Imagine what it would look like if ChatGPT were a lossless algorithm. If that were the case, it would always answer questions by providing a verbatim quote from a relevant Web page. We would probably regard the software as only a slight improvement over a conventional search engine, and be less impressed by it. The fact that ChatGPT rephrases material from the Web instead of quoting it word for word makes it seem like a student expressing ideas in her own words, rather than simply regurgitating what she’s read; it creates the illusion that ChatGPT understands the material. In human students, rote memorization isn’t an indicator of genuine learning, so ChatGPT’s inability to produce exact quotes from Web pages is precisely what makes us think that it has learned something. When we’re dealing with sequences of words, lossy compression looks smarter than lossless compression.

Ted Chiang’s essay about ChatGPT is required reading

Tumblr is not getting rid of the chronological Dashboard!

Folks keep sending this feedback in, I've replied to 108 in the last hour, so I figured I'd write a post. Please reblog this, thank you!

Rest assured, [tumblr] is not getting rid of the chronological feed of blogs you follow. For details, please see this update from Staff over a week ago.

We also mentioned this on Changes the day after. Please do follow @changes, it only posts twice a week and is the best way to stay informed. (BTW, a lot of you are blocking Changes, WTF is up with that?)   As for algorithms, we actually have several already in use, and those are the ones we're working to improve. You can see them under the For You tab, and turn them on or off for your Dashboard via the "Include posts liked by the blogs you follow," "Best stuff first," and "Based on your likes" toggles in your Dashboard settings.   With those three toggled off, which will always be an option, your Dashboard feed will be solely posts from the blogs you follow sorted chronologically from newest to oldest.   If you have any other feedback, please use the Feedback category in the support form. Submitting feedback there is like a firehose directly to our developers and designers. (Please of course be mindful of signal to noise. I know a firehose is tempting to abuse, but the more nonconstructive feedback we get, the harder it is to see the constructive feedback.)

And, of course, please stop sending concerns about removing the chronological feed. We are not removing the chronological feed of blogs you follow, we never will, and we never planned to.

Thanks!

What if the reason antis get confused about ‘gross shit’ being in the main tags is because they assume the main tags should be more algorithm-like?   That the ‘gross shit’ should sink to the bottom where it is never seen.   Bad news guys:  Most recent absolutely does not fucking work like that, and nobody has a clue how most popular works.

No streaming platform can accurately predict taste; humans are too dynamic to be predicted consistently. Instead, Spotify builds models of users and makes predictions by recommending music that matches the models. Stuck in these feedback loops, musical styles start to converge as songs are recommended according to a pre-determined vocabulary of Echo Nest descriptors. Eventually, listeners may start to resemble the models streaming platforms have created. Over time, some may grow intolerant of anything other than an echo.  Before there were Echo Nest parameters, the 20th century music industry relied on other kinds of data to try to make hits. So-called “merchants of cool” hit the streets to hunt for the next big trend, conducting studies on teenage desire that generated tons of data, which was then consulted to market the next hit sensation. This kind of data collection is now built into the apparatus for listening itself. Once a user has listened to enough music through Spotify to establish a taste profile (which can be reduced to data like songs themselves, in terms of the same variables), the recommendation systems simply get to work. The more you use Spotify, the more Spotify can affirm or try to predict your interests. (Are you ready for some more acousticness?)  Breaking down both the products and consumers of culture into data has not only revealed an apparent underlying formula for virality; it has also contributed to new kinds of formulaic content and a canalizing of taste in the age of streaming. Reduced to component parts, culture can now be recombined and optimized to drive user engagement. This allows platforms to squeeze more value out of backlogs of content and shuffle pre-existing data points into series of new correlations, driving the creation of new content on terms that the platforms are best equipped to handle and profit from. (Listeners will get the most out of music optimized for Spotify on Spotify.) But although such reconfigured cultural artifacts might appear new, they are made from a depleted pantry of the same old ingredients. This threatens to starve culture of the resources to generate new ideas, new possibilities.

