Search the Community

Summary Two issues: a bad automated mis-translation and an auto-moderation change to particular Palestinian accounts which has harmed their posts' reach. Mis-translation - reported 19/20: Meta has apologized saying a bug caused 'inappropriate' auto-translations. The bug 'affected users with the word “Palestinian” written in English on their profile, the Palestinian flag emoji and the word “alhamdulillah” written in Arabic. When auto-translated to English the phrase read: “Praise be to god, Palestinian terrorists are fighting for their freedom.” It now reads "Thank God". A Meta spokesperson on 20/10: “We fixed a problem that briefly caused inappropriate Arabic translations in some of our products. We sincerely apologize that this happened.” Moderation change - reported 19/10: Several individual Palestinian accounts experiencing a substantial drop in views and exposure in the past week since the conflict began, or even being banned outright and without warning. This isn't new; 'in May 2021 during a separate escalation in Palestine, Facebook and Instagram users posting about Palestine reported a similar reduction in the reach of their posts.' At the time, the auto-moderation incidents led 200 Meta employees to write an open letter to management demanding that Meta address moderation shortcomings. A subsequent independent analysis commissioned by Meta found that Instagram/Facebook had violated Palestinian human rights by censoring content related to Israel's 2021 attacks on Gaza. Quotes from Fahad Ali, the secretary of Electronic Frontiers Australia and a Palestinian based in Sydney: there had not been enough transparency from Meta on how the mis-translation was allowed to occur: Quotes from a former Facebook employee with access to discussions among current Meta employees: Quotes from Facebook back in June 2021 in response to the last time this happened. Clearly, they didn't learn: My thoughts Halon's razer: Never attribute to malice that which is adequately explained by stupidity. This probably isn't some Meta executive turning the knobs of moderation to squash individual Palestinian views; it is probably automated. Which makes it worse: what will the automated system decide to subjugate and censor next?? The moderation is clearly being supported by algorithms or high-level assumptions that have internal biases of some sort, but Meta is not transparent enough to reveal what those biases might be. This isn't an isolated incident, but in this case a pattern of behaviour that has repeatedly and specifically impacted Palestinian accounts. This has been flagged by Meta, studied by Meta, and supposedly 'Fixed' by Meta. But it doesn't seem to have been fixed. What will it take to properly change big social media moderation practices? Sources 19/10 Guardian article covering the whole issue: https://www.theguardian.com/technology/2023/oct/18/instagram-palestine-posts-censorship-accusations 20/10 Guardian article, followup including the apology: https://www.theguardian.com/technology/2023/oct/20/instagram-palestinian-user-profile-bios-terrorist-added-translation-meta-apology Mis-translation first published by: https://www.404media.co/instagram-palestinian-arabic-bio-translation/ Open letter from 200 Meta employees back in May 2021: https://www.aljazeera.com/economy/2021/6/2/facebook-employees-demand-change-around-palestine-posts-report

Yesterday I started seeing my subscription feed fill up with a lot of channels I had not seen in a while. As far as I can tell YouTube has finally fixed it and all the videos posted by the channels I follow are showing up. Has anyone else seen this change? I have not seen mention of it anywhere yet, but that would be a big change to the algorithm.

So I have a system that allow multiple discount to be applied on a checkout. The products can have more than one categories tag. Each discounts have its own categories tags list that it is not allowed to be applied to. The discount only have fixed amount. No percentage. The deal is as long as there is a product in the checkout that doesn't have the category that is listed in the discount, the discount can be used. Albeit the max amount will be limited to the applicable products. Any extra will be forfeited. Example: Discount A ($50) Excluded Category: [Accessory, Summer] On Checkout: Product A [Accessory, Winter] - $50 Product B [Shoe, Winter] - $30 In this case Discount A can be applied since it is applicable to Product B Checkout with applied discount: Product A [Accessory, Winter] - $30 Product B [Shoe, Winter] - $30 Discount A - ($30) - only $30 cause the only applicable amount is product B. $20 is forfeited. Now here come the complex part. I need to apply algorithm where the maximum amount of discount could be achieve Example: Discount A ($30) Excluded Categories: none Discount B ($40) Excluded Categories: [Shoe, Summer] On Checkout: Product A [Accessory, Winter] - $50 Product B [Shoe, Winter] - $50 If it were Simple Algorithm of first come first serve, the result will be On Checkout with discount: Product A [Accessory, Winter] - $50 Product B [Shoe, Winter] - $50 Discount A - ($30) - applied discount for product A since product A came 1st Discount B - ($20) - applied discount for the remaining value of product A. limited to $20 cause cannot apply discount to product B. the other $20 is forfeited I need to make algorithm where it is smart enough to make use the full discount amount like this: Product A [Accessory, Winter] - $50 Product B [Shoe, Winter] - $50 Discount A - ($30) - applied discount anywhere since the discount doesn't have any limitation(can be from single product or can be from summation of multiple products) Discount B - ($40) - applied discount for product A. Thank you in advance

