Tough decision. Will I post the context to some kind of “community breakdown”? Sadly, yes. Let’s have a short bookmark to not one but two remarkable points. Somehow both developed simultanously in the same thread, and couldn’t be more different.
Not exactly special so far, this happens every few weeks or even days. The script looks like this: the MR fanboys join, tell about their good experiences and try to help by putting links, tutorials and advice. Usually the newcomer either isn’t seen again, or even later reports about his progress proudly. Sometimes things go wrong and the mafia around MR tries to convince the poor soul that the problem isn’t exactly the “others”. At least this was the worst case, until today.
Two pages later things are getting stranger. At this point the guy had my attention:
Words fail me. I give up. I have seriously never come across this kind of obtuseness in my entire life.
Let us just stop here. I wouldn’t expect more knowledge gain, but the thread becomes harder to search for useful information and we might lose some of the newcomer’s questions.
He seems to be new to the internet. At this point the topic has around 15 different contributors where every single one supports Minorating and the specific point of discussion (N-grade), both regulars and new community members. In no way this will stop. The opposition now activates some self-reflection mode and … no. But you can read on for yourself, I wanted to get somewhere else:
@Minolin great respect to your calmness.
Calm? No exactly. He was boiling. Calm people don’t change their signature to double-facepalming Stefano.
They also don’t directly accuse people of being incompetent.
Did you ever hear about Dunning-Kruger?
By chance this confrontation was in a different thread so any potential discussion wouldn’t spam the Minorating-thread. Evil, calculating Minolin. To his defense, Parminio was offensive here as well and stated that gamepads have binary (on/off) inputs only until the evidence was overwhelming and he needed to change the strategy (not the stance).
Doesn’t this remind us of somebody? Hint: There’s a picture of him a few lines above. The Dunning-Kruger effect (paper) basically says that if you have little knowledge about a specific field, you also have a hard time to estimate your level of skill accordingly. More precise: If you don’t know how complicated something really is, you are overestimating your abilities relative to the maximum. Unfortunately the scientists turn this around into the short, precise and insulting: “Too incompetent to notice”. This is the core of Stefano’s problems with the community. The only solution he found was: The competent one leaves, the self-certains stay. This makes me worry a little bit. At this point we want to avoid thinking about tomorrow, right?
Let’s zoom back to the imaginary “abort” button. Minolin does another thing (it is more interesting to watch the people, not the clown!): He starts ignoring the intruder and instead gives sensational news. The hope would be to distract the people and give something else/better to talk about. Nice try (but: obviously new to the internet).
This isn’t exactly easy to read, but worth to understand. He tries to create a single computer program that is fed with the accident data from Minorating, and gives an opinion in the question of guilt. The system (or the next one) could improve big times I suppose, but of course there are catches. Neuronal Networks are not hard to program, but very hard to train for a decisive process. From the distance (remember Dunning-Kruger!) it looks impossible to generate many thousand test cases by hand.
The shortcut is: I know which grades are colliding in a given accident. Given that, I could only take A+D accidents and assume, that in most of the cases the D driver is guilty. So if we now give only the accident data to the NN, but train it by the grade assumption: The NN is trained to find a behaviour driven correlation in order to distinct A and D drivers, without knowing the grades. So the result is not really “who’s fault is it” but “who is the A grade”. Maybe the result means something.
Oh. This could be a shortcut indeed. If this is working, the result is directly dependent on the quality of the grades today. Given the difference between A and D grades is high enough, well, this could turn out amazingly. The idea is to let the AI find elements (we humans maybe can’t see) that correlate with a good driver in contrast to a bad one; instead of finding out who’s fault it was. The latter may be a hard task even for humans. The AI would be trained to tell who’s (more) acting like A or D, totally ignoring the rules and accident itself. Yep, that’s how the wizardry of correlations work.
Bonus points: The comment about feedback. I’m not a fan of the current system for its lack of feedback. You can’t just push people away without telling them what they did wrong.
Nice follow up containing the words: Neo, Skynet, Technical Singularity, Aliens, mankind.
One last look to the other news: Ian Bell gave an interview about features of Project Cars 2. It’s basically marketing, but also contains this:
We’re still working on multiplayer features. I have 10 full time dedicated multiplayer coding staff on the project now and we’re working very hard to deliver […] Our main focus has been on improving the ‘acuity’ of the online experience; on dealing properly with griefing (wreckers) […]
At this point features seem to be fixed. Was this the reason why Minolin pulls news about the new paths of Minorating now? Side note: 10 full time coders? Wow. I could imagine Stefano created the Assetto Corsa multiplayer with the effort that those 10 cost in project overhead only.
But: We shouldn’t become too certain about things we don’t exactly know, right?