Few thoughts on Artificial Intelligence and Machine Learning
Lately, machine learning, deep learning and other concepts which relate to data being used in order to adapt itself automatically have been very popular.
Although this is pretty cool to be able to make an algorithm adapt itself automatically and this can be practical in some situations when humans do not have the capability to created the statistical models to use in order to analyse the available data, this is a mistake to think that this will replace every other data analysis techniques anytime soon because of some critical points.
First, what are Artificial Intelligence, Machine Learning and Deep Learning?
I guess that I have already scared people less familiar with data science techniques. Anyway, thanks if you’re still with me.
Here are a few definitions and examples that may be useful:
Artificial intelligence (AI), simply put, is the ability to automate a decision using data. It is not a technology but rather a philosophy, as data and statistics are very old concepts, but the leap of faith to automate actions only based on data (and ignoring human instinct as a result) is a very modern approach to decision making.
Applied to a large set of data, AI can be very advanced such as powering autonomous cars, which can decide to break or accelerate based on real time feedback from thousands of variables. It can also be less advanced as well. For example, In NHL ’99 , a very popular video game, the other team and other players were already powered by artificial intelligence!
Machine Learning is a subset of artificial intelligence, which is based on live feedback. An example is an anti-spam, which could block an email based on the user’s preferences (the user is not opening emails talking about penis enlargement… makes sense) or the sender’s behaviours (if the sender send a .exe file, let’s prevent a potential virus).
Finally, Deep Learning is a more advanced and futuristic technique. A known application of Deep Learning is for image mining. For example, it can find the subject of an image by analyzing each pixels separately, and then together to be able to compare images together. Therefore, Apple can know that you are in Paris when you take a picture of the Eiffel Tower, Facebook can know who to tag on a picture, etc. Some more serious use case are in radiology, in topography and in anti-terrorism.
This automation sounds great. Why do you say this is not always a good idea?
Ok now that we understand the different buzzwords in data science, let’s understand why those automation are not always desired in business.
- What’s going on in my business
How can you evaluate your business if you do not understand the most important variable for your business? What if you do not know your customers? Awkward. Netflix in one perfect company in that sense. Although everything on your screen adapt itself based on your preferences and behaviours, the statistical models are all reviewed and adjusted by humans at Netflix. This way, the whole company can focus on ways to satisfy their customers, provide complimentary videos, offer new products, … There are so many opportunity to learn from data. Unfortunately, so many Machine Learning and Deep Learning techniques’ outputs cannot be deciphered by Data Scientists. In few cases, we do not care, in most business contexts, this is a big deal.
- Do I have enough data?
AI and machine learning are very powerful when you have enough data to explain and predict a certain behavior. In other words, if companies do not collect enough or the right information. it will have an hard time predicting it automatically, which makes AI and Machine Learning irrelevant.
In conclusion
Some AI use cases make a lot of sense and need to be promoted as they are disruptive and huge opportunities. However, the data science field of study still needs to focus on being able to interpret the output for better understanding.
At an higher level, many people foresee robots replacing us anytime soon when they read newspapers and online articles. For data analysis, I think this phenomenon of artificial intelligence will create new opportunities. However, analysis and insight generation will remain a key aspect going forward.
