I’m mixing it up this week by cornering one of my machine learning engineer colleagues to answer all the burning questions on machine learning, becoming a data scientist and the future of AI.
Pierre-Antoine is a colleague of mine based in Luxembourg. He is, among other things, a machine learning engineer and developer. We caught up a couple of weeks ago and I was able to ask him all my questions about machine learning and AI.
Let’s dive on in shall we?
What is artificial intelligence?
The dictionary artificial intelligence definition says:
Artificial Intelligence is the theory and development of computer systems able to perform tasks usually requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
Want to learn more about applications of machine learning? Check out this article on the benefits of artificial intelligence.
What is the state of artificial intelligence today?
People today often assume that artificial intelligence is at the level of human intelligence.
This is not the case.
The reality is that AI and the application of AI in machine learning are very narrow in what it is able to do.
The algorithms that are developed are very good at doing a specific task that they have been highly trained on.
For example, the algorithm that allows Alexa and other assistants to understand your speech and respond, are not able to pick out features of images.
What is machine learning?
A machine learning algorithm is a computer program that has access to data and uses it to learn for itself.
What is the difference between Machine Learning and Artificial Intelligence?
Machine learning is one application of artificial intelligence that is particularly popular now. Machine learning is not part of the artificial intelligence definition.
In machine learning, you give the machine (processor) the data and let it try to learn the solution to the problem or task by applying statistical methodologies (link to the definition of statistical methods) to the data.
What’s the most interesting problem you’ve seen solved using machine learning?
The Amazon Go store.
This is a fascinating case.
The way that they have been able to really eliminate false positives and negatives with such a high accuracy using visual ML is amazing.
What are the different types of Machine learning model?
There are three main types of machine learning model:
- Supervised learning. This is where you have labeled data to train the model, and you use this training data to predict the most likely value of a new item. Supervised learning then falls into two categories:
- Classification: Predicts the class of an object, i.e., if a product is a chair or not.
- Regression: Predict a value, i.e., house price values over time in different areas
- Unsupervised learning – This is where you do not have the labeled data, and instead you ask the algorithm to create a cluster and have a lot of news and cluster related items
- Reinforcement learning – where you have an agent in the algorithm that is able to learn from its environment
7. What is the difference between deep learning and machine learning?
Machine learning models are inspired by the neural network in the brain. In these models, each ‘neuron’ in the neural network layer tries to mimic what the brain does.
Deep Learning is a subcategory of Machine Learning (which spans across classification, regression and reinforcement learning)
However, 5-10 years ago, the development of enhanced computer processing power allowed people to experiment by adding multiple layers of networks into machine learning models. This can improve the model by making them very efficient for problems that require a lot more data to solve. This layering effect of neural networks forms a deep structure, and hence, deep learning.
How long does it take to train to become a machine learning engineer?
It really depends on your background and types of programs you want to create. There was recently an example of a high school girl who won a competition by writing a machine learning algorithm.
If you have some programming skills and the right documentation, it can take as little as two weeks for you to be able to start creating models.
There are even services by AWS Machine Learning that allow you to upload an excel file of data and they will then predict output values for you.
However, of course, if you want to work on more complex problems, it will take longer to learn.
I have heard a lot about python being the programming language to know if you want to get started with machine learning. Why is python an excellent language to learn if you’re going to learn ML?
Python has some great readily available libraries that allow a machine learning engineer to implement machine learning algorithms easily:
- Scikit Learn contains machine learning algorithms
- NumPy is excellent for mathematical computations
- TensorFlow, MxNet, PyTorch, Caffee2 are all used for Deep Learning (they are all more or less equivalent)
How do you train yourself on ML?
I studied machine learning briefly at University, however, keep up to date I use a variety of different resources:
YouTube has excellent resources – including courses by good Universities
- This channel has great introductory videos on Neural Networks: LINK
- Excellent tutorial by Fei Fei Li & all on Deep Learning: LINK
What’s the most exciting part of being a machine learning engineer?
See the algorithms in action and getting to play with it.
What’s the most boring part?
Definitely, data clean up and preparation, but it’s essential.
What jobs and tasks are being automated using machine learning by machine learning engineers?
A lot of the statistical analysis to do with loan approval will be automated. You also are increasingly seeing admin work in law is becoming automated.
What jobs are being created because of ML?
Data Scientists and machine learning engineers!
What book do you recommend for people who are interested in learning more about machine learning?
The Fourth Industrial Revolution by Klaus Schwab, who is organizing the world economic forum. It’s a fascinating read about how machine learning is shaping tomorrow’s society.
Thank you so much for your time Pierre-Antoine!