About machine learning

It has recently, in so many places, been suggested that "AI" poses the single largest threat to humanity. To me, this claim is just uninformed and rediculus. Some arguments against:

1. It has been stated that we will not be able to predict the evolution of AI. And it will grow in intellegence exponentially and BAM out-smart and kill us all. But looking at history this kind of prediction has been made lots of times befor with different technical advancememts. To error. I see no reason to beleive this piece of tech will be any more disruptive in that sense.

2. AI is, also amongst highly educated people, vastly miss-understood. In its current incarnation, it is a set of tools to solve detailed technical problems. No where is there any will or intention. AI people still say (and so do I)they do not even understand will.

3. They keep saying we will have an escalation of ai evolving it self away from us slow biologically evolved beings. And finally erasing us. But if an ai is intelligent enough, and the above is a problem, why would it want to have it self superseded by a smarter ai? Won't that feel like for us and the level 1 ai?

Rehammar's law

I just read about Godwin's law. I think I can make a generalisation: If the number of entries in a discussion thread on the internet (or some where else) goes to infinity, every statement that can be made, will be made. This is just due to that people tend to make all sorts of weird connections and sooner or later, every statement will be made. Woilla - Rehammar's law.

Why Windows is such a success

I have devoted a full blog post here to all things I hate about Windows, and it is growing. However, it is impossible to deny the huge success the operating system has had on the desktop. And I think I just realized how this came about. It is really one key thing they realized:
Distribute self-contained binaries and keep binary compatibility.
I think this is the genius move that *NIX did not understand, and still does not.

Open mobile

When will we see an android version for a phone running on openrisc or risc v and some free gpu? This would be a really nice thing that could potentially in the long run increase availability and lower costs for end consumers of mobile phones.
The last piece in this puzzle would then be a free lte stack and a free asic to execute it.
When will I be able to buy that?

Car fleet virtualization

With the current trends in automotive with electrification and automation, I think it is time to start talking about car fleet virtualization.
In the not-so-distant future, cars will be mostly software. Software will regulate the driving style and the functionality of the vehicles. Do I want acceleration or comfort? A car with a bed or a desk? The virtualized car pool will provide whatever I desire.

HFSS errors and their solutions

Lately I have been constructing som passive microwave components using Ansoft HFSS. It is a very powerful piece of software, but somethings the error messages can be very hard to interpret. So this post will be continuously updated with problems I have had and solved.

1. This error I believe is related to meshing issues. However, I could only reproduce it in a statistics analysis where it reoccurred until I changed the mean in the distributions: [error] Critical features could not be preserved correctly. (10:14:21 AM  Aug 22, 2016)    [error] Statistical Analysis failed - Solver error. More information may be available in...
2. 
   

Are we living in a computer simulation

This is a hot topic at the moment with many people having thoughts on the matter. My answer to the question is: of course we are not. The answer is depending on what is meant by a computer simulation, and here I am referring to a situation similar to the Matrix, which is what is usually considered in the current discussion. However, I think the discussion is not very different from what actually Descartes already had when he was discussing his demon and Kant with das Ding an sich.
Anyway, my arguments against the idea that we are living in a computer simulation are:
1. Evolutionary extrapolation. It is often claimed that extrapolating the current pace of technology advance indicate, on an evolutionary time scale, that there should be machines capable to performing these kind of simulations at some point. However, this is a logical error, because current pace of technological advance has not occurred on an evolutionary time scale. Thus, it can not be extrapolated on an evolutionary time scale either. There is no predictable way to tell what will happen with technology over long time scales.
2. Ethics. We know from our own civilization that once we get advanced, we also have ethical ruses. E.g. we do not allow for any type on animal experiments. I am guessing that the kind of simulation referred to in the discussion would be considered highly unethical, and not be allowed to be run.
3. Computing power. If the world is constituted in a way similar to ours. That is, the presumed simulated structure is similar to the world in which the simulation is run, then it seem to follow that it will be very hard to simulate it accurately. It is not easy to simulate it self.
4. Epistemologically. If there is no way to detect that we are in a computer simulation (actually Neo noticed in the Matrix), then the discussion is not a very interesting one at all. Since then we have no way to test the hypothesis. Further, is it not likely that if we are in such a simulation, that just by design, we would not be able to detect the boundaries of the simulation. Any "flaws" in the simulation would appear as how the world is constituted to us (maybe this is the reason behind cancer, chaos theory, the Republican party, ... and other crazy-ness).