Assembly complete content portal:
Can you do machine learning without writing code? !
This is amazon Cloud’s new tool for code-free visual machine learning, announced at amazon Cloud’s recent Re :Invent conference.
Amazon re:Invent Conference is the most important event in the cloud computing industry. Amazon’s Re :Invent Conference was renamed after amazon’s Headquarters in Seattle.
Over the past 10 years, this event has produced numerous cloud computing and EVEN AI industry benchmarking products, such as Amazon Mechanical Turk, Amazon Rekognition, Amazon SageMaker and so on.
Amazon Invent 2021 Re :Invent 2021 Amazon Invent 2021 re:Invent 2021 Amazon Invent 2021 re:Invent 2021 Amazon Invent 2021 Re :Invent 2021
Not only the machine learning model of code development, but also the 12 products of Amazon Cloud Technology have covered the entire AI industry chain.
From a horizontal perspective, from free computing power pool for individual developers to AI model optimization tools for big factory professionals, the corresponding release is also available. In addition, considering the rapid development of AI in The Chinese market, Amazon Cloud technology also provides the conference transcripts with Chinese subtitles on site B.
In the live speech, CEO also specially emphasized: “Amazon Cloud technology will provide the most extensive and complete full stack machine learning services.”
Please follow us to review the highlights of the whole process and have a comprehensive understanding of the AI product context of Amazon Cloud Technology.
Machine learning without code
Let’s see if the code-free predictive machine learning service that was mentioned in the beginning really works for people who don’t know code.
Officially called Amazon SageMaker Canvas, the product is aimed at people with zero machine learning experience, whether they are business analysts, hr, finance or marketing.
Predictably, most of these people have no machine learning experience or even knowledge of code, but have a need to use data to gauge current strategies and predict market trends.
Amazon SageMaker Canvas visualizes the steps of a machine learning model as an interactive UI, aiming to solve their business problem by quickly generating a machine learning prediction model without writing a line of code.
To prove its effectiveness, Amazon Cloud’s AI/ML division itself shared a case study.
Among them, the product marketing manager of the department wants to evaluate the current marketing activity through Amazon SageMaker Canvas to determine whether it has enough influence and effectiveness.
Just open up the Amazon SageMaker Canvas and upload the data. In this process, the platform can also automatically correct upload data errors, such as adding missing values or deleting duplicate rows and columns. Its technology is no accident, also from our own AI/ML.
Next, specify the target predicted by the model, and then click “Fast Generation”. The required model can be trained.
It turned out to be a visual diagram, with a model accuracy of 93%.
Once the models are generated, they can also be shared with partners such as data scientists to help business people further examine or refine the models.
After reading the official case, the visual interface does have some brushes
What about the partner experience?
At present, BMW Group has put Amazon Cloud AI/ML technology into more than 600 applications in actual business processes, from production line to sales end, in addition to 15 million connected cars. Millions of kilometers of data a day are analyzed and predicted by Amazon SageMaker Canvas.
Siemens Energy is also one of the early adopters. They use Amazon SageMaker Canvas as a complement to their machine learning toolkit. “Canvas allows us to share and collaborate with the data science team to help produce more machine learning models and ensure that they meet quality standards and specifications,” says a data science group leader in applications.
Many unknown giants are also Canvas users, such as Invista, a subsidiary of Koch Group, the world’s largest unlisted company, which has used Amazon SageMaker Canvas to assist in data science issues in business processes.
After reviewing the results of multiple evaluations and intuitive display, we can roughly judge that Amazon SageMaker Canvas is really worth looking forward to. After all, the principle that graphical interfaces unlock productivity and create value over code has been proven repeatedly in the past.
Free online AI lab
As mentioned above, in the annual blockbuster conference, Amazon Cloud Technology laid down its promise to provide the most extensive and complete full-stack machine learning service. Since it is “the most extensive and complete”, the release of a Single Amazon SageMaker Canvas is certainly not enough
For the majority of academic and research institutions, AI enthusiasts, cutting-edge technology giants also need to live up to their slogans.
To sum up, three words, lower the threshold.
The most intuitive, provide computing power resources.
In recent years, the high price of hardware and complex software configuration have been a barrier for beginners to learn AI, and also a huge obstacle to the development of the industry and the recognition and familiarity of more people.
