A Rebel Sunflower.
One sunflower is looking the wrong way. The morning sun is in the other direction.
Yesterday I played with Bing Chat, which is Microsoft’s AI engine and I noticed that I could play with generating images. I spent quite a bit of time generating a multitude of images, in part for fun, but also to get a grasp of the limitations of the opportunity presented by software like AI.
If you ask Dall-E via Bing Chat to generate a face then it can. It wanted to generate the face of a woman with curly hair so I did, and the image looks realistic. I then asked it to draw the face of a tired man and it did. It’s when I asked Dall-E to draw a woman cyclist exploring the Dolomites that I saw that faces are a mess. It’s fine with generating human faces, in isolation, but if you ask for a landscape, and a human with a face then the face is wrong. This is both great, and a shame. It’s great because it means that we still need to photograph real people doing real things, but it’s bad because if all we need is a face then AI is ready to provide us with some.
I asked Dall-E to create images of ducks cycling around the Vallée De Joux and it drew ducks, where instead of legs, the ducks had wheels. Sometimes these wheels were not round. In one case a duck had the front forks and a wheel coming out of the front of its body. Dall-E obviously hasn’t spent much time reading Duck Tales stories. If it had then it would draw flawless ducks.
I asked Dall-E to draw the Palais Wilson with the Ballerina in one instance and it failed. It took a generic building and showed it within the scene I described for it to draw. I also asked it to draw the broken chair at Place Des Nations in Geneva and it failed to do this too. Dall-E doesn’t imagine real things we describe as accurately as I would expect. In fact it doesn’t, at all.
I asked Dall-E to draw Firefish providing IT support, staring at a phone, and writing a post. In most of these cases Dall-E worked well. It was able to provide amusing or interesting images. When I asked it to draw a frog waving at a heron it generated one image that was better than the others.
I asked Dall-E to draw people cycling down a cycling route with a railway tunnel and the result was quite good. I then asked for the people to be replaced by meerkats and the result was amusing. If you use your imagination Dall-E can generate some interesting results.
I asked Dall-E to draw a camera operator on a crane filming a flock of geese bobsledding and it drew the camera and the crane well, as well as the geese but it ignored the bobsledding part, and the face of the camera operator was awful.
Rather than write more, I will include a series of images on my wordpress blog for you to enjoy.
In Switzerland a three hour drive feels long, but in France or Spain it does not. I drove from Switzerland to Spain and from Spain to switzerland. The drive to Spain is easier because traffic gets lighter as you get further south. Most of the time this is true. This time, as I drove from Spain to Switzerland I came across the opposite. From the moment I started driving I encountered traffic.
The problem with traffic is that some people are driving above the speed limit, while others are driving below it. You have to decide whether you want to slow down behind the car that is blocking you, or go out, blocking the fast car behind you. I usually let my speed go, rather than risk a high speed rear collision, flashing lights and other unpleasant behaviour.
This time when I arrived at La Jonquera there were plenty of cars so I stayed in the car and skipped this refueling stop, to refuel at the next petrol station in price. The difference in price was big but I thought that saving time was more important than money. I wasn’t wrong.
At the first péage in France I lost 20 minutes, and then I had regular traffic jams until 20 minutes or so before Grenoble. From there traffic was easy, and I even began to find driving between Grenoble and France easier. I have driven that route regularly enough. It has become a relaxing part of the drive.
This was especially true, because of how quiet the road was compared to the rest of the journey. I drove from around 0910 or so to 2220, so a thirteen hour drive. Yesterday I was tired from the drive.
I read two books during the drive. I listened to them with Audible. I listened to The Jewel That Was Ours by Colin Dexter and finished part two of three of The History Of the Ancient World. The Jewel That was Ours is a reasonable book, although I do not feel that it is ideal for road trips. It requires focus and attention, two things that driving in heavy traffic also require.
I think that listening to books where you can half listen are better suited to long drives. Big Mile Cycling by Sean Conway was a better choice for the drive down. I will see about finding similar books for my next long drive.
