Should you choose Data Science, Machine Learning, or Artificial Intelligence?
You've decided that data is the next thing for your hobby or career—but it seems like there are options. The choice is often formulated as Data Science vs Machine Learning, but it is a bit more complicated (and simple) than that.
What is Machine Learning?
Machine Learning is the most mainstream subset of Artificial Intelligence. In this case, artificial intelligence is utilized for the machine to learn and improve algorithms by itself. Solutions are often classified as powered by supervised and unsupervised machine learning.
Let’s take the development of self-driving cars that should be able to recognize road signs. Depending on the amount of data, the general system architecture philosophy, and other factors, an engineer can label road signs on “learning” images (training data) before the “lesson” or make the algorithm try and differentiate between signs on its own. The first solution is faster and more natural—road signs are consistent and new ones are rarely introduced—but it scales poorly. If you want to run the same solution on multiple continents, it would be much better if the algorithm knew that the circle-backslash symbol means “no parking”, even when it has a dark blue background in the UK while having a white background and the letter P in the US.
Various machine learning solutions are employed by hundreds of thousands of companies on a local and international scale. We’ve recently explored how ML is utilized in banking, customer support, and email.
What is Artificial Intelligence?
Artificial Intelligence is the pinnacle of processing data. It is a broader term than machine learning since the computer tries to mimic the human mind in more things than just learning. Among other things, AI replicates our cognitive functions: Natural Language Processing is used to interpret speech while Computer Vision powers the machine to “see” objects. The latter is utilized in self-driving cars.
It would be front to mistake fancy Artificial Intelligence solutions as new developments of the Facebook era. Stephen Hawking was using a text-to-speech device called CallText 5010 manufactured in 1988. As you can see, a beneficial AI solution does not require an unreasonable amount of computing power. We’ve learned as much from our friends at Peltarion: you can analyze tweets in under 15 clicks (14).
What is Data Science?
If we’re being scientifically accurate (not sorry), Data Science is everything about data: not just creating new products with it but also processing, handling, extracting knowledge, and more. Have a look at this chart.
We’ve already described the top two blocks of the pyramid. In a broader sense, Data Scientists can be doing anything from this pyramid (if not everything). The professional terminology, however, seems to be finally stabilizing.
- Learning & optimization is done by Data Scientists or specialized Machine Learning Engineers/NLP Engineers.
- Aggregation & Labeling goes to Data Analysts, although they often dive a block lower. Specialists that work at this level may still be referred to as Data Scientists
- Exploration & Transformation is the bread and butter of Data Engineers
Summing up, when you hear about a Data Science solution or a Data Scientist, it is likely (but no 100%) that there is no AI involved whatsoever. Similarly, a Data Engineer sounds quite fancy but they are actually at the bottom of the data hierarchy.
Which Field to Choose?
AI is a very math-heavy branch of Data Science. The relevant experience there would be an advantage while washed up high school skills will require a lot of studying to get up to speed. It may also prove very discouraging to be trailing behind Computer Sciences graduates for quite some time.
Data Analysis is wonderful if you have an analytical mind without hardcore math knowledge and/or solid business experience. It is more natural to advise your boss on what’s going on with their Lifetime Customer Value if you had heard about this metric before joining the company.
Data Engineering is a solid entry option, especially if you’re unsure about further options. You get to work with analysts and AI specialists, and companies often arrange external or in-house training to retain and promote talented data engineers.
Your advantage could be advantageous for AI, but, if you’re sick of abstract things, it’s probably better to go into Data Analysis. You’ll be more hands-on with the product, the customers, and stakeholders. When switching careers, it’s a good idea to go for something more enjoyable—especially if you’re burnt out.
Machine Learning Engineers are expensive. The same level of experience usually nets you some 20-30% more compared to a Data Analyst. If you live in a developing country or looking to get into IT just after the recession, there will most likely be a shortage of AI-related jobs. Starting as a Data Analyst would give you a leg over fresh graduates who keep waiting for ML opportunities.
All data journeys start with Python. We still have a couple of spots left for the winter Python group. You’ll get a solid foundation for data fields as well as pick up the necessary skills for a Junior Python Developer job.