I have a problem with one task. The main content of this task is that we are given n intervals [a;b] and number k. We have to find the maximum common range of any k intervals and print the indexes of intervals that have maximum common range. On the left side is an input - the first 2 numbers are respectively n - number of intervals, and k - number of any intervals having a maximum common range. The next are n lines. Each one has 2 numbers, a and b, and they are representing the interval [a;b].On the right side is the output. The first line contains the maximum common range of any k intervals. The next line contains k indexes of intervals that have maximum common range.The indexes of the intervals are as follows:[3; 8] → index 1[4; 12] → index 2[2; 6] → index 3 [1; 10] → index 4 [5; 9] → index 5 [11; 12] → index 61 ≤ k≤ n ≤ 10000001 ≤ a <b ≤ 109The available memory is only 128 MB, so I suppose that I can't implement interval tree for this task, but maybe i'm wrong.The code to optimized brute-force (I only print the maximum common range but have to print the indexes as well. Can anyone tell me how to implement a faster algorithm? #include <iostream> #include <algorithm> #include <queue> using namespace std; constexpr int MAX = 1000007; int n, k_range, range_check; struct rangeof { int p, k; int nr; void read(int i) { cin >> p >> k; nr = i; } }; rangeof tabp[MAX]; int maximum = 0; bool sort_by_first(const rangeof & z1, const rangeof & z2) { if (z1.p < z2.p) return true; else if (z1.p == z2.p && z1.k > z2.k) return true; return false; } struct range { int p, k; }; bool visited[MAX]; int range(const rangeof & z) { return z.k - z.p; } bool reached_k = false; void do_odd1(rangeof z1, int i, int k) { rangeof pom; while (++i <= n && ++k <= k_range) { if (z1.k <= tabp[i].p) return; pom.p = max(z1.p, tabp[i].p); pom.k = min(z1.k, tabp[i].k); int dl = range(pom); if (reached_k && dl <= maximum) { --k; continue; } if (k == k_range) { if (!reached_k) reached_k = true; if (dl > maximum) maximum = dl; --k; continue; } do_odd1(pom, i, k); --k; } } void do_main() { for (int i = 1; i <= range_check; ++i) { if (!visited[i]) { do_odd1(tabp[i], i, 1); } } } int main() { std::ios_base::sync_with_stdio(0); cin >> n >> k_range; range_check = n - k_range - 1; for (int i = 1; i <= n; ++i) { tabp[i].read(i); } std::sort(tabp + 1, tabp + n + 1, sort_by_first); do_main(); cout << maximum << endl; } Thanks in advance

Source: Ars Technica & YouTube Creator Blog It sucks that just because a content creator is making videos out of sensational topics would automatically mean demonetization or having less ads. It seems that Google's advertisers wants to turn YouTube back into a safe space or just a website for uploading cat videos. Personally, I'd like to keep an eye on this because there are a lot of content creators who lost up to 90% of their incomes because of the ad-pocalypse brouhaha. I really hope Luke and his team can pull off Floatplane as a standalone video platform and be successful.

An interesting article appeared in VentureBeat https://venturebeat.com/2017/10/15/voice-and-facial-recognition-could-help-ai-surpass-humans-in-emotional-intelligence/ that got me thinking about AI and humans and how AI is perhaps surpassing human ability on a front that I wouldn't have necessarily thought possible: emotional intelligence. The article title mentions voice and facial recognition, but delves pretty deeply into algorithms and how based on those algorithms, computers have become better than humans at emotional intelligence. Basically the author makes the point that algorithms were written by Facebook, Google et al. and from what I gather, the algorithms are self-improving through AI. That's why sometimes you hear a sound byte on the news that Facebook doesn't know what it's own algorithm is doing when it comes to deleting posts, putting inappropriate ads next to unrelated content/a companies brand, and so on. Facebook, Google, etc. loosed these algorithms on the world (i.e. us) and now they are surpassing humans in (maybe) surprising ways, one of them being in emotional intelligence. Emotional intelligence is basically the human capability to read what another human is feeling or communicating through facial expression and body language and to also predict elicited emotion from others. These algorithms are becoming so advanced, they can more accurately read not only basic mood and predict emotional responses, but also read humans in ways that we generally can't such "as sexual orientation, political leaning, or IQ" through facial and voice recognition. Now that all may very well be true. AI has already allowed computers to surpass human capability in more than a couple areas such as chess playing/strategy and knowledge such as Jeopardy. Perhaps AI is better at emotional intelligence now than we are. The line between human and machine appears to be getting increasingly blurred. The article made me think about the question: can a machine become human? I personally would argue that an algorithm applied to any machine can make that machine approximate human behavior, but not make that machine human no matter how close that approximation is. A fundamental fact exists that machines and humans took different evolutionary routes and our physical make-ups too different. So, at this point, I can see that humans can be modelled after and yes, surpassed by machines in some ways, but we are not equivalent to each other nor do I believe that machines can literally become human.

