121–129 of 129 posts
Earlier quoted context omitted.
Or use the word “trigger” to mean mere annoyance. It is tragic that people with a hatred of people living with conditions like PTSD and epilepsy have turned that word into a joke. (I guess most people who use that word now are unaware of this history.) I know people who experience actual triggers and it's a very painful thing.
I am falling behind. Can someone send me the most current, approved vobaulary lists, please?
Earlier quoted context omitted.
> Or use the word “trigger” to mean mere annoyance To be clear, the people that took away meaning from this word were the people that added "trigger warnings" to the most trivial of things. "trigger warning: meat eating." "trigger warning: bad words." > It is tragic that people with a hatred of people living with conditions like PTSD and epilepsy have turned that word into a joke That's really not what's happening.
> To be clear, the people that took away meaning from this word were the people that added "trigger warnings" to the most trivial of things I would object to this on two grounds. Firstly, seemingly insignificant things can be triggers. It doesn't have to be something people without PTSD consider a big deal, because by definition it isn't them who are affected. I have seen this first-hand. The source of the trauma and…
Are you a yoga teacher? Because WOW!! What a stretch!!
Earlier quoted context omitted.
All due respect - I wouldn't really consider this to be an example of anything close to approaching the subject of AI alignment - at least not in terms that people talk about in the past few years. "Exhaustive short horizon exploring strategy with simple terminal heuristics" - is this in reference to the for loop which calculates the future possible states of the game up to a certain depth? It seems to be a state exp…
The AI is optimizing for a different goal than the player : Like a loving helicopter mom, this AI is trying its best to help you reduce tower height now (to reduce risk) whereas the human player has different objective often related to increasing its score over a certain horizon. The thing is from the programming point of view there are not a unique way to define what should be a "best move".
My issue is with you using incorrect terms and conflating machine learning with a relatively simple graph based algorithm with a relatively simple heuristic.
There is no machine learning going on in this codebase whatsoever as far as I can tell, but you seem to be claiming there is. Do you care to respond to that?
Earlier quoted context omitted.
Making a single line is a very bad thing to do in Tetris. You always want to make four lines simultaneously, and there's no easier way to do that than by stacking long pieces.
I'm not trying to argue what is a good or bad thing in Tetris. I'm trying to explain the algorithm. Hatetris tries to maximise the minimum time to removing a line; Lovetris, as explained by its author[1], flips a sign, thus it tries to minimise the minimum time to removing a line. If that contradicts your notion of "best piece" then either the author is being too imprecise or you are being too literal. Either way I d…
There is a worthwhile observation to make here, which is that even though Hatetris assigns scores to all the pieces and correctly determines, based on those scores, which one is the worst, Lovetris cannot correctly determine which piece is the best by picking "the opposite of what Hatetris does". Good is the opposite of bad, but the metrics for badness are not simultaneously metrics for goodness.
Earlier quoted context omitted.
The AI is optimizing for a different goal than the player : Like a loving helicopter mom, this AI is trying its best to help you reduce tower height now (to reduce risk) whereas the human player has different objective often related to increasing its score over a certain horizon. The thing is from the programming point of view there are not a unique way to define what should be a "best move".
Are you responding to someone else? My issue is not with understanding the algorithm's flaws and the difficulty surrounding providing players with an optimal choice considering players have different goals than the heuristics used in the algorithm. My issue is with you using incorrect terms and conflating machine learning with a relatively simple graph based algorithm with a relatively simple heuristic. There is no m…
Afaik I haven't coined the term "machine learning" in this thread and made no such claim.
To clarify : There is no machine learning going on in this codebase (though the problem is interesting enough that you can replace the simple AI by a more advanced one).
I was just merely pointing that AI alignment problem already affect even simple decision making agents.
Earlier quoted context omitted.
I'm not trying to argue what is a good or bad thing in Tetris. I'm trying to explain the algorithm. Hatetris tries to maximise the minimum time to removing a line; Lovetris, as explained by its author[1], flips a sign, thus it tries to minimise the minimum time to removing a line. If that contradicts your notion of "best piece" then either the author is being too imprecise or you are being too literal. Either way I d…
> Either way I don't think it's an interesting discussion. There is a worthwhile observation to make here, which is that even though Hatetris assigns scores to all the pieces and correctly determines, based on those scores, which one is the worst, Lovetris cannot correctly determine which piece is the best by picking "the opposite of what Hatetris does". Good is the opposite of bad, but the metrics for badness are no…
Earlier quoted context omitted.
> Either way I don't think it's an interesting discussion. There is a worthwhile observation to make here, which is that even though Hatetris assigns scores to all the pieces and correctly determines, based on those scores, which one is the worst, Lovetris cannot correctly determine which piece is the best by picking "the opposite of what Hatetris does". Good is the opposite of bad, but the metrics for badness are no…
Hmm, it wasn't my understanding that Hatetris did that. I thought it tried to maximise time to completing a single line, but I may be misremembering or may have misunderstood. Moreover, if it did do what you claim it does then I believe that reversing a sign would end up always giving you the long piece.
Earlier quoted context omitted.
> To be clear, the people that took away meaning from this word were the people that added "trigger warnings" to the most trivial of things I would object to this on two grounds. Firstly, seemingly insignificant things can be triggers. It doesn't have to be something people without PTSD consider a big deal, because by definition it isn't them who are affected. I have seen this first-hand. The source of the trauma and…
> Nice use of “bad words” to describe racial and other slurs by the way. That isn't trivial for the people affected Are you a yoga teacher? Because WOW!! What a stretch!!