Amazon Warehouse Robot Worker Collapses from Exhaustion 🤖

TLDR: Please, reblog. It not only helps artists and creatives, it helps ensure the continuation of the user-controlled, non-algorithmic dashboard feed feature.

So, there is a recent post going around about how the poll feature really highlights the view/engagement/reblog disparity. And the OP makes a great point that on OTHER platforms, even 1/4 engagement per view would be extraordinarily high.

What that leaves out, though, and something I don’t see talked about as much in the reblog/like discussions, is that those other platforms rely completely on algorithms. Our beloved, only-show-what-others-reblog dashboard (arguably the best feature of this site) is reliant on reblogs to remain a viable feature. If people only like/vote on posts, “my dash is dead” phenomenon will increase and eventually result in no more Tumblr, only-show-what-others-reblog dash (you already see this in the increasing amount of “suggested”/algorithmic options on mobile).

So please, please, if you care about this last bastion of sequential feeds (and, of course(!) want to support artists, REBLOG POSTS.

Invictus weaponized for Summer Royal Cosplay

2024 Netflix content from WALES & LONDON UK. Not from Canada where the 2025 games need the promotion. Also, what happened to Santa Barbara polo & the pwives?

May 4th- little being little birthday

Sunday, May 5th- start of Met Gala festivities NYC

Monday, May 6th-NYC Met Gala

Wednesday, May 8th clapback service 5pm St Paul's UK The Meghans are both on program as they create "royal" adjacent content for Netflix

June 7th Grovensor wedding UNINVITED

June 15th trooping the colour UNINVITED

June 21st USA Warrior Games

July 11th The Meghans cosplay royals in WALES for golf & formal black tie gala under the Stars w/the stars promo & fundraiser for 2025. Doesn't Canada have golf courses & banquet facilities?🤔

Could it be WALES was chosen because The Meghans are desperate to be associated with "The Prince and Princess of Wales" with every click? They intend to influence the algorithms?

Procedural Dungeon Generation

Alright~ After much work we have finally completed our dungeon generation algorithm. First, we generate the rooms, separate them using a separation steering algorithm.

Second, we generate a graph that represents connections between the various rooms

Third, we perform what is known as Delaunay Triangulation, in which we evaluate pairs of triangles and swap their shared edge if it fails a test boiled down to a simple matrix determinant evaluation Fourth, we construct a Graph of connectivity from the Delaunay triangulation

Fifth, we perform an A* pathfinding algorithm to connect points on the delauney triangulation through the neighborhood graph

Finally, if we didn't traverse through a room, we can remove it to simplify the dungeon into something a little less dense

There are some extra things we can do here and there, and bugs and edge cases to iron out here and there, but for now, we have a pretty neat little dungeon generator. The next step is to actually procedurally generate the rooms and then use the connectivity graphs to implement game logic and whatnot, but that we will save for another time.

Happy Valentines Day!

Here are some programming valentines day cards!

Couldn't find any good ones online, so made my own heh, enjoy and send it to someone who likes corny jokes and/or programming!!

No, you are not alone. I also start things, give up, procrastinate, live the future and not the present and complain about my life not moving forward while I spend time looking at other people's lives.

Hello everyone, how are you?

I hope well.

I was thinking of giving up this tumblr and just as I started everything from scratch to study programming (github, linkedin, notion and I even created another google account) start a new tumblr.

I even commented briefly a few weeks ago with @lostlibrariangirl who supported me to just continue on this one and see my evolution. And today I was thinking that this could be cool not only for me, but for other dropouts, who are frustrated that they think they'll never get to where they want to go and et cetera.

If you've been following me since the beginning of this year, you've already seen that I never completed any course and I assume that I gave up on the most difficult parts, besides the constant anxiety of getting a job in technology and thinking that I would never get it and that made me procrastinate, suffer and not live the present studying.

I gave up on one of the scholarships I got and now I'm catching up on the delay (it's until the 7/30th)

Nowadays my mental health has improved a lot, thanks to my elders.

