In today's intelligent era, it is not difficult for an enterprise to market, but if it wants to market well, it has not become easy because of the digital age.
The internet isn't short of traffic, but user engagement is being diluted by a variety of platforms, making it increasingly challenging to capture users' attention. From traditional portals to mobile information to short platforms, channel changes are only the foundation, users are affected by channels, their needs and interests become more diversified, and the demands of advertisers or brands also change, which is a daily occurrence.
It can be said that the core competitiveness of marketing is to "respond to change", which is measured every day or even every minute and second, vertically, it is the mining of the needs of a single user, and horizontally, it is the in-depth understanding of the latest hot spot tracking and platform rules.
In view of these two pain points, today's large models and related technical routes can be well solved. From a technical point of view, generative large models improve the efficiency of data analysis and material output, and more importantly, the AI agent that cooperates with large models has the ability to understand, plan, and execute complex tasks independently.
To put it simply, the agent can complete tasks independently, has the ability to adapt to complex scene problems and the advantage of solving complex problems, which is a layer of buff added to the large model.
In the second half of this year, 360 released the 360 intelligent marketing cloud driven by the 360 intelligent brain model, which has taken the lead in adopting the agent model and integrating the enterprise knowledge base. Under the 360 intelligent marketing cloud, AI digital humans, AI digital employees, 360 intelligent painting and other products have also appeared one after another.
The gameplay of the new marketing era may be able to start from the 360 intelligent marketing cloud.
This year, the battle between major manufacturers is no longer news, with ChatGPT as the goal, domestic technology companies have caught up with its AI results, and are committed to continuously narrowing the gap with ChatGPT.
But how big is the technical difference between the underlying large models?To what extent can these differences affect the effectiveness of AI services?This series of problems was ignored in the sudden AI involution.
Huang Jian, vice president and commercialization president of 360 Group, said in an exchange with 36Kr that at this stage, in terms of the technical route of the entire large model, there is not much difference between everyone, so how can the final difference be reflected?At the application level, what problems and what scenario requirements to solve with large models are the biggest differences in the future.
Huang Jian, Vice President and President of Commercialization of 360 Group.
Everyone knows that the big model is the future, but who really implements the new technology in the vertical industry first is the final competition, and the latter not only needs technical capabilities, but also the ability to combine technology and business. The experience of 360 is that only after a trip in the industry can we truly understand that "the winning score of the large model competition is the best".
According to Huang Jian, at the beginning, when the GPT trend swept the country, everyone plunged into the competition of large model technology, and invested a lot of resources and time in the process of parameters, data volume, computing power, training and learning in order to brush up the score of the large model, "but later, we found that the gap between the general large model itself will not be too big." ”
Especially after many large models were open-sourced in the second half of the year, the 360 team found that simply increasing the size of the large model could not achieve the real value of efficiency improvement, and what was really meaningful was the gap in the vertical scene.
There are more down-to-earth trial and error to help 360 understand the big model and the industry. For example, at the beginning, in the application layer, the 360 team will also be large and complete, "We believe that user needs are very different, so we must do 'more'." But after doing it, I realized that too much force will lead to many demand points that cannot be penetrated. ”
Huang Jian realized that the reasoning and generation capabilities of general large models in vertical scenarios should not be overestimated, and there is still a big gap between general large models and specific industry applications.
In June this year, Zhou Hongyi pointed out that the vertical model is the future development direction, and the real value is to make the large model change from "know-it-all" to "jack-of-all-trades", "industry and enterprise".
Based on these experiences, 360 quickly adjusted its strategy, from long to refined, from focusing on the bottom to focusing on applications, and the final preliminary result is 360 Intelligent Marketing Cloud.
The solution of "large model + enterprise knowledge base + agent" pioneered by 360 intelligent marketing cloud is the key barrier to the application of large model. The agent mode allows the large model not only to recognize human intentions as a "brain" and intelligently think about and decompose tasks, but also to grow "hands and feet", automatically use tools, call various APIs, perform tasks, solve problems, and achieve target results, becoming a general agent system.
Relying on the support of large model capabilities, knowledge base training capabilities, agent studio, digital humans and other technologies, 360 Intelligent Marketing Cloud has successively launched products such as AI digital humans, AI digital employees, and 360 intelligent paintings, using AIGC to empower the whole link of marketing.
The reason why the marketing industry is chosen to take the lead in the commercialization of large models is determined by the industry attributes behind it.
Compared with other industries, the marketing industry presents the characteristics of fragmentation, rapid change, and high degree of personalization, and the so-called "thousands of people, thousands of faces" marketing is the result of the initial application of artificial intelligence, but at the same time, the problem is that artificial intelligence and other high-tech marketing is no longer inclusive, and a high threshold has been erected in front of the marketing business of small and medium-sized enterprises.
