Blog overview

Part 2: Maths, Technology and Vicarious Embarrassement

The first contact with AI Programming


Uwe Weinreich, the author of this blog, usually coaches teams and managers on topics related to strategy, innovation and digital transfor­mation. Now he is seeking a direct confron­tation with Artificial Intelligence.

The outcome is uncertain.

Stay informed via Twitter or linkedIn.

Already published:

1. AI and Me – Diary of an experiment

2. Maths, Technology, Embarrassement

3. Learning in the deep blue sea - Azure

4. Experimenting to the Bitter End

5. The difficult path towards a webservice

6. Text analysis demystified

7. Image Recognition and Surveillance

8. Bad Jokes and AI Psychos

9. Seven Management Initiatives

10. Interview with Dr. Zeplin (Otto Group)

Here we go...!

No, we jumped the gun. Before we get going, prepa­ra­tions have to be made: registration on the MOOC Platform, downloading the sample data, installing Python and IDE, setting up an Azure account...

Wait a moment: setting up an Azure account? Yes indeed: it's only possible to take part in the course if one opens an account and thereby becomes an Azure customer. It's fair to say that an ulterior motive is in play here. Hmm, admittedly that wasn't my intention, but at least Microsoft offers a starting credit of $200 (€170), valid for the next 30 days. That will surely be enough time to complete the course, but the time pressure will be constantly increasing...

What is machine learning?

Behind AI lies the ability of machines (computers) to learn independently, by building models and thereafter generating and sorting their own solutions.

Computers being able to do this is nothing new, of course. But this method differentiates fundamentally from the previous approach of solving problems by using algorithms. Let me try to show you in graphic form...



What's the difference? In the best-case scenario, the results are the same. But the effort and outlay differ considerably between the two options, and this means that the two approaches are best used for different scenarios, as the table below shows:

  simple, stable problems complex dynamic problems
Algo-rithms Relatively quick to program, stable problem-solving over time. Long and complex development process, prone to mistakes, every dynamic change creates problems. Possibility of not addressing some aspects of the problem.
AI Enormous outlay (acquisition, processing power, energy, training) to solve a straightforward problem. Possibility of differing results over time despite same input data. Reasonable outlay in comparison to the complexity of the problem. Possibility of discovering solutions (models) that wouldn't have been found by humans. Independently adapts to change.

Therefore one needs to be alert when defining the problem. AI is good for complex, dynamic problems, while algorithms are better for simple ones.

Of course one can also use AI for simple problems. The problem then is that it kind of feels as if one is sitting in a seminar led by an average tutor, who's trying to encourage the participants to find the solutions on their own: "Let's say 1 times 1 is 1, 1 times 2 is 2, and 1 times 3 is 3. So, can you figure out how to continue? What would 1 times 4 be? Who can tell me? Mr Smith, you maybe?" In such situations it's embarrassing to answer, and just as embarrassing not to: a lecture hall full of vicarious embarrassment. We should thus avoid this type of situation with AI as well, and use it only where it belongs, right?


A Microsoft video uses exactly this off-putting example to show how AI learns. OK, but only for this fleeting moment as a demonstration!



The Basics - pure maths

But thereafter it gets more challenging. The process will include computation of regression, data classification, and clustering. Yes, these processes are used in Machine Learning, but at the moment this doesn't have much to do with AI. All of these processes are also known from the world of classical analysis.

Let's see how it goes...


⬅ previous blog entry      next blog entry ⮕

published: June 15, 2018, © Uwe Weinreich


June 18, 2018 - Uwe Weinreich

Installing Python and an IDE is not really necessary. Everything works also fine by just using the online editing options.

new comment