With some of the top tech entrepreneurs in the U.S. either dropping out of college or not attending, there is some debate about whether college is the right choice or not. This post will focus on college for data science. However, for college in general, if you know what you want to study, then college or graduate school is a great option. If you are going to college because you do not know what else to do, I would say college is too expensive for that.
College?
Most would agree that an undergraduate degree in some highly analytical field (math, CS, economics, physics) is definitely beneficial. Plus college has a strict set of guidelines and a specific order for the learning. A formal degree program often provides the necessary motivation for a person to continue learning. The U.S. college education system is not perfect, but if it keeps a person from quitting, it will help to reach the goal of becoming a data scientist.
All this leads to a second point. Only a few colleges offer undergraduate degree programs for data science. Thus, graduate school or more learning will still be required. College should provide the necessary prerequisites and many employers will pay for the continued learning.
No College?
A highly motivated person could probably learn most if not all the data science skills on the internet for free or very low cost. The key is being a highly motivated person. That person must have the drive to not quit when the learning becomes difficult. Also, there are no classmates or professors to help with difficult concepts. Sure, the internet can help there, but it requires a bit more work to find the help. Plus, knowing what topics to learn and in what order can be challenging. Already, this blog has much helpful content, but it is not organized based upon a sequence of learning. Not attending college presents some obstacles that only the most highly motivated students will overcome. As more and more learning resources appear online, the no college option may become more popular.
What is the Answer?
Strictly speaking, I would say the answer is NO. However, many people will not succeed without the rigor of school, and some companies will not hire a person without a degree. So, college is not 100% essential to being a data scientist, but for many it is probably the best option.
It’s nice to read a new perspective on this topic. The amount of free and open source resources in data science make it seem like one of the better areas to get started in. Here are some resources that seem like they’d be helpful to someone wanting to go solo:
* Free online Udacity courses provide elements of a Stanford education in some CS topics.
* OpenIntro offers free introductory statistics resources, including a free textbook — by association, I’m obligated to plug this resource 🙂 (But I wouldn’t be involved if I believed in the quality of the resources.)
* StackOverflow is a great resource after getting stuck somewhere in programming, and it is highly educational just to browse around and see what others are asking.
* Meetup is an excellent, free way to interact with learn up-and-coming technologies and network with real people… nice way to land that first job, even without a degree.
(I have no doubt this is the tip of the iceberg.)
This said, I agree that an unmotivated individual probably won’t learn much without classes to sit in. But I’m also unclear how much unmotivated individuals learn at college as well. Too bad we can’t run a randomized experiment.
David,
Thanks for the additional resources. I have heard good things about Udacity, but I have never used it. You will be happy to know, that I had already planned to mention OpenIntro in a post later this week.
College does provides enough motivation for some people to learn. A truly unmotivated individual would probably have difficulties learning anywhere. That is why I think it is best to have a motivation (know what you want to learn or major in) before going to college. As a thought (and kind of a joke), would the overall incoming freshman class at a university be random enough for the experiment?