Report: The number of students who want to learn AI has increased by 200%, and there are not enough teachers

丰色 发自 凹非寺
量子位 | 公众号 QbitAI

如今,With the comeback of neural networks and the boom in deep learning——

let want to learnAImore and more students,以至于There are not enough teachers in college.

一份报告显示,从2011年到2020年,computer science major学生人数增加了两倍,从60661人增至182262人.

But computer science majorsThe number of teachers is from4363人增加到6230人,增幅不到50%.

也就是说,The average teacher taught14个学生,Now a teacher wants to teach fast30个学生.

The teacher is more tired,The teaching quality of the student to obtain more than it was before.

所以,Why the number of teachers in computer science has not kept up？

老师数量x1.5,学生数量x3

This report comes from an American think tank“安全与新兴技术中心”(CSET).

该调查收集了140Data from multiple public and private university computer science departments.

负责人表示,Survey although there is no dedicated statistical machine learning course,But the data from the computer science department is enough to illustrate the trend.

正如开头所说,2020Years of computer science students tripled from nine years ago,不管是本科生、Both master's and doctoral students are on the rise.

The number of teachers is also increasing,But the increase is far from keeping up with the students.

This situation has made teachers complain that their workload is increasing,Less time for research.

The quality of education for students also suffers：

In response to this situation, many schools,开始Raise the bar for admission,or cancel small classes.

就比如2018年时,In the survey is in the early years of school60%of schools limit the number of students in popular courses、61%of schools have or plan to cancel small class sizes、42%of schools have or plan to improveCSprofessional entry threshold……

Why this imbalance？

Not all because of the flow of talent to industry

Experts have previously warned,Because industry pays higher wages、easier access to resources, etc.,Universities face the dilemma of brain drain.

此话不假.

一方面,确实有很多University professors poached by some tech companies.

The most famous cases include2015Uber in a few months fromCMUDig near robotics lab40名专家,2013年Yann LeCun成为Metahead of artificial intelligence research.

其中,LeCunStill retains a teaching position at NYU.

这种“dual affiliation”A very common relationship,Professor a lot of people only10-20%time to work for the company,但总的来看,There are more examples of full-time employment in industry.

另一方面,AI相关的博士毕业生虽然越来越多,But the number of people entering academia20In the coming year basic do not have what change,Entering the industry but growing rapidly.This trend can be clearly observed from the curves in the table below.

CRAhas criticized the industry for admitting PhDs before they graduate..

然而,The loss of talent to industry is not the whole reason.

事实上：

（1）调查显示,过去二十年,“By poaching”的现象并没有显著增加,A common cause of loss of teachers colleges and universitiesInstead, retirement and job changes within academia.

（2）同时,The industry also has a small part of the talent flow to learn the industry.

2020年,有32%The school said,Their teachers are hired by industry;同时,也有38%of schools hire experts from industry as teachers.

同样值得关注的是,Some teachers choose to go back to academia after a few years in industry,他们的理由是：Industrial wages are high、资源多,但无聊,短视.

This has alleviated the brain drain of university teachers to a certain extent..

（3）另外,For many PhD students,their career choices,does not represent career preference.

相反,调查显示,Many PhD students interested in academic research and teaching work more.

In this reportanother important reason,Is the school recruitment speed not to catch up with him.

看下面这张图就明白了.

从2006年到2020年间,The difficulty of recruiting teachers in colleges and universities has remained basically stable,not much increase.

由此我们可以推断出：

The most plausible explanation for the relatively slow growth in the number of teachers,is not recruiting enough.

那么,Is the school deliberately not recruiting?？They are fully aware of the shortage.

报告指出,Mainly because of lack of money.

总的来说,The cost of computer science is much higher than that of humanities,But government funding has grown slowly,For many schools with rigid budgets,to accept more students、After spending a lot of money on more courses,You can only pay for the additional teacher recruitment costs yourself.,Or reduce the subsidy for the original teacher.

The implementation of both plans faces huge internal resistance.

于是,The number of teaching positions has not kept pace with the growth of students.

几点建议

If this shortage is not resolved,后果也很明显,其中,Worst impact for PhD students.

调查显示,近年来,Ph.D. enrollments grew the slowest compared to undergraduates and masters,This is closely related to the growth in the number of tenure-track teachers.

Due to limited tutors,Many applicants who did not publish in top journals were rejected,这个要求其实很高.

反过来,Ph.D. training slows down,It will further restrict the supply of qualified teachers in colleges and universities.

At a deeper level,This phenomenon isAIIn the field of long-term development is also very bad.

最直观的表现,Teachers spend all their tasks on teaching,Where's the time for research?？

After all, the industry pays more attention to profit,Create a fundamental progress is less likely.

所以,应该怎么办？

The report concludes with some recommendations：

（1）学术界：简单粗暴,增加经费;

（2）工业界：Encourage employees to teach at school at the same time;

（3）政府：Provide schools with more data and basic computing facilities,In this way, talents will not go to industry because of insufficient school resources.;

……

最后,Does this trend also exist in the country?,Please speak freely.

原报告：
https://cset.georgetown.edu/publication/ai-faculty-shortages/

参考链接：
https://www.theregister.com/2022/07/07/comp_sci_students_ai_lecturers/