Din ce în ce mai mulți studenți CS sunt interesați de AI – și nu sunt destui lectori

Departamentele de informatică din universitățile din SUA nu au suficienți lectori pentru a preda un număr tot mai mare de studenți interesați de IA, a sugerat luna aceasta un raport al Centrului pentru Securitate și Tehnologie Emergentă (CSET).

Interest in machine learning and artificial intelligence has risen and fallen since the field was formally founded in the 1950s. Neural networks have made a comeback in recent years, exploding in popularity with deep learning. Demand for machine-learning courses at universities has skyrocketed, we’re told, and there aren’t enough lecturers to support students’ interest.

Date compilate de Sondajul Taulbee, și citat în raportează, a arătat că între 2011 și 2020, numărul studenților înscriși la programe de informatică în America s-a triplat de la 60,661 la 182,262. Dar numărul profesorilor din departamentele de informatică a crescut de sub 1.5 ori de la 4,363 la 6,230. Raportul total studenți-facultăți în cadrul departamentelor chestionate sa dublat de la 14 la 1 la 29 la 1. 

Acum, pentru a fi clar, acestea sunt statistici care reprezintă toți studenții înscriși în 140 de departamente de informatică din SUA, mai degrabă decât acei studenți care se înscriu în mod special la cursuri de AI, deși raportul susține că cifrele indică o creștere a interesului pentru cursurile care conduc în primul rând predarea învățării automate. Rezumatul executiv a concluzionat:

While there is an undeniable rise in those taking computer science, you have to take the report’s word for it that this in all probability equates to a rise in interest in ML. “Many AI courses are taught in computer science departments, and AI specialists account for a growing portion of CS faculty overall,” the report noted in an appendix.

Some universities have had to cap the number of students for particular classes due to a lack of teaching staff. Limiting education will have a detrimental effect for the United States, the report’s authors Remco Zwetsloot, a fellow at the International Security Program at the Center for Security and International Studies, and Jack Corrigan, a research analyst at Georgetown’s CSET, explained.

“Teaching capacity gaps limit the amount of talent flowing into the US AI workforce, which in turn negatively impacts economic and national security,” they wrote. “Research has shown innovation is partly a function of the absolute number of researchers in a particular field and the act of generating new ideas is becoming more labor intensive. Less talent therefore means less innovation.”

Experts in AI have previously warned that universities are suffering from a brain drain of talent. Instead of going into academia, they’re moving towards research positions in industry due to higher salaries and access to better resources, leading to fewer tutors in colleges.

But Zwetsloot and Corrigan believe the data shows this isn’t the whole picture. It’s not that universities are struggling to hire faculty to support more students, it’s that they aren’t hiring at a fast enough pace. Some academics wooed by industry often continue to stay at their departments and only spend 10 to 20 percent of their time working for a company. 

“We found little evidence to suggest the outflow of AI faculty from academia to industry has increased in recent years, and, though a larger share of new PhD graduates is indeed taking jobs in industry, survey data does not indicate that they are disinterested in academic careers. However, we did find evidence that suggests universities have not increased the number of computer science faculty positions in line with the growing demand for AI-related education,” the report said.

Percy Liang, an associate professor of computer science at Stanford University, however, told us: “It is true that the number of available faculty positions has not grown nearly as quickly as the number of industry positions, but I think brain drain is real: researchers choose industry over academia or leave academia to go to industry because of higher compensation, more data and compute.”

Între timp, Zachary Lipton, profesor asistent de învățare automată și cercetare operațională la Universitatea Carnegie Mellon, a spus Registrul he doesn’t see a huge brain drain of researchers going into industry. After working for a couple of years for a company, many often return to academia.

“Yes, there is more pay in industry but it’s boring in a sort of way,” he told us. “Their focus is more myopic. There are more important interesting problems in foundational, theoretical research that are still best studied in academia.”

Lipton said the surge of interest in machine learning is for introductory courses that cover the basics, and these classes are useful for a wide range of careers outside of academia. There isn’t as much demand for advanced graduate-level study. To cope with the increased demand, universities should boost teaching faculty rather than researchers seeking tenure. 

“Universities should make the teaching track more attractive,” he told us. “These faculty members don’t have to worry about grants or running a lab, but it’s very hard to accept a pay cut just to focus on teaching. Tenured academics may teach some introductory courses but their primary focus is research. We need to find more people that have a passion for teaching, who are able to connect with a broad base of students.”

Raportul sugerează că guvernul SUA ar trebui să intervină și să crească finanțarea pentru universități, astfel încât acestea să poată angaja mai mulți profesori. Ar trebui să existe mai multe opțiuni în afara mediului academic pentru ca oamenii să își îmbunătățească abilitățile și să urmeze aceste cursuri introductive la AI la colegiile comunitare sau online. Sectorul privat poate ajuta, de asemenea, donând universităților, continuând să finanțeze granturi și sprijinind noi posturi academice. ®