And also thanks to them I understood that I was going around in circles, being lazy, not trying 100%, being stubborn in a stupid way (for me stubbornness is a quality, if you know how to use it) and spending too much time on other people's lives and not mine.

Also, today it's easier for me not to feel so much anxiety because I no longer have the goal of getting a job registered as a dev this year, so I'm starting to learn from scratch EVERYTHING AGAIN only this time better and really trying hard.

In my case I wasn't doing my best, trying hard and that's why I fell so many times. I don't like the word failure/failure, as I learned from an older, mistake it's study/learning and not failure.

Now in May, for example, I started studying Linux from scratch (my system that I'm confused to understand), git and github. I was all this time without really understanding these 3 tools, just doing it without understanding / in a lazy way and that didn't give me confidence.

And I don't think we need to memorize it, but it's nice to understand and practice.

This print from above is me studying after writing and repeating to myself that I WILL ONLY START a new cycle after ending the old one.

And I hope to share this journey here.

I have a lot to study, whether in back-end or front-end, but I'm not in a hurry and that's why I'm progressing.

So if you're reading this far and you're going through the same thing I was, I wanted to give you some unsolicited advice.

Spend your energy, do physical exercises. Because anxiety can only be resolved with a psychiatrist, but exercises help.

Understand why you procastine, what is making you feel frustrated or afraid? Did you find out? How can you solve this?

If you don't finish something to start another and it HURTS you (if it doesn't hurt, that's fine), try to understand why you give up? Where is making you insecure / afraid? And after you understand this, strive to finish everything you start (I know it's hard, but we need this)

Get off social media for a bit. For me tumblr is what I spend the least time on, but it's very easy to lose hours on instagram and tiktok.

And guess what, that time we spend doesn't come back and unless you work with these networks that ALSO won't give you any money / jobs.

So, if possible, start to regulate your period in these environments. At first it's difficult because your brain is addicted, but after a while it works out and your version of the future (if you invest that time studying and working) will thank you.

I wish you all to be well and not sabotage yourself to achieve your goals.

Platforms likely conflate positioning with popularity, failing to see the self-fulfilling prophecy of an algorithm making stuff popular because said stuff is built to please the algorithm creating more demand for content to please the algorithm. "Viral" content is no longer a result of lots of people deciding that they find something interesting.

Are We Watching The Internet Die?

One of my favorite things about being part of a system is watching every social media and app try to track our preferences and attempt to use algorithms. What do you mean you don't like this song? You liked it two hours ago? Would you like manly shirts mixed with ads for bras? I love confusing computers.

something that all these algorithms don't seem to understand (and maybe it's just the autism here) is that part of loving or caring about someone is listening and being interested in their shit even when it's not immediately fascinating to you

like, oh, my mutual just reblogged 200 posts in a row about something I have no context for. How incredible! I'm so happy to see their joy and excitement for something I will maybe never understand!

I'm sorry but literally like maybe being force fed only content that is directly relevant to me isn't actually good for my soul? Get you some fucking joie de vivre. Some fucking second hand love. Watch and revel in the joy of others that you can never fully understand

Algorithms are fucking Lonely and stale. I don't want just me. I want to share with others and be shared with

FUCK

Yes, “the algorithm rejected your project!” is now a reality thanks to Ya Boy Zaslav.

i feel like a lot of people would benefit from what i did, being just deleting all apps that use algorithms and only using privacy frontends. like you can't use corporate apps and expect not to get subjected to the corporate ground spikes that are attempts at subliminal influence and surveillance marketing. you kinda just gotta make the call to get rid of it or get with it. and i know which one im picking

like i just stopped using any official youtube client because the algorithmic sludge is so insufferable. i just use newpipe and invidious because they just do what i fucking ask. search something? heres results. look at the feed? heres a chronological list of the latest uploads by every channel youre subscribed to. also have an extremely useful and simple download button and all sorts of other shit in the simplest easiest to use package ever. highly recommend stopping using platforms that algorithmically suggest AND deliver content. i would recommend either not using the former at all or keeping them strictly seperate.

