Exploring Yunqi: Omnipotent AI, Serverless at Your Fingertips

Original source: CSDN

Author: Wu Xingling

Image source: Generated by Unbounded AI

Enter "Help me write a plane battle game", and the AI programming assistant Tongyi Lingcode completes more than 100 lines of code in just a few seconds; The Tongyi large model series family was collectively unveiled to show the application of large models in various fields; Experience 5 minutes to quickly build a merchant user system; Use AIGC to design personalized short sleeves and become the most beautiful cub in the audience; You can get a variety of AI photos with one click with your mouth; There are also large-scale immersive hands-on labs and experience areas, so that every developer can get their hands on the latest cloud technology...

That's right, the 2023 Apsara Conference is here! With the theme of "Computing, for the Value of the Uncalculable", this year, Yunqi has set up two main forums and more than 500 independent topics, comprehensively covering the technical topics of "cloud + AI", and at the same time opened three theme pavilions: Computing Power Pavilion, Artificial Intelligence + Pavilion and Industrial Innovation Pavilion, exhibiting more than 3,000 latest technologies, products and cutting-edge applications.

After in-depth exploration of this year's Yunqi, the biggest feeling is that Yunqi has made cutting-edge AI and cloud technology within reach, which is no longer a lofty concept in people's impressions, but allows developers to truly experience and enjoy these future-oriented black technologies. As the saying goes, "it's better to see than to hear", the following editor will take you to the exhibition immersively and explore what black technologies are available at this year's Apsara Conference?

All-powerful AI, new applications for future-oriented large models

As a well-deserved "top stream" in the science and technology circle this year, as soon as you step into the artificial intelligence + hall, you find that it is crowded, and everyone is talking about and trying out the latest achievements and applications in the field of various AI large models.

The full range of products of the Tongyi family was collectively unveiled. In the past 6 months, Tongyi Qianwen 2.0 has made a huge leap in performance, compared with version 1.0, Tongyi Qianwen 2.0 has significantly improved its ability to understand complex instructions, literary creation, general mathematics, knowledge memory, and hallucination resistance. At present, the comprehensive performance of Tongyi Qianwen has surpassed GPT-3.5 and accelerated to catch up with GPT-4. In addition, Tongyi Qianwen APP has been officially launched in major mobile application markets, and you can experience the latest model capabilities through the APP.

In addition to Tongyi Qianwen, the 8 major industry models trained based on the Tongyi large model also attract everyone's attention, including the personalized character creation platform Tongyi Xingchen, the AI reading assistant Tongyi Zhiwen that helps understand and summarize obscure papers, the health assistant Tongyi Renxin, the understanding of legal documentsTongyi Farui, and the intelligent customer service Tongyi Xiaomi, the intelligent investment research assistantTongyi Dianjin, and the Tongyi Tingwu that processes more than 50,000 audio and video every day**, At present, the cumulative number of users exceeds 1 million. The most surprising thing for the editor is the AI programming assistantTongyi Lingcode.

The Tongyi Spirit Code officially released today, the exhibition area is very hot, and many developers are in the same mood as the editor - gearing up and eager to try. Xiaobian tries to make Tongyi Ling code write bubble sorting, explain code, generate code interpretation, and generate unit tests, almost all of which are one-click generation of the desired code. Xiaobian continued to challenge it and asked it to write a program for a mini-game of plane battles and a snake, which was basically completed in a minute or two.