Amazon Sagemaker Studio Lab offers a large roll of wool. An online lab that requires no additional environment configuration, no account registration, and direct access via email.
In this environment, anyone creating a project can directly have 12 hours of CPU time, 4 hours of GPU time, and 15GB of storage:
This configuration throughout the industry, indeed in place.
Note that 12 hours of CPU time is generally sufficient for data preprocessing with Pandas or XGBoost for classical ML algorithm training. For deep learning training, the GPU backend can also be selected to obtain 4 hours of computing time, which is sufficient to train or fine-tune the model on a smaller data set.
In other words, beginner AI models can be trained for free with the above resources.
At the same time, the most popular machine learning tools, frameworks and libraries are pre-packaged to provide subscribers with the ability to customize the Conda environment and install the open source JupyterLab and Jupyter Server extensions. The lab environment is tightly integrated with GitHub, making it easy to copy and save the projects you create.
In addition to free “online LABS” and computing power resources, another part of the “wool” is more intuitive – scholarships.
This time, Amazon Cloud technology has set aside a total of $10 million to launch an Amazon Cloud Technology AI&ML Scholarship program, which aims to help high school and college students over the age of 16 pave their path to machine learning-related careers.
In addition, Amazon’s 1:18 scale autonomous racing car Amazon DeepRacer is also offering a more fun and less expensive way for autonomous driving and machine learning enthusiasts to get into machine learning and train their own reinforcement learning models.
Amazon DeepRacer is powered by reinforcement learning and can deploy its algorithms in a 3D racing simulator in the cloud, as well as experiencing the thrill of racing in the real world via a physical car.
Of course, high performers also pass through the scholarship program.
Amazon Cloud Technology has also teamed up with Intel and Udacity to launch a joint campaign to distribute 2,500 scholarships to disadvantaged people over 16 years of age in economic difficulties, disabilities and other social conditions.
In addition to financial support, the underprivileged also get a year of mentoring from Udacity mentors, Amazon Cloud tech gurus and Intel tech gurus.
Machine learning “industrialisation” is being reshaped
Whether it’s the launch of zero-code machine learning, or the adoption of universal benefits to a wider audience, the technology is behind it. After all, functional development needs a deep understanding of the scene and technical accumulation, and the “universal benefits” two word test, or the cost reduction level of technical enterprises.
Compared with the above two, Amazon Cloud Technology Re :Invent2021 released for professional practitioners Amazon SageMaker new features, more intuitive display of Amazon cloud technology technology level, from which, more visible technology giants for the future of AI/ML planning.
For extensive MLer, a complete machine learning process includes data preparation, data annotation, training, reasoning, and deployment. The final reasoning effect of the model depends not only on the level of the developer, but also on external factors such as architecture, computing power and data.
In doing so, Amazon Cloud technology wants to reduce the level of influence at the individual level and, in their words, move AI/ML from artisanal to industrial.
Specifically, To solve the problem in a package, Amazon SageMaker offers a combination of approaches covering the whole process of machine learning:
During the data preparation phase, data engineers often need to leave the current development environment and manually configure a cluster that meets the requirements of the running model or analysis.
To do this, Amazon SageMaker Studio integrates with Amazon EMR, You can use SparkUI directly from Amazon SageMaker Studio Notebook to monitor and debug Spark jobs running on Amazon ECR clusters.
Given that you don’t have to leave the environment to perform data preprocessing, development, or model deployment, this is a step toward an ideal fully integrated development environment.
Data annotation stage is also bidding farewell to labor intensive, avoiding artificial to artificial intelligence:
This used to require manual tagging or processing through data tagging programs, but now, after raw data and requirements are presented, Amazon SageMaker Ground Truth Plus will assist human experts in tagging with pre-tagging assisted by machine learning.
This approach reduces error rates and reduces labeling costs by 40%, enabling more efficient error detection and avoiding low-quality labels.
The training phase is more critical.
It is as strong as BERT, the classic deep learning model in the industry. Complex neural networks with billions of parameters need thousands of hours of GPU training. Even if tuning and optimizing, it still needs several days of training.
But now, Amazon SageMaker Training Compiler, a machine learning model optimization Compiler provided by Amazon Cloud Technology, can improve the Training speed of GPU instances without adding too much code.
