhipu Qingyan, which is affiliated to Zhipu AI, presets many intelligent twins (GPTS) such as "Soul Question for the Year of the Dragon" and "New Year's Greeting Assistant", as well as a Spring Festival special area (Prompt) that combines AI Q&A, AI writing, and AI painting, to provide users with services such as writing couplets, writing blessings, and drawing portraits of the Year of the Dragon.
Wenxin Yiyan and Ali Tongyi Qianwen also launched similar functions during the Spring Festival, and it is obvious that these functions also have the shadow of Coca-Cola and Master Kong's Spring Festival activities. The embarrassing scenario of homogeneous competition of AI products has begun to take shape in CNY. And homogeneous competition is nothing more than two roadsVolume efficiency or volume brand.
The battle for efficiency is about cost and effect. Better model fine-tuning, lower computing costs, and better data quality can all become barriers. OpenAI is the ceiling of the effect, and all the other model companies are in a state of catching up. However, the Chinese large model is limited by the limitation of high-quality Chinese corpus, and the competition at the efficiency level is anxious, and none of them can open a big gap with others.
The battle for brands may be the next stage to enter.
Large factories have pressure to boost their stock prices, and small factories have pressure to raise funds. This year is a critical year for the landing of domestic AI products, and everyone is looking for suitable scenarios to cut into. If the gap cannot be opened at the effect level in the short term, then the user's mental cognition brought by product differentiation positioning will be a new increment**.
Behind all the problems of marketing is the problem of growth. Whoever can help businesses solve the problem of growth will be the biggest marketing winner of the year. AIGC brings a window dividend like **short**, PC to smartphone, with the switch of infrastructure and the leap of generational demand, it will usher in the reconstruction of the entire marketing ecology.
The Year of the Dragon has come to an end, but the war on AI marketing has just begun.
WeChat*** Silicon-based letter (ID: GH F908794E917C), Author: Silicon-based letter This content is the author's independent point of view and does not represent the position of Tiger Sniff. Do not do without permission**, please contact hezuo@huxiu for authorizationcom
People who are changing and want to change the world are all on Tiger Sniff app