Judging from these performances, the editor would like to say that the Tongyi spirit code can be benchmarked against GitHubCopilot. As an intelligent coding assistant based on Alibaba Cloud's general code model, Tongyi Lingcode can automatically generate line-level/function-level code according to the current code file and cross-file context. You can also describe the functions we want in natural language, and generate code directly in the editor area; Unit tests can be generated based on frameworks such as JUnit, Mockito, Spring Test, unit test, and pytest. In the past, it took several minutes to write tests manually, but according to Xiaobian's on-site observation, the test code generated by Lingma at a millisecond speed has a good coverage rate. Every time a developer has a headache with code comments, click "Generate Code Comments" to generate method comments and interline comments with one click, saving us time in writing comments and improving code readability. Look, isn't it convenient:

According to the on-site staff, Lingma supports JetBrains Family Bucket and Visual Studio Code, and supports more than 30 mainstream programming languages such as Java, Python, Java, Type, C/C++, etc., which can automatically identify programming languages and generate code interpretations. In addition, the Tongyi Lingcode brings together a lot of knowledge in the R&D field, and when developers encounter technical problems, they can also evoke the Tongyi Lingcode with one click, and they can quickly get answers and solutions.

It is reported that Tongyi Lingcode has been applied on a large scale within Alibaba Cloud. Now, all developers can experience Tongyi Spiritual Code! Tongyi Code is open to all developers: give it a try!

The exhibition hall also showcased many best practices in the era of large models, and there are rich cases of Alibaba Cloud large models in transportation, electric power, finance, education, intelligent search, automotive, medical and life sciences and other fields. From these displays, Xiaobian found that the large model has covered many areas of our life and work, and it is thriving, and the future is indeed here.

After visiting the Tongyi exhibition area, I came to the Magic Community exhibition area. As the first Chinese model open source community in China, the Magic Community currently brings together top domestic AI large model players such as Baichuan Intelligence, Zhipu AI, Shanghai Artificial Intelligence Laboratory, and IDEA Research Institute to open source the core large models in the Magic Community. With more than 4,000 AI models and 2.8 million AI developers and more than 100 million AI model downloads, there is no doubt that the community has become the largest and most active AI community in China.

At the scene, the editor got started with the Face Chain model of the community, said the photo style he wanted in a conversational way, and uploaded a photo to get an AI photo. After taking a photo with the computer's camera, I upload the photo, and an AI photo will be generated after a minute or two. To be honest, this is the AI photo that the editor has experienced that is most like himself, without over-beautification and cookie-cutter formulaic processing, it is more real.

With large models and large model communities, how can we better help developers develop large model applications in one stop? Move to the exhibition area next to itAlibaba Cloud Bailian, which shows us that Alibaba Cloud Bailian has integrated mainstream high-quality large models at home and abroad, providing services such as model selection, fine-tuning training, security suite, model deployment and full-link application development tools, simplifying complex work such as underlying computing power deployment, model pre-training, and tool development for users.

So, how much time does it take to develop a large model app? According to the on-site staff, with Alibaba Cloud Bailian, developers can develop a large-scale model application within 5 minutes, and "refine" an enterprise-specific model in a few hours.

Xiaobian has to sigh: the era of large models is really coming, many developers and enterprises are actively embracing, such as CCTV, LongShine Technology, AsiaInfo Technology and other enterprises have developed exclusive models and applications on Alibaba Cloud Bailian, LongShine Technology has trained an exclusive large model of electricity on the cloud, and its development of intelligent assistant for power bill interpretation, power industry policy analysis/data analysis assistant, for customers to receive 50% more efficiency, reduce complaints by 70%.

Immersive, hands-on experience with cloud technology at your fingertips

With the advent of the AI era, it is inseparable from the blessing of powerful computing power. Walking out of Hall B, enter Hall C with the theme of computing power, displaying Alibaba Cloud's cloud technology stack in an all-round way, from the underlying infrastructure to platforms and services, it is worth mentioning that a large-scale immersive hands-on experience area is provided, allowing developers to fully experience Alibaba Cloud's full-stack serverless products.

In the experience area of "Alibaba Cloud X Semir AIGC T-shirt Design Activity" surrounded by crowds, Xiaobian tried to deploy Stable Diffusion based on Alibaba Cloud Function Compute FC with one click, and after entering the keyword "squirrel", AI generated a favorite picture, and Xiaobian only needed to wait a few minutes to get a DIY exclusive T-shirt.

In the Serverless Hands-on Lab showcase, Xiaobian experienced the process programming CloudFlow, which can quickly build the system through simple drag and drop, and simulate a transaction system in which a user places an order and the merchant receives the order in a few minutes. It is believed that developers can simplify complex business processes and workflows on the cloud based on CloudFlow, and easily build applications on the cloud.

What surprised me the most was that there were multiple hands-on labs, the SAE Hackathon Challenge, and the Serverless Productivity Academy area, which was convenient for developers to get started with Alibaba Cloud Serverless products. Have you tried a hackathon that doesn't require you to write code? Developers can adjust the optimal resource utilization by adjusting the application parameters and SAE console parameters under the unified traffic pressure test, and the real-time score of each developer is displayed on the big screen, with a strong PK atmosphere! In order to help everyone fully understand Serverless technology, in the Serverless Productivity Academy experience area, learn new knowledge of product technology for 30 minutes, and receive electronic certification on the spot. We try to quickly deploy web applications through ACR, and the steps of the experiment are clear, step by step to create experimental resources ECS, activate container image service ACR, install Docker service, etc., and finally deploy web applications.

Currently, Alibaba Cloud has more than 40 serverless cloud products, including Function Compute for Compute (FC), Serverless Application Engine (SAE), PolarDB database, and E-MapReduce. Among them, Function Compute has been upgraded to version 3.0, the cold start has been reduced to seconds, the resource cost has been reduced by 92%, the complexity has been reduced by 40%, and the integration has been deeply integrated with the Tongyi product series, the Magic Community, HuggingFace, etc., and the deployment time of AI applications has been reduced by more than 80% with one-click deployment. In addition, Alibaba Cloud also announced that the price of all serverless products will be reduced, and the price of serverless versions of big data, databases, middleware, etc. will be reduced by 30%-70%.

In order to ensure the stable and efficient parallel computing of large model training, Alibaba Cloud has upgraded its artificial intelligence platform PAI to effectively support the training of the Tongyi large model series. In addition, at present, Alibaba Cloud has become the public AI computing power base of China's large models, and half of China's large model companies are on Alibaba Cloud, such as Baichuan Intelligence, Zhipu AI, 010000 Things, Kunlun Wanwei, vivo, Fudan University and other leading enterprises and institutions. For the development of global computing power, Alibaba Cloud has built 89 cloud computing data centers in 30 regions around the world, providing more than 3,000 edge computing nodes. This year, Alibaba Cloud also successfully supported the high-intensity traffic explosion of Miaoya Camera in a short period of time.

As can be seen from the above, in the AI era, Alibaba Cloud is continuously upgrading its cloud computing system through full-stack technological innovation from the underlying computing power to the AI platform to model services, creating the most open cloud in the AI era.

** In the 15th year of Yunqi, developers from all over the country are here! **

This year is the 15th year of the Apsara Conference, which has witnessed and led China's cloud computing from its inception to its current emergence as a new infrastructure. Over the past 14 years, Apsara has attracted 700,000 industry professionals and developers from more than 70 countries around the world. This year, technology enthusiasts from 44 countries and regions around the world came as scheduled.

As always, there were many young developers at the scene to experience black technology or listen to the sharing of bigwigs, including some silver-haired participants and primary and secondary school students. There is a 71-year-old CTO who is still struggling in the front line of technological innovation, he is very interested in digital human technology, and hopes to discover the application of technology in the fields of community management and elderly care services at the Apsara Conference. 10-year-old Sonam Choddin crossed 3,129 kilometers from Zhiduo County, Qinghai Province to Yunqi just to witness and learn the latest science and technology.

The Apsara Conference is like a window, open, cutting-edge and interesting, allowing technology enthusiasts from different regions and different ages to participate in and witness the new development of China's technology. With the development of large-scale model technology, the curtain of the intelligent era has been slowly opened, and the Apsara Conference allows developers to see and touch the new future of intelligence.