Preface
Recently, I chatted with a senior sister about artificial intelligence. Like me before, she was interested but didn’t know where to start. So, I updated the previous article I published on Zhihu, combined with some recent new ideas.
I must admit that I am still an amateur myself. I don’t even know how much reference value this path has, but I can assure you that I wrote it with sincerity.
Why Recommend These Materials?
Non-CS background, zero basics, after using many learning materials, I have carefully selected these, which all meet the following criteria:
- At least studied seriously more than 4/5 (I only recommend after I finish and feel good about the course)
- Case-based, with practical explanations (purely theoretical courses are hard to get into)
How to Allocate Your Energy?
You can watch the foundational courses one by one or together. They are relatively boring but can help build a systemic framework, understand principles, grasp concepts, and stop being afraid when seeing technical terms. Later, combine your own direction with practical courses, reading others’ papers to attach “muscles, nerves, and blood vessels” to the framework.
Each course notes the estimated study time. It is not recommended to spend more than the upper limit on one course because extending the study period too long (like studying python for one or two semesters) can make you forget earlier content. It’s better to concentrate your time and finish the system quickly. After all, our starting point is “understanding,” not “mastering,” and our goal is “using,” not “creating.”
I have also noted my actual study time (roughly). I do not recommend studying in less time than I did because I had seen other similar courses before, so I had some basics and skipped some contents, making it faster for me.
Cultivating Interest Basic — Estimated Time: 6-7 hours
It’s well known that interest is the best teacher. If you learn based on interest, efficiency improves and learning becomes more fun.
This is a podcast series about the history of artificial intelligence development. It has nothing to do directly with specific technologies but can cultivate your interest and help you get familiar with some terminology.
https://www.xiaoyuzhoufm.com/collection/episode/6427ac43c5aa3e738cb122e9
Python Syntax Basics
Python Programming Design_Beijing Institute of Technology
Estimated time: 2 weeks to 1 month (I spent 2 weeks)
Introduction to python, environment setup, case-based learning to get a preliminary understanding of the python system. After finishing, you can start programming and understand some code written by others. For example, after finishing this course, I learned some python automation and completed my first project: iTutor—a tool to help you select tutors.
I personally think after finishing, it’s good to do some small projects related to your interest. One, to solidify the language foundation; two, to improve your interest in python. Otherwise, directly starting machine learning and taking two courses consecutively is very boring 
.
Machine Learning Basics
Washington “Machine Learning Foundations: A Case Study Approach”
Estimated time: 1-2 weeks (I spent 3 days)
This is the first course of a series. You can complete all four courses in the series, which will make all professional terminology easy to understand. If you find the last three courses very challenging, you can pause for now and come back later.
Andrew Ng’s Deep Learning Specialization
Estimated time: 1-2 weeks (I spent 2 weeks)
With a python foundation, you can start machine learning! Andrew Ng’s deep learning course is quite popular. I missed it by accident and instead took the Washington course. Both teachers teach together, with interesting dialogues and memes
. Most importantly, this course also combines case studies, teaching knowledge points while practicing. Difficult knowledge comes later to keep your hands busy and keep interest high.
Note: Don’t worry if you don’t understand the math knowledge in the courses; just finish watching. We are not algorithm scientists and don’t need to understand everything. Knowing basic knowledge and being able to use others’ algorithms well is enough.
Both courses have their own focus. It’s better to watch both, but watching just one is okay. After completion, you’ll have a good understanding of AI’s main basic concepts and various algorithms.
Machine Learning Medical Practice
After learning basics, it’s time for practice. Here we use medicine as an example; other fields should have similar tutorials.
DXY “Application of Artificial Intelligence in the Medical Field”
Estimated time: 3 days to 1 week (I spent 3 days)
Based on specific literature, it explains how AI achieves those results, which fields are suitable for AI, and ideas for designing topics. It’s like a stepping stone for practice, giving a certain understanding of professional practice after finishing.
After this course, you basically know the technical threshold but still lack real hands-on experience. Another bigger problem is designing projects: What problem are you solving with AI? This requires long-term clinical experience and communication with others.
Find a Teacher for Guidance and Solidify the Basics (I spent 1 month)
I found a teacher from the management school specializing in AI. Under this teacher’s guidance, I finished our medical statistics elective course, relearned commonly used algorithms like linear regression, logistic regression, decision trees, and combined with medical field papers to consolidate foundational knowledge. Below is the study path the teacher arranged (left to right) 
https://kdocs.cn/l/crtYGAxNrDo0
Although aiming at medical image recognition, medical statistics seemed useless but helped me solidify tool use skills (numpy, pandas, matplotlib, etc.), and also revived old math knowledge (I hadn’t studied math for 4-5 years; now I can’t even calculate derivatives )
Continue with Teacher’s Guidance to Expand Frontiers (I spent 1 month)
In summer I signed up for the teacher’s advanced class. By then the foundation was solid. I started expanding frontiers (transformer, multimodal, etc.) https://forum.beginner.center/t/topic/295
After school started, I worked on projects with the teacher. Some progress has been made but this is just the beginning, with a long road ahead. The difficulties I encountered and solutions are also recorded https://forum.beginner.center/t/topic/485
Reviewing History
Overall, it’s based on interest, learning while using, confused all the way until someone pointed me to a place to apply the knowledge. What I learned before was not in vain, and now I have started producing some results.
More detailed reviews see https://forum.beginner.center/t/topic/427