The right words
The right words
It is Machine Learning? Is it Artificial Intelligence? (Does it quack like a duck? 🦆)
Two weeks ago I presented the Flood Forecasting project at a Vivatech booth. Flood Forecasting is a non-commercial project by Google Research that predicts river floods in 80 countries up to 7 days in advance, helping local institutions and NGOs to prepare for emergencies.
One visitor's question stuck with me: “Does this project use Machine Learning, or is it Artificial Intelligence” ?
Given the amount of time we spend discussing both Machine Learning and Artificial Intelligence (especially in this newsletter), let’s try to align on what these concepts mean.
Machine Learning is computers learning from examples
Whenever we humans explain something, we use one of two approaches:
giving instructions, e.g. “in order to make scrambled eggs, first open the fridge, then find the eggs, then …”, or
providing examples, e.g. “this is a cat, this is a dog”
The same goes for programming computers. Traditional programming is giving the computer instructions (in a programming language such as C++ or Python). On the other hand, Machine Learning is any technique that allows us to explain something to a computer by showing it examples.
Machine Learning is in a way a “more developed” way of programming computers, in the sense that it relies on the traditional instruction-based programming as the foundation. We use traditional programming to build Machine Learning systems, which can then learn from examples.
Artificial Intelligence
While Machine Learning has a very crisp meaning I covered above, Artificial Intelligence is shrouded in somewhat foggier cloud of different meanings.
To me, Artificial Intelligence is any automatic system that aims to emulate a task normally associated with human intelligence. It could be programmed the traditional way (like the Deep Blue chess program that beat Garry Kasparov in 1997), or the Machine Learning-way. So Artificial Intelligence is a broader term than Machine Learning.
I like the slide form Cassie Kozyrkov excellent Introduction to ML and AI:
Artificial Intelligence is when a computer emulates human intelligence by any method ( and regardless how well / badly it’s doing :))
Machine Learning is when a computer emulates human intelligence using a method that finds patterns in previously given examples.
Deep Learning is a specific method of Machine Learning – more on this in a future post 💫 !
(At the same time, much of the industry conversation these days seems to use a different meaning of “AI”, e.g. here “AI engineering” seems to mean “Applied ML” )
So the answer is…
So to answer the Vivatech question about the Flood Forecasting project: yes, it’s both Artificial Intelligence and Machine Learning! Because the project makes predictions (a task associated with human intelligence), it is Artificial Intelligence. And because it uses a model that was trained on historical data, it’s Machine Learning too.
If you’re new to the newsletter, join us below to follow along ⬇️
Learn more
Flood Forecasting by Google Research
🎞️ Introduction to ML and AI - MFML Part 1 – great introduction to ML and AI concepts, friendly to technical and non-technical audiences
Postcard from Paris
Days are long and warm in Paris. This weekend was marked by unrest widely reported in the media.
Have a great week 💫 !
– Przemek