What can you buy for a dollar?
At the 2024 Spring Volcano Engine Force Motive Force Conference held this morning, Tan Bei, president of Volcano Engine, gave an unexpected answer.
One yuan can buy 1.25 million tokens of the main model of bean bags, which is about 2 million Chinese characters, which is equivalent to three copies of "Romance of the Three Kingdoms".
The price war is open! Byte released a large model of bean bags
The key challenges in the implementation of large model scenarios are the model effect, inference cost, and difficulty of implementation.
In order to help enterprises solve these challenges and help enterprises do a good job in AI transformation with better models, lower costs, and easier to implement solutions, Volcano Engine officially launched a new generation of full-stack AI services today.
A good technology must be formed in large-scale application and continuous polishing in the case of a large number of calls.
A year ago, the bean bag model, formerly known as "Skylark", became one of the first large models in China to pass the algorithm filing.
After a year of iterative development, the Doubao model now processes an average of 120 billion tokens of text per day and generates more than 30 million images.
Starting today, the bean bag model will be officially available to the public through the volcano engine.
The model family released this time includes a variety of models such as the bean bag general model pro, the general model lite, the role-playing model, the speech recognition model, the speech synthesis model, and the Wensheng diagram model.
Among them, the Doubao Large Model Pro is the strongest representative of the Doubao model, which has excellent performance in multiple dimensions such as understanding generation logic and memory, and supports 128K context windows, which can help users quickly understand the long text content of the earrings and summarize the difficult content.
In some scenarios, for latency- and cost-sensitive customers, the bean package generic model lite is a better choice.
Thanks to the blessing of the role-playing model, whether it is playing a Sichuan teacher or promoting the interpretation of the story of the script killing, the actual demonstration on the scene seems to be at ease.
Speech is an important part of interacting with AI.
Speech recognition models and speech synthesis models can accurately identify the user's content, language, and context, and through the learning of timbre, tone, and intonation, large models can express real feelings, allowing AI to communicate as if it were a real person.
Only by using a large amount of data can a good model be polished, and the unit cost of model inference can be greatly reduced. The volcano engine is directly incarnated as a "price butcher" today, bringing 100 million points of shock to the large model market.
The pricing of models of the same specification on the market is generally 0.12 yuan/1000 tokens, while the inference input price of the pro-32k model of the bean bag general model is only 0.0008 yuan/1000 tokens, which is 99.3% lower than the industry price.
Or, the input price of the pro-128k model is 0.005 yuan/1,000 tokens, which is 95.8% lower than the industry price, truly achieving the price involution of "cents to cents".
In addition, in order to help enterprises better apply large models, Tan Cheng also announced the launch of the Volcano Ark 2.0 platform and the release of three large model plug-ins:
- Networking plug-in: Search for relevant data information from the data of the whole network
- Content plug-ins: Provide rich video and graphic content
- Knowledge base plug-ins: Enables fine-tuning of proprietary data to minimize AI illusions
Application-oriented, landing is king
With an AI-native development platform, we have the opportunity to empower everyone to become a developer of AI applications.
At the press conference, Tan Bei made the above judgment loudly. As a next-generation AI application building platform, users can quickly build various model-based bots on the buckle, regardless of whether they have programming foundation or not.
Users can also publish bots to various social platforms, messaging software, or other channels such as websites.
The opening scene of the conference showed an AI bot for a fifth-grader who used the language knowledge he learned in school to create an AI bot that can be called an English teacher.
In an enterprise environment with higher requirements for application scenario capabilities, Haidilao uses buttons to simulate customer dialogue drills to help customer service ladies improve service levels.
Super Gorilla uses buttons to help users know how to get fit; China Merchants Bank has built a handheld life discount bot that recommends restaurants with preferential restaurants based on buttons, as well as wealth highlights that analyze market conditions......
The bot created by the button supports mixed interaction of text, pictures, voice, video, GUI cards and other modalities, and the user can even communicate with the bot like chatting with a real person, and he will always respond to every request of the user in the most intelligent and natural way.
In addition, Volcano Engine has also officially released the professional version of the button - the Volcano engine is further packaged based on the button platform according to the needs of enterprises, and supports many advanced features.
In the past ten years, ByteDance, which knows the most about the mobile Internet era, seems to have always been able to make a product hit, and has quietly become a behemoth today.
The most impressive thing about this conference is how ByteDance understands the application and development of products in the era of large models.
Zhu Jun, vice president of product and strategy, said that more important than the code running on the server side is to create the right product form and natural interaction mode to meet the actual needs of users, so that users are really willing to use these products.
At the press conference, Zhu Jun revealed the origin of the name of the bean bag.
"Bean Bao", which doesn't seem to have much to do with AI, is actually the first general principle that was determined when it was first named: simple, easy to read, and easy to remember.
At the same time, in order to shorten the distance between the product and the user, they also defined three product design principles for products such as bean bags: anthropomorphism, proximity to the user, embedding the user's use environment, and personalization.
In the past year, ByteDance has done a lot of exploration in the form of large-scale model applications, and his biggest feeling is that compared with before the AI era, there are both commonalities and great differences in making applications.
So what are the commonalities? The essential needs of human beings have not changed. For example, if you want to get information quickly and easily, you need to be more efficient in your work......
In his opinion, it was relatively simple to make products in the past, because at least the underlying technology was mature and stable, so you only need to exert your empathy and think about what the needs of users are, but the era of large models is completely different.
The new difficulty is not only to consider what the big model can do at this moment, but perhaps more importantly, to guess what new user scenarios can be implemented in 3 months, 6 months, or two years.
So this is a new challenge, and it is also necessary to constantly predict the PMF of the next product under the dynamic technological evolution.
Taking the AI search engine as an example, Zhu Jun revealed that in the first half of last year, the search task was often evaluated and 6 out of 10 questions were wrong, which means that the application scenario of search is completely untenable, but with the evolution of the model's capabilities, the AI search task is now at least usable today.
This transformation from scratch and from usability to optimization is not only a breakthrough at the technical level, but also the result of a deep insight into user needs.
According to a McKinsey report, by 2030, the global economy will be driven by the big model to reach 49 trillion yuan, of which the Chinese part of the economy will increase by 14 trillion yuan.
The huge economic increment includes not only the improvement of existing work efficiency by large models, but also the new scenarios and new business formats brought by new technologies. ByteDance's exploration is a microcosm of the implementation of AI applications, and it is also a common topic that the entire industry needs to think about.
And this is exactly what Tan Cheng repeatedly emphasized at this press conference, a good model must be used by everyone and every company.
#欢迎关注爱范儿官方微信公众号: Love Fan Er (WeChat ID: ifanr), more exciting content will be presented to you as soon as possible.
Love Faner|Original link· Sina Weibo