AI technology stack is too huge! Wu Enda gives career planning: lifelong learning

 Datawhale dried food  

Share ： Wu enda , source ： The new intellectual yuan

Wu Enda recently wrote two letters , by AI Guide the career of practitioners , The most important thing is lifelong learning ！

The rapid rise of artificial intelligence has led to the rapid increase of artificial intelligence jobs , Many people are starting new careers in this field .

But a career is a journey that lasts for decades , And the road is not always flat .

these years , I have the honor to see thousands of students and engineers of large and small companies' career planning in the field of artificial intelligence .

I want to share some thoughts , These ideas may help you plan your route .

Three steps for promotion and salary increase

The three key steps of career development are learning （ Acquire technology and other skills ）, Working on a project （ Deepen skills , Build a portfolio , And create influence ） And looking for a job .

These steps are superimposed .

• first , You need to focus on acquiring basic technical skills .

• After acquiring basic skills , You can work on the project . in the meantime , You may keep learning new knowledge .

• later , You may occasionally apply for a job . In the process , You may continue to study and work on meaningful projects .

These stages apply to almost all occupations , but AI The industry also includes some unique elements .

Artificial intelligence is a new thing , Many technologies are still evolving . Although the foundation of machine learning and deep learning is maturing , And course learning is an effective way to master basic knowledge . But beyond these foundations , Compared with those more mature fields , In the field of artificial intelligence , Keeping up with changing technology is more important .

Project work often means working with stakeholders who lack AI expertise , This may make it possible to find a suitable project 、 Estimate the project schedule and return on investment , And setting expectations becomes a challenge .

Besides , The highly iterative nature of AI projects leads to special challenges in project management . When you don't know in advance how long it will take to achieve the target accuracy , How can you come up with a plan to build a system ？ Even after the system reaches its goal , Further iterations may also be required to solve the post deployment drift problem .

Although looking for a job in AI may be similar to looking for a job in other industries , But there are some differences . Many companies are still trying to figure out what AI skills they need , And how to hire people with these skills . What you do may be very different from what your interviewer sees , You may need to introduce some aspects of your work to potential employers .

In these steps , Finding a support community will be of great help . There are a group of friends and allies who can help you , And the people you try to help , It will make your road easier .

Whether you are taking the first step or have been on this road for many years , Is so .

I'm glad to work with all of you , Develop a global Ai Community , This includes helping everyone in our community develop their careers .

Learning technology

There are more published papers on artificial intelligence than anyone can read in his life . therefore , In the process of your hard study , It is essential to give priority to topics .

I think for the technical profession of machine learning , The most important topics include ：

Basic machine learning skills

for example , Understand linear regression 、 Logical regression 、 neural network 、 Decision tree 、 Clustering and anomaly detection models are important . In addition to specific models , It is more important to understand the core concepts behind how and why machine learning works , Such as prejudice / differences 、 cost function 、 Regularization 、 Optimization algorithm and error analysis .

Deep learning

This has become an important part of machine learning , If you don't understand it , It is difficult to excel in this field ! Understand the basic knowledge of neural network 、 Practical skills to make it work （ Such as super parameter adjustment ）、 Convolution network 、 Sequence model and Transformer It's valuable .

Mathematics related to machine learning

Key areas include linear algebra （ vector 、 Matrices and their various operations ） And probability and statistics （ Including discrete and continuous probability 、 Standard probability distribution 、 The basic rule , Such as independence and Bayesian rules , And hypothesis testing ）.

Besides , Exploratory data analysis （EDA）, That is, using visualization and other methods to systematically explore a dataset is an underestimated skill . I find EDA In data centric AI （data-centric AI） Especially useful in development , Analyzing mistakes and gaining insight there can really help drive progress .

Last , It is also helpful to have a basic intuitive understanding of calculus . In the previous letter , I described how the mathematical knowledge required for machine learning changes . for example , Although some tasks require calculus , But the improved automatic differentiation software makes it easier to invent and implement a new neural network architecture , Without doing any calculus . This was almost impossible ten years ago .

software development

Although you can only rely on machine learning modeling skills to find jobs and make great contributions , But if you can write good software to realize complex AI systems , Your job opportunities will increase .

These skills include the basics of programming , data structure （ Especially those related to machine learning , Such as data frame ）, Algorithm （ Including those related to database and data operation ）, software design , be familiar with Python, And be familiar with key libraries , Such as TensorFlow or PyTorch, as well as scikit-learn.

This is a content that needs a lot of learning ! Even after you have mastered everything in this list , I hope you can continue to study , Continue to deepen your technical knowledge .

I know many machine learning engineers , They are in applications such as natural language processing or computer vision , Or in technical fields such as probabilistic graphical models or building scalable software systems , Benefit from deeper skills .

As for how to acquire these skills ？

There is a lot of good content on the Internet , Theoretically , It is also possible to read dozens of web pages . But when the goal is deep understanding , Reading disconnected web pages is inefficient , Because they often repeat each other , Use inconsistent terms （ This slows down your learning ）, The quality of different , And leave some technical gaps .

This is why a good course , The material is organized into a coherent logical form , It is often the most time-saving way to master a meaningful knowledge system .

When you absorb the knowledge in the course , You can turn to research papers and other resources .

Last , please remember , No one can cram all the necessary knowledge into it in a weekend or even a month . Everyone I know who excels in machine learning is a lifelong learner .

in fact , Given that our field is changing so fast , If you want to keep up with the times , In addition to continuous learning , Have no choice but to .

If you form the habit of studying a little every week , You can make significant progress with less effort .

Reference material ：

https://read.deeplearning.ai/the-batch/issue-151/

https://read.deeplearning.ai/the-batch/how-to-build-a-career-in-ai-part-2-learning-technical-skills/

Sorting is not easy to , spot Fabulous Three even ↓