spring recruitment season, the ai topic is still hot.
No, Ant Group’s spring recruitment poster this year was produced with the assistance of AI technology. The comment area was very lively as they talked about how difficult it is to submit resumes during the recruitment seasons for gold, silver, and four, as well as about the AI technology achievements conveyed in the posters.
Nowadays, the development of large models is in full swing. It is not difficult for fresh graduates to want to participate in the process of this era of AI. It is not difficult to understand.
Why did Ant Group choose to put these technical people on the spring recruitment poster?
After taking a closer look, I discovered that there is actually a lot of hidden information behind these six posters. So graduates, let’s take a look at each one first and then make a decision. What information is revealed behind the
ant poster?
At first glance, these six posters record the daily work of Ant technicians.
is different from the work scenes we are familiar with every day. Some of them work with satellites in farmland; some participate in the Asian Games billion-level project; and some deal with our basic necessities of life in the streets...
uses this method to attract people. Graduates, nonono, that really can't be so superficial.
After some digging, I found that the selection of these six people as spokespersons for spring recruitment into colleges and universities was indeed based on Ant's own considerations.
Let’s first take a look at the departments where these people work:
ai innovative R&D department nextevo;
basic intelligent technology department;
terminal technology department;
strategic line innovation business;
wealth insurance business group service and sales intelligence department;
privacy intelligent computing department.
's innovative business, technical departments, and industry application business groups are all covered. In the specific technical field, various technologies such as privacy computing, AI large models, AI infra, terminals, and digital humans are also included.
Okay, okay, isn't this a proper show of technical muscle?
Looking carefully at the general content of this poster, I found that there is actually a lot to say.
For example, in this poster, it looks like this guy is standing in the field holding an iPad, but do you believe that he is actually injecting large model technology into the satellite in ? !
This is an online merchant bank "Big Tit" project that Wang Jian from the Basic Intelligence Technology Department is participating in.
With the support of large models, the remote sensing satellite's ability to identify crops in a wide range and with high precision has been enhanced . In this way, people-land-plant data are connected to support the development of a new paradigm of agricultural credit.
In the past, it was difficult to obtain labeled data for these crops, and there were many long-tail varieties, especially cash crops, which affected the efficiency of agricultural credit. But this happens to be a scenario that is naturally suitable for multi-modal implementation.
Through unsupervised training of massive all-landform remote sensing image data, large remote sensing models can effectively learn high-precision and high-generalization feature representations. On long-tail classic varieties with few samples and complex terrain, large-scale and high-precision recognition is achieved.
also relies on core technologies such as spatio-temporal decoupling architecture, multi-granularity contrastive learning, and large-scale geospatial context information enhancement to break through the problems of complex terrain, sparse samples, and poor generalization in operational identification.
In the past year, they have achieved the ability to identify 4 types of cash crops and agricultural facilities , covering more than 20 million cash crop farmers.
At a time when the Thousand Model War is in full swing, they have already given full play to the value of large models in the intersection of agriculture and finance.
Go to the fields and then to the bustling streets.
Ma Chenguang, an algorithm engineer in the Terminal Technology Department, deals with small and micro businesses. The problem he wants to solve is how to let them experience cutting-edge technology in the most accessible way, thereby helping them reduce costs and increase efficiency.
seems to be a simple creative poster generation, but it actually requires a lot of technical challenges behind it.
For example, the adaptation between the image processing capabilities of large models and the performance of various mobile phone terminals; the fool-like operation method that is accessible to everyone, and the ability to truly meet their marketing needs. In addition to industry applications,
has also been implemented in the c-end field.
For example, this poster is the wealth and insurance team of Ant Group.They have created a professional AI assistant for practitioners in financial scenarios: and , which are based on the Bailing large model.
Version 1.0 includes AI assistants in six different fields, including "Service Expert Edition", "Investment Research Expert Edition", "Claims Expert Edition" and "Insurance Research Expert Edition", supporting financial market research and judgment, automated coding, content generation and writing, and intelligent Q&A and other functions. It can be said that professionals in different scenarios have been taken care of.
and the amount of work that can be completed is quite large. For example, take the investment research support assistant, which can assist analysts to complete the financial logic and point extraction of more than 100+ research reports and information with high quality every day, and the reasoning and attribution of 40+ financial events.
Work efficiency straight up~
There are also service support assistants and AI assistants that can help financial consultants and insurance agents expand their effective account management radius by more than 70% per person.
Like this poster, the digital torch bearer of last year’s Asian Games must have impressed many people. The same is true for Xian Ying, product director of ’s strategy of innovation first, .
Alipay's technical staff wrote 200,000 lines of code to create a web3d interactive engine, supporting hundreds of millions of digital torchbearers around the world, and jointly completed the first digital ignition ceremony in the history of the Asian Games. The engine
can render a 3D digital human only through a small program, and it can be adapted to different mobile phones. And combined with AI and other technologies, from personality face shaping, dynamic capture to clothing design, etc., two trillion digital human images can be created. Even if everyone in the world participates, thousands of people can have thousands of faces.
has talked about so many down-to-earth applications, and now he will also show you the hard-core ones. There are two posters of
"leaking the secret" this time, one represents privacy computing , and the other represents ai infra.