Today I listened to two Le Cours De L’Histoire podcasts. One was about Cretinism and Iodine deficiency, and the other, which I still need to finish is about vertical exploration of the mountains. For a while I found that I had no interest in listening to podcasts but for once I felt ready. I try to find pandemic friendly content, that explores ideas, concepts and history, rather than self-pity. I don’t think self-pity is the right term, but something along those lines. I like to read and listen to things that distract me from pandemic solitude.
It has taken a while, and some effort but I finally feel that I am getting to understand JavaScript better, and that I am more confident with it. I am not rushing, but rather seeing the code, and re-writing it with my own terms, words and more. I am pushing the envelope slightly, to deepen my learning. Today I played with encoding letters to learn about functions. Tomorrow I will experiment with something else.
Tomorrow I get my booster, so we will see how it affects me. I want to eat a raclette, to see whether my dreams are more interesting, afterwards. 🙂
In the near future we may see many more self-driving cars from many different manufacturers. If this is the case then we should think about reducing the number of roads and lanes devoted to cars. As robots drive cars, rather than humans so safety distances no longer need to exist, because every car is talking with every other car, as a swarm.
As I listened to a podcast about drones one of the people said “they write about swarms of drones but these are not swarms, for it to be a swarm each drone should talk with every drone and they should work as a single entity. Given this context I look forward to when we live in a world where cars are programmed to be part of a swarm, rather than individual units. This is because if cars drive themselves, they can be much closer, at a higher speed, whilst still being safe. It means that rather than have two or three lanes of traffic you would have just one, per direction. Imagine a motorway where cars behave as train cars, rather than driven cars.
If you look at a map of most parts of the world, whether countryside or cities, you will see that there are five or more roads exiting every village. With self driving cars you could reduce the number down to one or two. The other roads could be given over to pedestrians, cyclists, horse riders, skateboarders and rollerbladers. That’s because humans are human, and humans can be predictably irrational.
With self driving cars they just run programmed. This means that traffic can be synchronised to avoid having to stop, or feel any doubt. A machine runs as it is programmed to run, so it should not deviate from its course.
Although machines just run as programmed humans do not, whether cycling, walking or other. I do not like the idea of having self-driving cars driving by me with no human at the wheel. it’s bad enough when humans are at the wheel. It would be worse with machines.
With a car that drives itself you do not need to have it parked a short walk from where you are. With a self-driving car it can park in a place that is optimal for when you request it to come back, but not in the way of others. At the moment, in summer, with human driven cars, when they are not in use they take up large parking spaces. They also take up pavements and cycling lanes. If cars park themselves then pedestrian lanes and cycling lanes can remain free of cars in summer.
I started off with the idea that we should separate self-driving cars from pedestrians and cyclists but came to the conclusion that by having self-driven cars the need for roads will decrease as automated cars can drive along designated routes. Roads that were once used by cars would now be free of them, and they could be dedicated to healthier walking and cycling.
In my view it is a mistake to look at how to get city people out of their cars, because city people don’t need cars. When I lived in London I never felt the need for a car, but I deeply regret that no one encouraged me to get a bike and enjoy London that way. Imagine taking a bike rather than a night bus.
The issue is from villages to towns, and from village to village. Between villages and towns the distance is a kilometre or more, and if you walk over a kilometre on a dangerous road it is unpleasant. With self driving cars you could route traffic along two or three roads, and release the other roads to cyclists and pedestrians. Imagine having a slow up along certain routes every day. In fact you don’t need to imagine it. Via Verde, Voie Verte etc. are old train lines that are now wide spaces where cyclists, pedestrians and other users can be safe from the noise and pollution from cars.
For weeks, or even months, by now I have been playing/experimenting with Hugo, 11ty and other solutions. I really like that with Hugo I can use FrontMatter as a CMS to create new posts, add the appropriate meta data, and keep track of what is published and what is in draft form. It allows me to create posts with the right metadata in seconds, rather than having to write the date, time, draft status and more by hand. It also generates the right file title for good archival practices.