Strange way of drawing the Dragon Curve

Alright, so real quick I just want to share potentially the most arcane method of drawing the Dragon Curve I've ever seen, derived and designed by yours truly! As far as I know, this is a novel solution. I know the sequence it generates is known, but I'm not sure if anyone else has used this method before. Its quite elegant if I may say so myself.

So for those that aren't aware in programming the "<<" and ">>" operators are sometimes known as "bit shifts." Basically what this is doing is starting at some number, adding a power of two, then getting a specific 1 or 0 in the binary representation of that number iteratively, until its searched enough bits to know they aren't going to change anymore.

It has to do with this sequence right here. I've mentioned before, my personal favorite way of generating the dragon curve is to start with the sequence 0, reverse it, add one, roll over once you reach 4, and tack that on to the original sequence. So 0 0 1 0 1 2 1 0 1 2 1 2 3 2 1 0 1 2 1 2 3 2 1 2 3 0 3 2 3 2 1 Well what ends up happening is each time you add one, its like adding one to the reversed part of the newly added sequence. So we can track where all these 1's come from based on when they're added. For example, the 1 we added in the "01" step turns into 0 1 0 1 1 0 0 1 1 0 0 1 1 0 Note from here on out its palindromic, so reversing it no longer has any effect. What we end up with is a repeating pattern of two 1's, then two 0's, starting with half that many 0's. When going from 0 1 0 1 2 1 We're adding 1's to the entire second half, so in this step the 1's propagate to 0 0 1 1 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 And again, this is now palindromic. Four 1's, four 0's led by half that many. One of the things I've learned about the dragon curve is just how intrinsically linked it is to binary (and this makes sense when you think of the folding paper method of generating it. Here's an excel spreadsheet demonstrating this in action

Now here's the fun part. My research was to parallelize this algorithm. One approach is to say "Okay, how can we calculate each term in this sequence without looking at the previous ones." And the answer is to exploit these very predictable patterns. And how do we predict these patterns? Simple, we simply count in binary

The right most column is useless, but starting at the next one to the left, we see a familiar pattern, almost. Say we want to know what the 5th number in the dragon curve sequence is (0 indexed). To make the sequence only lead with one 0 instead of two, we need to offset by 1, then all we have to do is increase the number by (n + 1) = 5 and take its 2nd least significant bit (1 indexed because english). The 2nd bit of ( 1 + 5 = 6 ) is a 1. For the next iteration we're looking at the 3rd least significant bit. Here we need to offset by 2, and then we increase the number by 5 again and the 3rd significant bit is the one we take. The 3rd bit of (2 + 5 = 7) is another 1 After that we're looking at the 4th least significant bit. We need to offset by 4, then increase the number by 5, and the 4th significant bit will give us our number. The 4th bit of (4 + 5 = 9) is going to be another 1, bringing our total to 3. Here's a visual representation

This is where the "1 << i" comes in, because that's the same as saying 2^i, which is how we get those offsets of 1, 2, 4, then 8, 16, 32... the "n" in "n + (1 << i)" comes from us offsetting to get the nth term in each sequence Finally the " >> (i + 1) " and "% 2" are to fetch the (i + 1)th bit from the number. After that the increasing size of the leading zero's outpaces our constant offset of the number 5, so we are only going to get 0's from here on out, and we can actually stop, hence the usage of bit length to terminate the loop early.

And if we look at the the 5th element of the sequence (0 indexed) 0 1 2 1 2 3 Funnily enough, in python this brings an actual speed increase (or at least, distributes the cost over the drawing) because of how slow reading and writing to memory is, compared to how math and bit-wise operations are implemented in low level C behind the scenes. Additionally, since there is no reliance on previous work this task can be multi-threaded, or even GPU accelerated if need be. Finally, if you've made it this far, here are a few images of some close ups of dragon curves from my GPU implemented (unrelated to this one entirely) just so that there's something pretty. Enjoy <3