With the compiler, the training speed of many classical deep learning models including bert-Base-uncased, bert-Base-uncased and distilbert-base-uncased can improve 50% directly.
Finally, the improvement of reasoning stage. Amazon Cloud Technology took out the previously famous concept of “Serverless Inference” and provided a set of Serverless Inference function.
This function can allocate resources to the cloud and enjoy an elastic resource space service in the case of strong fluctuation of data computation volume. Let programmers focus on high-level languages, not the underlying hardware, and let professionals focus on what they’re good at.
Considering that in reality, many customers have special needs, but it is difficult to judge how many computing resources are appropriate, Amazon SageMaker Inference Recommender, another function, provides the recommendation of configuration and actual operation parameters in the Inference stage, so as to find the best balance between cost and speed.
From data preparation to reasoning stage, the product functions of the above processes are released to serve the whole machine learning cycle rather than a single point. Its purpose is to help enterprises realize the large-scale application of machine learning, connecting the dots and opening up an AI/ML industrial-scale application process.
So how did this one-two punch work?
For example, Vanguard, one of the largest fund management companies in the US, can reduce the deployment time by 96%, astrazeneca, a pharmaceutical giant, can complete the machine learning environment deployment in 5 minutes, and NerdWallet, a financial management company, can reduce the cost by 75% on the premise of increasing the original training demand.
In addition, a more diversified landing scene can also see the deep mining of AI/ML by Amazon Cloud technology.
For example, DevOps Guru for RDS can be used to help developers detect, diagnose, and resolve performance and operational issues in Amazon Aurora.
CodeGuru Reviewer Reviewer (S), for example, will identify passwords, API keys, SSH keys, and access tokens in source code to improve code review efficiency and help the traditional software industry improve performance.
Interestingly, CTO Werner Vogels took a break from his time at Amazon’s Re :Invent 2021 conference to post a blog post about the tech guy’s high hopes for the AI/ML industry:
Software development will shift from being labor-intensive, and software development supported by ARTIFICIAL intelligence will dominate.
Finally, on hardware, Amazon Cloud technology also released its own chips, and still rolled out three at a stretch.
The Graviton3 CPU chip features machine learning.
Moreover, the machine learning customized training chip Trainium supports Trn1 instances, providing users with higher cost performance and faster training of deep learning models in the cloud.
Whether it is to open AI/ML industrial-scale application process, or the release of self-developed hardware chips, the macro level —
The above announcement indicates the visible extension of Amazon cloud technology in AI/ML business.
Amazon Cloud technology is pushing the boundaries of AI
According to DATA from IDC, from 2013 to 2020, the global AI/ML annual expenditure has rapidly expanded from 0 to about 50 billion US dollars, which is almost twice the growth rate of cloud computing, the original business of Amazon cloud technology.
It is to see this trend, Amazon cloud technology multi – way attack seems to be inevitable.
From no code machine learning, no server application in-depth AI/ML, to the underlying computing power continue to upgrade, and many inclusive programs… The dizzying rollout shows that Amazon Cloud technology is redrawing new boundaries in machine learning.
Although the above release has not yet been implemented, the value displayed by the general public may not be visible to the naked eye at the moment, but from another perspective, the so-called pursuit of long-term value, the so-called emphasis on infrastructure layout, is not clearly written in the DNA of Amazon cloud technology?
Think back to 1997, when Mr Bezos issued his famous “letter to shareholders”.
When Amazon argued that customers, sales, and brand growth were all for long-term value, bezos emphasized continued investment in “systems and other infrastructure” for long-term value.
Since then, Amazon Web Services has operated independently, turning cloud computing from a “concept” into a real industry. Amazon Redshift and Amazon Lambda have also developed into cloud native data warehouse and serverless route.
Everything seems to have been written in an early prologue.
It’s no surprise that Amazon Cloud continues to bet on AI/ML with the same mindset.
This is not only the responsibility of a technology giant, but also the expectation, which echoes the Slogan “Lead the wind and reshape the future” of Amazon Cloud Technology Re :Invent 2021 this year.
We can already see that the AI/ML field is expanding its population coverage, its industry landscape is expanding, and its technology continues to be explored, a process that is being continued by industry technology pathfinders.
How big will AI/ML be in the future? Little by little, Amazon’s cloud technology is taking a new shape.
Assembly complete content portal: