Business World VS Achievement in Machine Learning | AI Specialized News Media AInow
To summarize the characteristics of AI research in the business world, it is as follows.
On the other hand, the characteristics of AI research in the academic world are as follows.
The above features are based on MEHTA's experience, and if you are actually worried about the course of AI research, we strongly recommend that you consult with each person in the business and academic world.。
The following article is translated after direct contact with Purvanshi Mehta and permission to translate.In addition, the content of the translation article is his view, not a specific country, region, or organization, nor does it have expressed the principle of translators and the Ainow editorial department.
Image source: From UNSPLASH's Lubo Minar
table of contents
How did I make a decision
Choosing a career is difficult itself.Comparing the Ph.D. to get a Ph.D. to get 5-6 years with the advantageous work in the business world, it will be a concern for those possibilities.After graduating from the university undergraduate and master's program, I fell into a situation in the same way on my course (both times).
I think there are many readers who know my career path, but fortunately I was led to the lab in the summer of a second -year college student, where I studied a semi -religious extraction.After that, he worked full -time in the lab before entering the graduate school.With this experience, I had a glimpse of my full -time researcher / doctoral student.
However, despite such a research experience, I joined Microsoft after acquiring a master's degree (currently the current workplace).The decision process was annoying, so I decided to write down my final thoughts.Before the readers take this article as one opinion (and how I made this decision), before making a decision before making a decision.I strongly recommend talking to multiple people.
AI evolution level
I worked in full -time in both the academic world and the business world.that is,
Whether it is a business world or an academic world, you can work on exciting things and take part in the AI revolution.The rest is what level of evolution you want to be involved in.
Do you want to transplant an old rule -based system into an ML/DL model?Or do you want to scale the existing ML regression model into a more state -of -the -art one?Or do you want to answer questions that will help you to build a system five to 10 years from now (although it depends on how you apply it), not just right now (depending on how you apply it)?I asked myself about what level I wanted to work in the "evolution of the" from the two perspectives as follows.
Most business research is product -oriented
However, if you belong to a pure study group, it is different (it is difficult to be assigned to such a group just to complete a master's course, or even a doctoral course).Usually, there is a problem setting related to the product and try to find research related to how to solve it.If there is any new discovery in the process, it may announce it.
Most study groups in this industry do not consider papers published in the process of performance evaluation.Therefore, basically you have to increase your motivation to study in -house.
If you do not stick to the problem setting, a research group in the business world may be good.
In my case, I was actively published a dissertation, and most of the members found a doctoral study group.The study group was very suitable for me.
ML research is interesting, but not all research is an impactful.Let's look at that in numbers.
The number of AI -related papers published in ARXIV has increased more than 6 times to 34,736 in 2020 in 2020 ( * Translation 1).The quantity is definitely increasing, but is the quality the same?
The annual transition graph of the number of Neurlps quotes in Neurlps was quoted.With the increase in the number of papers, the average number of quotes in papers is decreasing year by year.In other words, if a paper was published in Neurlps in 2017, the average number of quotes, that is, the number of people using the paper is 4..It will be 6.This figure looks pretty tough to me.
We know that there are exceptional papers, and the average number of quotes is not the largest scale that measures the value of the dissertation, but it is at least an approximate evaluation of the influence of the dissertation.
Some of the papers I wrote have contributed exceptionally (actually leading the field), but there are many factors that I think I contributed to such papers (some of them are instructors,There are labs, research fields, and mental conditions at the time that I was acquiring a doctorate).
（※訳註1）AI関連論文数は、論文に関する統計情報の調査を目的に作成されたウェブサイト「Microsoft Academic」を利用して算出されたと推測される。同サイトを使えば、記事本文に引用された論文の被引用数の推移のほか、以下のような最近5年における人工知能分野で発表された論文において引用されたトピックトップ5に関する推移をグラフ化できる。最近5年における人工知能分野で発表された論文のトップ5トピックの被引用数推移Changes in the business world
In 2019, 65%of the doctors in the AI field in North America entered the business world, 44 in 2010..It has been highlighted that the role played by the business world in AI development has begun to grow in AI development ( * translation 2).
When I worked as an applied scientist on Amazon, all the people in my team had a Ph.D.As I interacted with them, I realized that master's students could do the same work.
The boom of AI development in recent years has been amazing.A lot of engineering is required to make a good product, but it reads the paper and thinks a use case that applies the content, or the first you already have a usage case, and solves the problem that the use case has.You need the skills to read a dissertation.
In fact, acquiring a doctorate in machine learning does not mean that all of the doctoral acquisitions have advanced to research.
What would happen if I worked in a place similar to the current group after I got a doctorate?He would not have found much significance in the entire process (until he got a doctorate).
（※訳註2）記事本文で引用された北米のAI業界における博士号取得者の割合は、アメリカ・スタンフォード大学が発表したレポート「2021 AI INDEX REPORT」を出典としている。同レポートの「4.2 北米におけるAIとコンピュータサイエンス学位取得者」によると、AIに関する博士号取得者の進路は、2011年頃までは学界と実業界が40%程度で拮抗していたのだが、2012年以降は実業界の就職が多くなり、2019年には65.7%となった（下のグラフ参照）。北米におけるAI博士号取得者の進路Furthermore, among the new Ph.D. acquired AI, the proportion of international students from other than North America is increasing, and in 2019 64..It reached 3%(see the graph below).Changes in the proportion of international students in North America to acquire the Doctor of AI
Also, among non -North American students who have acquired a PhD on AI, I got a job outside of North America..While 6%, I stayed in North America and got a job 81..It was 8%(see the graph below).The report suggests that a doctorate acquiring the world AI in North America is gathering.Employment of international students who acquired a doctorate in North America
The academic world is difficult
To be honest, it is difficult to live in the academic world.Such drawbacks are as follows.
However, despite the above factors, I highly appreciate those who work in the academic world.I understand that researchers are seeking novel ideas and working on basic problems.You can also sympathize with the intellectual simulation that is over your head when you find the truth of the problem!
Advantages of academic world
When I have to do the mediocre work in my daily work (welcome to the business world), I still think of returning to the academic world.Doctor's work has many advantages:
Conclusion
After all, it's important to work on exciting themes.Whether it is a product that involves the theme, whether it is a research or a research.I don't know if I will get a doctorate in the future.But if so, you'll definitely be ready.The experience in the business world will add to me what is important and what is important.
Original "Industry vs Academia in Machine Learning"
MediumIndustry vs Academia in Machine LearningHow I took a decisionAuthor Purvanshi Mehta
Translation Yukimoto (Free writer, JDLA DEEP Learning for General 2019 #1 acquisition)
Edit