As I was looking for a CMS tool to make managing 11ty content easier I came across Decap CMS and it seemed interesting. I installed a version locally, and then I started to look at the code manually, rather than using the CMS tool. It felt complicated so I did some more research. Eventually I learned that in order to play with Decap CMS you need to setup a netlify account, a github account and then expose yourself to accidental charges when playing with a static website generator. I was struck by the paradox. Why would you use a CMS tool that requires you to commit to an external hosting tool? Why not use ClassicPress or WordPress and cut out the middle man. Of course the short answer is “because you still generate a static tool, but the interface is intuitive for non coders.
By requiring us to set things up via Netlify we’re forced to use yet another service, which is fine, when you’re using the service in the first place. I am not.
Within a few minutes frontmatter.codes could be setup locally do do what I want, to manage documents and frontmatter for an 11ty site. In so doing I keep development on the local machine, only connecting to the external server when I’m uploading site changes. I can use the same workflow as I have for Hugo, once I set it up.
It’s easier, for me to setup a ClassicPress or WordPress CMS and use that. ClassicPress feels very fast and I can use markdown or html for pages that I am creating, or that already exist. Within a short amount of time I can do what Decap CMS does, anywhere I want.
For WordPress you can use this method/tutorial or with the free playground option. Within seconds you can have a wordpress instance running on azure, up and ready for a new site and content.
In particular, while App Service F1 will not generate any cost, database usage is chargeable for “pay as you go” plans or when the usage limit of 750 hours per month for 12 months is exceeded. So, in order to ensure they will not pay for the WordPress playground, developers should monitor and track their database usage.
With this tool a wordpress instance is prepared for you, and for a month you can see what the cost would be, before jumping into a financial commitment.
If I am experimenting with a Static website generator like Hugo or 11ty I want to have local versions to play with, rather than remote ones that may cost something if I am not careful. If I’m reading it correctly the basic plan I’m experimenting with is 3 CHF per month for a server in Northern Switzerland. With this “playground” I have the opportunity to experiment, and see whether that is the case.
The testing options are cheap, but for production Azure and other cloud solutions are expensive, which is why we use other cloud solutions, especially for personal sites. I will spend time experimenting with Frontmatter, set up for 11ty, following this learning experience.
I had a theory that if I wanted to I could transport a raspberry pi running ubuntu server from one place to another and connect by wifi, with a little tweak, or by ethernet if that didn’t work and today that thought was proved wrong. I spend at least an hour experimenting, before calling it a day, because of lunch time, rather than a loss of desire to find a solution.
I had the theory that if you plugged an ethernet cable into a Pi it would become visible on the network with ease but this didn’t seem to be the case. The Pi wasn’t happy. It was configured for one wifi hotspot but when I tried to change the config file for a second wifi access point it simply didn’t work.
There are two things I learned. The first is that raspi-config works when you have Raspian installed. it doesn’t work when you have a flavour of linux installed. The second problem I encountered was that the tools which various sites, including Bard and Chat GPT recommended required a working connection to download the tools that they recommended. I tried this, but had no luck, because for some reason the ethernet network was not detected.
The lesson is that I should download those tools before migrating a Pi from one location to another, so that if I encounter such an issue I have a simpler UI that works, rather than modifying config files by hand, and failing. I wrote failing but this wasn’t a failure. My entire goal was to take a Pi, configured for one environment and transpose it into a second. It didn’t work and I familiarised myself with the tools I need to install before attempting such an experiment again.
Don’t forget. One of the strengths of Pi is that you can download instances and install them on microSD cards. You can attempt something with one card, and when it fails swap it for a second and try a variant, until you succeed. If you succeed you can remove that SD card, run an experiment, and see if that experiment works, before reverting to the original card with the install that works.
The beauty of Raspberry Pi, and staring and stopping instances is that you can try various ways of doing the same thing several times, and make mistakes, and invest time in learning how to problem solve, before achieving your goal and moving forward. In this situation the experiment was the goal, so my ideal outcome was not reached, but the opportunity to learn was.
With today’s experiment I learned that to make a Raspberry Pi running Ubuntu easier to port from wifi network to wifi network, for example to backup a phone while traveling, I need to add the network manager tools. If I had done that my day’s experiment would have been a success. I also learned that raspi config is not on Ubuntu installs via the Raspberry Pi imager, and this is a useful thing to learn. Now I can see whether I can recover the system before having dinner.