To Hire or Not to Hire Online AI Grads: That Is the Question

It all started with a tweet from Google Japan Data Project Manager Suzana Ilić: “Yesterday someone (ML, CS PhD, Stanford) said he would not hire a person who is online educated in Machine Learning. Who here agrees and who thinks differently?” The question triggered a long and occasionally heated discussion that spread from Ilić’s twitter across the machine learning community.

Typically, candidates for ML research positions are expected to have a strong academic background in CS (computer science), ideally from a respected tech university. Carnegie Mellon University, Massachusetts Institute of Technology, Stanford University, University of California at Berkeley and Tsinghua University are global leaders for ML, and these schools’ grads are heavily targeted by tech giants and startups alike.

As Ilić pointed out, “any job description related to ML requires a (technical) degree, in many cases a PhD (and I’m not even talking about research, but engineering). Without a degree, one would probably not even pass the resume screening.”

However, with the rapid evolution of ML both in labs and the real world, online courses have emerged as an increasingly popular choice for those interested in the field. The question for Ilić is whether “a formal degree provides something essential that cannot be replaced by online education and if so, what it is?”

online education

A tremendous number of online AI education options have appeared in recent years. MOOC (Massive Open Online Course) platforms such as Coursera and Udacity are now wildly popular across the ML community. AI guru Andrew Ng’s Stanford University machine learning course remains the most popular on Coursera, the world-leading online education platform he co-founded in 2012. In 2017, Ng launched Deeplearning.ai, a specialized deep learning education project which so far has attracted over 250,000 enrollments. Ng’s courses are aimed at students or engineers with mathematics and computer science backgrounds. The Deeplearning.ai “AI for Everyone” project meanwhile is designed to help non-technical business professionals such as CEOs, product managers, marketers, designers and financiers better understand AI and what they can do with it.

Udacity was also born out of Stanford University, when two instructors began offering their “Introduction to Artificial Intelligence” course to the online public for free. More than 160,000 students in more than 190 countries have since enrolled in Udacity AI courses. Aiming to offer a specific skill set for professionals who wish to bridge the gap between learning and career, Udacity programs are decidedly job-oriented. Udacity has also teamed up with Georgia Tech to offer an online Master´s degree in Computer Science.

Kaggle, an online community of data scientists and machine learners, offers Kaggle Learn as a short-form AI educational tool that covers topics such as deep learning, machine learning, Python and so on; to teach “practical data skills you can apply immediately… to become a data scientist or improve your current skills.”

Fast.ai meanwhile offers a variety of free courses in coding, software library, research, etc.: “the world needs everyone involved with AI, no matter how unlikely your background.” Courses “covering computing from its fundamentals to fully-deployed, scalable applications” are also available from the Lambda School with no upfront cost.

Opinions

Ilić’s tweet naturally caused concern for the many people studying AI online. For her part, Ilić said she would “absolutely hire someone without a [traditional] degree.” The Founder of fast.ai and former Kaggle President Jeremy Howard agreed, tweeting “I never had a formal technical education” and “I didn’t actually go to any lectures or tutorials. I thought they were a waste of time.”

Creator of the Keras open source neural network library François Chollet chimed in: “It’s an antiquated and elitist viewpoint. There are no ‘online’ and ‘offline’ educations, nor ‘formal’ and ‘informal’. The best people are 90%+ self-educated, whether they have a degree from Stanford or not. The value-add of degrees in CS is increasingly marginal.”

 

Lambda School CEO Austen Allred let figures do the talking: “Five $130k+ offers for Lambda School grads this week. None of them had a four-year degree. None of them had prior professional experience. None of them had ever made more than $30k before.”

Google AI ML researcher David Ha presented his personal experience: “I was online-educated in Deep Learning. I started from Andrew Ng’s ML MOOC, Stanford tutorials, to Hinton’s MOOC on Neural Networks.”

Many on the other side of the debate pointed out that new hires bring risks, and those from established universities can be a safer bet for companies.

Twitter ML Engineer Sijun He drew from personal experience: “MOOC still doesn’t provide the same rigor as a formal education yet. i.e., I have taken Andrew Ng’s ML course both in person and on Coursera and the difference is huge. To cater the class to the mass audience, most of the math was taken out and the focus on intuition instead.” He also suggested what matters most is interview performance.

The discussion revealed how ML professionals and the AI community in general view the emergence of online education opportunities in the field. In an informal poll in Ilić’s twitter thread, only 18 percent of respondents agreed with the idea of not hiring online grads, while 82 percent disagreed. One thing is certain: As online AI education programs continue to turn out new grads, to hire or not to hire is bound to become an even more controversial topic across industries.

---

Journalist: Fangyu Cai | Editor: Michael Sarazen