High cost, insufficient informatization capabilities, and long time consumption are all pressures that small and medium-sized enterprises can hardly bear in marketing.
In the past, advertising was limited by the knowledge and experience of optimizers, but the large model can be compliantly encapsulated, you only need to tell the large model an end goal, and it will help you pick up traffic and generate ad-related elements for delivery. In fact, at the level of algorithms, materials and platform operations, large models can be used to transform. Huang Jian said.
According to Huang Jian, the daily Xiaohongshu grass planting articles are a headache for many ecological service providers. "I recently communicated with Xiaohongshu ecosystem partners, and their pain points are not about the follow-up results, but about helping them solve the problem of communication with brands first. Just in the step of communicating needs, many service providers have been defeated because they do not understand the brand's whimsical ideas. ”
After in-depth insight into marketing pain points, 360 Intelligent Marketing Cloud has helped many brands achieve marketing reform. For example, Pirate Shrimp Rice is the first case in China to apply AIGC to the whole process of new product development in the catering industry.
According to the general food and beverage new product development process, it usually needs to go through seven steps: demand collection, internal research and development, product internal testing, factory version of products, factory version internal testing, single store testing and store-wide promotion. However, with the support of AIGC, the entire process has been greatly reduced, and the efficiency has also been greatly improved.
Specifically, how does AIGC empower the new product development step?
First of all, the first step is to address the category demand. In the absence of professional data analysts, decision-making is time-consuming and laborious, 360AI digital employees are used to build an enterprise knowledge base, create industry expert workflows, establish the role of AI product managers in the catering industry, quickly read a large number of industry reports and brand management data, and output new product category suggestions with reasonable evidence.
Secondly, solve the taste needs. In the past, for the question of "is it a tomato or a black pepper", most of them used internal voting and blind testing experience, the test cycle was long and the cost was high, but now there is no need to actually shoot, through 360 intelligent painting (AI raw picture tool), you can generate different flavors of dish materials, with the help of 360 as a strong traffic advantage of the marketing platform, with 360PC lock screen pictorial as the position for taste preference testing, which not only expands the test sample size, but also reduces the test cost.
In response to the needs of single-store testing, Pirate Shrimp Rice created the role IP of the captain through the method of 360AI digital human + large model, based on the learning and training of different knowledge bases, and took on multiple roles such as training lecturers, welcoming store managers, exclusive customer service and private nutritionists instead of the brand founders.
Finally, in order to solve the problem of high cost and single means of promotion in the whole store, Pirate Shrimp Rice uses 360AI digital employees, 360 Zhihua (AI mapping tool) and other product portfolios to assist the brand's social media operation, and the per capita single-day output of copywriting has increased from 3 to 50+, and the interaction rate of Xiaohongshu's official account reviews has increased by 89%.
The perspective is retracted to the interior of 360, and the large-scale model ecology based on 360 Intelligent Brain reconstructs the overall business and products of 360 from top to bottom.
This innovative AI platform not only inherits the accumulation of 360 in the PC era, but also promotes the business transformation and upgrading of the entire group.
In the past, 360 has accumulated rich experience and data in the PC era, and these resources and achievements have provided valuable nutrients for the research and development of 360 large models. The 360 model deeply integrates and optimizes these resources to form a more comprehensive and efficient technical system. In this process, AI technology continues to amplify the advantages of PC-related scenarios, and the original products and data of 360 also provide strong support for the training of large models.
On the user side, 360 Intelligent Brain has been fully connected to 360 search, 360 secure browser and other full-end products. On the enterprise side, 360's intelligent marketing platform, 360 Eyeball, has also been fully upgraded.
The comprehensive upgrade of the overall business by the large model is reflected in the "one technology for multiple purposes", which means that the commercialization efficiency of the past business will be exponentially improved.
The 360ai digital human is the best illustration of this. Driven by large models, digital humans can be applied to many different industries to maximize business value. According to reports, different from other digital people, 360 AI digital humans can obtain the latest knowledge by calling search and browsers, and can strengthen the cognition of facts by large models through private knowledge bases, and even, it will learn hot search content every day.
Huang Jian revealed that 360AI digital human is currently one of the most popular products among customers, and in the first half of this year, the revenue of 360 intelligent marketing cloud has reached tens of millions of yuan.
From a horizontal point of view, when other large manufacturers are still struggling with commercialization problems, 360 can achieve landing results in a short period of time, which is the result of the combined effect of 360's past product accumulation and large-scale model technology. When talking about the advantages of 360 in the commercialization of large models, Huang Jian said: "The scenarios that 360 large models quickly enter are ready-made, and the scene coverage and market share of the first application position of large models are also the highest. ”
The wave of AI is sweeping, but compared with looking up at the stars and pursuing technological breakthroughs, perhaps it is better to feed back the underlying technology by doing a deep and thorough job in the industry application. 360 is the first city to commercialize large models, and this is just the beginning.