I worked with Merel on a new HRI Podcast Episode: The Case For Rude Robots. It is surprising how hard it is for a robot to understand politeness.
Author: Christoph Bartneck
Swim Training Patterns books travel around the world
Swimming is much more fun in a community. We share a passion for the sport and I would like to give others the opportunity to enjoy my latest book, Swim Training Patterns. I sent the first two hardcover books I received on their way around the planet. One will travel predominantely westwards and the other eastwards.
The books can only travel by being given from person to person. It should not be send by mail or courier. Given that we are all only six degrees apart, it should not take long for the books to return to Christchurch. Both books contain a tracker, so I can roughly know where they are. Right now, both books are still in Christchurch, but I hope that this will change soon.
If you receive the book, please sign the log and enjoy the book. Once you are done then please simply give it to the next person. This is not a race, so don’t worry if it takes a long time for the book to move forward. If you want to, please send me a photo of the book from where you are.
Motorising the V100 train
I motorised Bluebrixx’s V100 LEGO train with some Mindstorms motors and sensors. The train worked well during the Model Train Show for two days. I documented the major technical challenges and solutions.
Information Is Beautiful Talk
Please join me for this public talk.
Keynote-poster-OMG25-Information-is-beautifulchatGPT visualisation
openAI published a working paper on how people use chatGPT. They include a visualisation that triggered my curiosity.
The width of the columns represents the cumulative proportion of the main category, and the height of each sub-category represents the proportion of the sub-category within the main category.
The first problem is that the percentages of the main categories do not add up to 100.
| Main Category | Percentage |
|---|---|
| Multimedia | 6.00 |
| Other / Unknown | 4.60 |
| Practical Guidance | 28.30 |
| Seeking Information | 21.30 |
| Self-Expression | 4.30 |
| Technical Help | 7.50 |
| Writing | 28.10 |
| Grand Total | 100.10 |
The percentages of the sub-categories, however, do add up to 100. Next, I measured the width of the columns to test if they actually got it right.
I then (ironically) used chatGPT to extract the numerical values from the graph.
Next, I divided the width of the column by the percentage it is supposed to represent.
| Main Category | Calculated Percentage | Column Width | Column Width Ratio |
|---|---|---|---|
| Multimedia | 5.9 | 43.0 | 7.3 |
| Other / Unknown | 4.5 | 32.5 | 7.2 |
| Practical Guidance | 28.3 | 208.5 | 7.4 |
| Seeking Information | 21.3 | 157.0 | 7.4 |
| Self-Expression | 4.3 | 30.0 | 7.0 |
| Technical Help | 7.6 | 54.0 | 7.1 |
| Writing | 28.1 | 207.5 | 7.4 |
While the columns approximate the main category percentages, they vary considerably. So, how can we do this better? The answer is a good old area graph. It scales each box according the the percentage and groups them into the main categories by color.
A new type of visualization is not always a better visualization.