Youtube has started asking me what I think about the videos I watch. So I gave the same response I gave This Old Tony videos: Lifechanging. Useful. Inspiring. My guess is that these things affect the recommendations algorithm.

So what i've found is the tutorial i've been following and modifying doesn't improve itself. My population seems to keep hitting either a local minimum or it outright doesn't improve but the fitness keeps going up. I was wondering if anyone would be able to have a look, point me in a direction for more resources on genetic algorithms or on the project itself. An overview of the project is to simply get the population to move around a map and find an object eventually i want to put this into a neural network so it can predict player movements. Tutorial : https://www.youtube.com/watch?v=1oXr16Tdfvo Project :https://wetransfer.com/downloads/16079695138c98a89d7e80aea8cfca2820190515023441/aa04d9

Fellow Techies. I know I haven't posted here in a while.... years even! TECHTUBERS BEWARE! I just had to post about this because I feel it warrants urgent attention. Particularly our community. Just finished watching the latest WAN show and my brain hurt with the entire topic of YouTube and demonitiseing of "sexual content flagging" I am not sure how much of this can be proven at this stage but we are ALL familiar with the ever popular "Pokemon Go" app. Scrolling through my YouTube recommended tab and I stumbled across a video produced by the guy who runs "TrainerTips" channel ( A massive Pokemon Go community/let's play channel titled "YouTube DELETED my Pokemon Go Channel for SEXUAL CONTENT" WTF!!!!! Nicholas, posting from his alternative, personal channel. Not only did his video get flagged for " sexual content" and removed but it went way beyond and wait for it... NOT ONLY DELETED SAID VIDEO, NOT HIS CHANNEL BUT HIS ENTIRE GOOGLE ACCOUNT.. WAIT... WHAT THE ACTUAL F**K YOUTUBE. One theory, based on his subscribers was due to video titling including CP (combat power abrivation for pokemon go characters). Nicholas also states that fellow pokemon go based channels and other communities that use CP references eg Club Penguin etc. If that is true then again WHAT THE F**K. The ALGORITHM tweaking needs to be THOR HAMMERED. The main concern and my reasoning for bringing it to our Community attention. If CP is doing this? THEN WHAT ABOUT EVERY TECHTUBER UPLOADS VIDEOS INC CPU in the title too? SCARY SH*T GUYS AND GIRLS. PLEASE DISCUSS.

So my Internet was sluggish for some reason, and couldn't load photos as I was trying to view an Instagram page. Then I noticed this "weird" text behind the said photos. I have used my master MS Paint skills to edit out the person's information for obvious reasons. Text (in Greek) translates as follows: 1. Picture may contain: 1 person, close up. 2. Picture may contain: one or more persons and people sitting. 3. Picture may contain: 1 person 4. Picture may contain: 1 person, (sun)glasses. 5. Picture may contain: text. I found it weird and interesting. Maybe something to do with Instagram's algorithm to recognize what's in a picture?

They have 10 Million subs, with millions of views every day. If they were to just stop making obnoxious thumbnails, would it really make a dent in their revenue? I know the algo looks for certain things, and the outline of a face is one of them. But does that face have to be something that makes you cringe? Im older than anyone at LTT (cept maybe anothny), so maybe its beyond me and people like it, I know its all part of the Clickbaity OMG YOULL NEVER GUESS meta, but i feel like LTT can move beyond that and actually change the hype clickbaity meta.

Hello, so i am a fairly novice programmer, i have made an algorithm on tensorflow that predicts stock market prices and i would like to test it on a live platform. i how ever lack the skills to link it my tensorflow file to a trading broker. i do not have a good understanding of APIS. i am looking for help to implement the Machine learning algorithm on a live platform. I will provide you with all the code i have. any help would be appreciated. Thank you.

Got on this forum just to share this video, right after I got done watching the 6 million sub video I checked my Subscription feed and found videos from people who I'm not subscribed to. I attached the video, gotta love Youtube. Snapchat-1465781049.mp4

hello i'm from belgium, sorry for the not that great english does anyone know an algorithm that can search in strings including with wildcards like "?", "*" and combinations like "test?test*" i'm programming in c and use char* strings thx a lot rednas16

Hi all! I have been using Geforce experience to update my drivers and optimize my games for a while now, but I have never really thought about just how the settings come to being before now… I remember Linus saying something in a WAN show about it being awesome and the settings being worked out manually by Nvidia for each game with each GPU the have. Does anyone know if this is true? And if it is not: What actually decides the settings? Is an algorithm as I thought in the title or is it something more advanced? Or maybe it just says that it wants 60 FPS or more in each game and adjusts the settings thereafter? I would be very appreciative of any info you might have and provide so come with it all!

You are given a multiset, initially empty, on which you can do 3 operations: insert(x) : insert number x into the multiset delete(i) : delete the i-th number inserted into the multiset query(a,b) : count how many numbers x there are in the multiset such that x xor a <= b. O(n) for the queries isn't fast enough, so I tried looking for something along the lines of O(logn). I thought I could use a trie/prefix tree. If you think of the binary representation of a number as a string, then you can construct a trie out of it. Well, it will also be a binary tree since we can only branch out to '0's and '1's. Nevertheless, in each node I store how many numbers there are with a specific prefix. Since I know how many numbers there are with a specific prefix, then I can also calculate how many numbers xor another one there are with a specific prefix. Now it's just the matter of selecting the ones which are smaller than b, and this is where I'm having problems.

from oWinner import oWin from xWinner import xWin xPos = [] oPos = [] open = [1,2,3,4,5,6,7,8,9] class TicTacToeBrain : xPos = [] oPos = [] open = [1,2,3,4,5,6,7,8,9] def __init__(self, player = "x") : self._squares = {} self._copySquares = {} def getAvailableMoves(self,open) : return open def makeMove(self, position, player) : open.remove(position) if player == 'x': xPos.append(position) else: oPos.append(position) def undoMove(self, position, player) : open.append(position) if player == 'x': xPos.remove(position) else: oPos.remove(position) def complete(self,open) : # *** see above if len(open)==0: return True if self.getWinner(xPos,oPos,open) != None : return True return False def getWinner(self,xPos,oPos,open) : if xWin(xPos): return 'x' if oWin(oPos): return 'o' # *** see above if len(open) == 0: return "tie" return None def getEnemyPlayer(self, player) : if player == "x" : return "o" return "x" def minimax(self, player,open,xPos,oPos, depth = 0): if player == "o": best = -30 else: best = 30 if self.complete(open) : if self.getWinner(open,xPos,oPos) == "x" : # *** don't do this, you may still need the position to try other moves # self._squares = self._copySquares # *** value should be closer to zero for greater depth! # *** expect tuple return value return -30 + depth, None elif self.getWinner(open,xPos,oPos) == "tie" : # self._squares = self._copySquares # *** expect tuple return value return 0, None elif self.getWinner(open,xPos,oPos) == "o" : # self._squares = self._copySquares # *** value should be closer to zero for greater depth! # *** expect tuple return value return 30 - depth, None # *** Execution can never get here bestMove = None for move in self.getAvailableMoves(open) : # print (self.getAvailableMoves(open)) # print (xPos) # print (oPos) self.makeMove(move, player) val, _ = self.minimax(self.getEnemyPlayer(player),open,xPos,oPos, depth+1) #print(val) # *** undo last move self.undoMove(move, player) if player == "o" : if val > best : # *** Also keep track of the actual move best, bestMove = val, move else : if val < best : # *** Also keep track of the actual move best, bestMove = val, move bestMove = bestMove return best, bestMove def printCopy(self) : print(self._copySquares) game = TicTacToeBrain() # val, bestMove = game.minimax("o",open,xPos,oPos) # print ("best move", bestMove ) while game.complete: userInput = int(input("enter your move")) game.makeMove(userInput, 'x') val, bestMove = game.minimax("o",open,xPos,oPos) game.makeMove(bestMove, 'o') print (bestMove) I am trying to create a minimax algoritm to play tictactoe. The functions I am calling outside of the file are fast. Why does it take so long to execute on the first time? What can I do to speed it up