Artificial intelligence and more

(Picture: Walldorf near Heidelberg, photo by Andreas Roesel, permission granted)

(You are reading the English original, translations are available here: Deutsch)

Following my last post, the nitty-gritty work began to actually set up the company and relocate. As always, what seems straightforward on paper, can easily turn into many tasks that require a lot of labor.

Needless to say, in the last four weeks there wasn't really time for software development. As a bare minimum I would have liked to playtest another great map for Heroes V and add it as a recommendation to our strategic map pack, but there wasn't even time for that. But as my mind is always connected to the realm of AI, I still managed to work on its fundamentals. Somehow, maybe through a combination of the change in location and the stress, I could make large strides in this realm. More on this further below.

Apologies to everyone who is waiting on my work.

What happened

I am leaving Hamburg and are relocating to Heidelberg. Who doesn't know Germany, the distance is nearly 600 km, one city in the North and the other in the South. So you can't simply hop in your car in the morning, check out a few interesting flats in the afternoon and be back in the evening. You need to plan ahead, and ideally find a solution for a temporary home while you look for a permanent one. Luckily, a friend could help me here. I am very grateful for this.

Deciding how you split your household, the part that remains in Hamburg initially and the part that you take with you, is a logistical challenge itself. I needed to make sure that I have everything needed for living and also to set up my work environment. I am not the type who can do everything from a mobile phone or tablet. Sorting all my stuff plus packaging and loading everything that was coming with me on a transport took more than a week.

Many people, also a lot of people I met in Heidelberg, see Hamburg as a very beautiful city and asked me why I leave it behind. But if you come from Northern Germany, where the landscape is mostly entirely flat, you appreciate the steep hillsides and climbs with their lush forests and the different weather this brings. It has its own beauty.

Finding a place to live in Heidelberg and the surrounding area is hard. Of course there are offers for flats to rent, but going by statistics there are at least 7,000 flats lacking in the town, so that you usually have 70 or more people who express an interest in renting any reasonable property. And as a guy who is self-employed you are always at a disadvantage because the landlords usually ask for a proof of income.

Luckily a friendly and very competent estate agent could provide me with a good background about the local housing market and great practical tips. Nevertheless it was hard work to identify good properties and to sign a rental agreement, which together took nearly three weeks.

Add to this the time for sorting out your stuff, plus the time that is still needed to move everything into the new home, and you are out of commission for six weeks. It's hard for an entrepreneur for whom time is everything.

The bright side

As explained in my last post, moving to Heidelberg makes sense as a place to set up and grow my company, Tesla Minds.

Another big plus is that I can play my favourite team sport here, Ultimate Frisbee. Of course there are many places in Germany where you can, but the situation is different from town to town, and it is definitely favourable here.

Once everything is set up, I believe a lot of things will develop much faster and further here than these would have in Hamburg. A lot depends on a good choice for the place where you live and how you set it up. Distances to the places that you visit frequently matter a lot. That was another reason why finding the right flat wasn't easy.

AI fundamentals

Of course a lot of people, from all walks of life, know little about how AI works. But a lot of people, who call themselves AI experts or AI programmers, don't know much about the fundamentals either. AI programmers in the games industry usually have a limited repertoire of algorithms they use, and machine learning experts may be able to train neural networks to spec, but don't know much about fundamental AI beyond that.

What I am doing is much closer to the fundamentals, i.e. I research how AI in general works and what technological barriers need to be solved to make big strides.

A lot of this is maths and envisioning the fabric of space and time, how reality forms from quantum events. It is a big field and there are many angles from which you can tackle the challenges, but the goal is of course to master this field sufficiently that you gain access to universal AI. Something that you can use to tackle and develop applications that seem not feasible today, or are outright thought of as not in reach in the near future.

One angle to tackle this is topology. For a game like HoMM you typically process paths leading from one location of the map to the next. It is a bit like the travelling salesman problem to find the best possible paths for each hero. There are existing algorithms to find a best path. But from an AI developer's point of view, these are insufficient because these algorithms are many orders of magnitude too slow. Typically a reasonable fast AI needs to process tens of millions of paths in a second, plus the paths are much longer because these require a higher granularity.

In principle we need a better approach. These are the types of problems I usually tackle while doing my conceptional work. And here, in Heidelberg, I had one of these very bright spells, where you simply have to envision structures, mechanics and a solution that brings all together. What emerged is a method to process very large volumes of paths, with a variable level of abstraction and a high level of parallelism.

Progress like this is quite significant to AI development in general. In my case the progress is even bigger, because it was a missing link in a larger picture. With other words another big step towards a universal AI.

This step is big enough to give me a boost, and help me to believe that I am right here.

You’ve successfully subscribed to Quantomas AI
Welcome back! You’ve successfully signed in.
Great! You’ve successfully signed up.
Your link has expired
Success! Check your email for magic link to sign-in.