First of all, it is no exaggeration to say that privacy computing is Ant’s specialty: the number of patent applications ranks first in the world, and it is at the world’s leading level in terms of technology layout, innovative breakthroughs, and implementation practices. The Linyu team behind it is what supports Ant’s hard core strength.
But even though it is so hard-core, it actually has many wide-ranging uses.
Cui Jinming, an expert on privacy computing open source framework algorithms, has polished a replicable barrier-free identity authentication technology solution to minimize the disclosure of information about visually impaired people, and developed an "accessible online viewing platform" for this purpose. "Shadow" program allows them to enjoy the convenience of digital life like ordinary people.
now relies on data-driven large model intelligence to emerge, and the issues it contains such as user privacy protection and data security require privacy computing technology to protect it. The Linyu team is sharing the technology with the society through open source, which took 7 years to build a trusted privacy computing product.
Let’s look at ai infra. The development of large models requires strong infrastructure support. Improving computing power efficiency and optimizing video memory capacity are all challenges that need to be solved.
Zhang Ke's team has developed a complete set of innovative systems, which can make the effective time of 100-calorie training of the Ant large model account for more than 97% , and the computing power utilization rate of 100-billion-parameter model kilo-calorie scale training can reach 2% .
In terms of video memory capacity optimization, they have self-developed a GPU memory + transmission optimization open source project glake, which can achieve integrated optimization of video memory and transmission at the bottom layer of the training and inference framework and the system layer. For example, in the evaluation of mainstream large models such as gpt, glm, and vicuna,
has expanded the available video memory by up to about 33%, which means saving about 25gb of video memory on the 80gb a100 gpu, and increasing the training throughput by up to 4 times.
is such a secret show of AI technology muscle, which actually shows an idea of Ant AI technology layout.
uses the most cutting-edge hard-core technology to solve our daily practical needs .
includes AI applications that are widely used in agriculture, finance, remote sensing, digital people, marketing, etc., as well as hard-core technology layouts, such as AI infra and privacy computing. For current graduates who want to embrace the dividends of AI, star companies like Ant that are not commonly known may also be a good choice.
How fresh graduates embrace the AI bonus
On one hand, the annual spring recruitment is in progress, and on the other hand, the continuous progress of large-scale models has made the world sleepless... Participating and witnessing this era-class opportunity may be the heart of many fresh graduates. direction.
However, there are many popular large model companies, and the competition is fierce. We have seen a lot of such feedback on social networking sites.
A consensus emerged among this that instead of rushing to the current star companies, scenario companies have more potential and are more cost-effective .
Since its inception, after one year of AI technology development, AI has accelerated its trend of application implementation.
In this new round of technology trends, scenario companies have a faster closed loop of technology and business model innovation.
is an intuitive reflection of AI applications such as ants. When the world is still fighting for thousands of models, they apply large models in finance, marketing, agriculture, insurance and other fields. In this process, they have also accumulated rich scene data to drive iterative upgrades of large models.
The development of science and technology throughout history is actually the scenario where companies finally seize opportunities and enjoy dividends.
Take the previous generation of AI wave as an example. The four little cv dragons attracted so much attention in the early stage, but now it is the scene players in specific fields who are still benefiting from it.
may have more stable development prospects for fresh graduates.
Secondly, don’t underestimate companies in these scenarios. Their AI content is actually very high, and employees may have more application practice opportunities.
Maybe, he can also participate in global-level billion-level projects ~
, for example, like the Ant AI digital man that debuted at the Asian Games last year.
At that time, the "Asian Games Digital Torchbearer" launched globally by the Hangzhou Asian Games Organizing Committee started an online torch relay event. Each user can have a unique digital identity in the digital Asian Games world, and the total number of participants ultimately exceeded 100 million. Behind
is Ant's self-developed web3d interactive engine. It has been previewed several times during the Spring Festival Five Blessings event, with over 100 million participating users. You must have experienced it without even realizing it.
Joining these companies can also participate in global hot events. While watching the excitement, you can work and make money at the same time. To be honest, it is quite fun ~
Therefore, fresh graduates do not need to worry if they do not get offers from big model star companies. Companies in these scenarios have greater potential and are worth looking forward to.
Besides, technology is developing so fast now that many scenario companies have emerged. They are fully embracing AI and growing rapidly. There will only be more employment options and more opportunities in the future.
For example, Ant has also launched a top technical talent training program " Ant Star ", focusing on core technical fields such as artificial intelligence, privacy computing, and security technology. Every year, it recruits top fresh doctoral graduates from well-known universities around the world, and a small number of outstanding academic achievements. of fresh master's degree graduates.
They are mainly concentrated in Ant Technology Research Institute and Ant Group AI Innovation R&D Department nextevo, engaged in the research and development of cutting-edge technologies in the AI field. The starting point for entry is level 13, which is 1-3 levels higher than the starting point for enrollment in other schools; the annual salary package for entry starts at 1 million, with no upper limit...
In short, in the AI era, new opportunities for young people are coming one after another~
— Finished—
