Section Seven - Algorithms
A story about code, cats and dogs and cooking (4 minutes)
What have we learned so far?
- Data is never objective;
- We influence what we measure;
- More data means more coincidental patterns;
- Correlation and cause and effect are tricky concepts.
This is the quicksand on which we are going to do 'something' with data. That 'something' often has to do with algorithms and/or artificial intelligence.
We are going to talk a lot more about algorithms in the next courses.
For now, we would like you to think about the video you watched in section three. The one about algorithms being just values embedded in code? Remember? If you forgot, you can watch it again below:
So, now you have a firm grasp on the initial fundamental problems with data. You also understand that doing something with algorithms is very problematic. Especially when you are not constantly aware of the underlying problems with data.
This problem becomes even more urgent when we think about new software that can become better at certain tasks on its own. This software is often called machine learning of neural networks. We will talk about them a lot more in the next course. For now we would like you to watch this 2 minute video explaining machine learning with an example of software recognizing cats and dogs.
So, in this example the computer gets better and better when it has data to 'train." In this case pictures of cats and dogs that are labelled cat or dog. But, what if this data is more tricky? Subjective? Influenced? Biased? Flawed? Then the results of the software will also be subjective, influenced, biased and flawed. As we have seen already, all data is subjective, so all machine learning is subjective too.
The difference between a simple algorithm and high-end artificial intelligence like this, is that machine learning is often a black box. It can no longer explain how it reaches certain conclusions.
This is very troublesome, especially when you think about the issues with data. We will talk about the subject in more detail in the next two courses.
Take aways from section seven:
- Doing 'something' with data is even more tricky, because there are issues with the data;
- Being aware of these issues helps thinking about, assessing, designing or using a technology.