#estimate re-evaluation

[jamaicahomescom]

The Good, the Bad, and the Ugly of Project Management for Beginners

The lack of project management training or experience of many people can be an enormous stress factor for them. Whilst natural organizational ability is enormously helpful, in itself it is no guarantee of any project being both successful and low-stress. It’s Monday afternoon at the office. The week has only begun, but you’re already swimming in a sea of memos, spreadsheets, and schedules. Just…

[paarksunghoon]

you plus me (teaser)

SUMMARY: it’s been six years since heeseung stopped being your friend and the thought of him tagging along an annual camping tradition makes you feel like the world must be crashing round you. one misunderstanding and one trip later makes heeseung re-evaluate all he knows, and it makes you believe that there might life after love.

PAIRING: heeseung x fem!reader (featuring enhypen)

WORD COUNT: no estimate because who really knows but this baby sits at 28K right now. the teaser stands at 2.7K.

NOTES: usually I don’t post teasers but I’m so proud of this story so why not!!!!!! I don’t think I’m going to open a taglist but that could change. I’ll let you know if I do. :) hoping to publish by October 26! thanks for reading!! xx

GENRE: angst + fluff + smut

edit: it’s out!

***

“Please don’t make me go.”

“Y/N, you already said yes. We’re only gonna be gone for a week.”

“I don’t think this is a good idea, Jungwon. You just said that Heeseung is gonna be there.” 

Your best friend sighs and sits down on your bed, inspecting the duffle bag you have that’s half-packed. Your clothes are haphazardly strewn all over your bedding while you plead with him to no avail. You’re so desperate that you consider getting on your knees to beg.

“I’m sorry for telling you now but he was able to get people to cover his shift last minute and paid for a spot on the kayaking rental.” 

“If he’s going, I’d rather save us all the trouble and stay at home.” Jungwon watches you cross your arms over your chest. “Every time we’re in the same room, it’s just a matter of time before things become awkward.” 

“We’ll be outside in the suuuun,” Jungwon says, tilting his head to the side and giving you those amused eyes that he always gives you when he’s trying to convince you to do something with him. You scoff and look away. It almost works. 

“I bet that it’ll be worse since we have a few things planned with the guys already.”

“So what? You two don’t get along. Big deal. We’ve already made reservations to secure a spot on the campsite and set a deposit for kayak rentals.”

“Won, I think you and I view Heeseung very differently. He doesn’t just not like me. He hates me.” 

“Hate is a wrong word.” 

You huff. “I don’t think you grasp just how weird it is every time we’re together. You could cut the tension with a knife.”

“Seriously, Y/N. It’s one week. I’m sure you can survive that. You’ve never missed a camping trip and it’s the first time all of our friends are coming.” Jungwon deadpans and throws a shirt towards your chest, which you hastily grab after being startled by his sudden movement. You know better than to argue with him when he gets like this. “Just help me pack your clothes, dude. Jay’s gonna be here to pick us up tomorrow morning and you don’t want to be under-packed.” 

You relent and grumble. “Are you still staying over?”

He nods. “My apartment’s in the opposite of where we’re going and I didn’t want to make him drive an extra twenty minutes since he needs to pick Riki up. Just need to drop Maeumi off at my mom’s before coming back here. ” Your eyes fall for a flat second before you squash that feeling down.

“I didn’t invite you over, you know.” 

“No, but don’t pretend like you’re not excited,” Jungwon says with a laugh as he pulls your clothes out of the bag and starts to readjust the clothing you’ve folded poorly. Seeing your best friend smile tugs a bit at your heartstrings and you can’t say that you aren’t happy to have him with you. “We should get you packed now so you don’t stress out later.” 

Begrudgingly, you allow Jungwon to sort out your clothes for you and pull last minute items you’ve yet to pack. It annoys you, watching him be so calm when you’re simmering with worry. But you know he’s right—you’ve invested some money into this getaway and it’ll be the last big outing before you move away from Korea for a year-long job opportunity in Okayama before pursuing your Master’s degree. Jungwon knows you a little too well and sometimes it irks you. 

The end-of-summer camping trip is always one for the books. For as long as you can remember, the two of you have been going camping just before everyone goes back to school to celebrate the beginning of a new academic year with your families. But this time, the trip wasn’t just about continuing an annual tradition. It was also to commemorate a new chapter in your life. 

You’re a year older than Jungwon. He’s known you since you were obsessed with learning how to double dutch and you’ve known him since he first learned how to ride a bike. The two of you started out as neighbors when you moved into the house next to his and his family had adopted your own like old friends, eventually inviting you and your parents into their annual camping tradition. Even when dynamics changed and people had left, the tradition was the only thing that remained a constant for you.

This is the first summer that your loved ones announced they wouldn’t be coming along. They all thought it was time for you to embark on new traditions with new people and nobody seemed to mind the change that much except for you. Jungwon had been ecstatic about it since he invited his friend, Jake, to the camping trip last year. You’d been wary at first since Jake is friends with Heeseung, but he never brought up your confusing arch-nemesis and chose to have a great trip before you all started university again.  

Sure, you had a lot of fun. You might even consider last year’s trip as one for the books. But your mom pulling out of the camping trip and everyone around you agreeing that it was for the best made you feel like your world was crumbling around you.

When you graduated university three months ago (Jungwon swears he didn’t cry but you know better than to believe him) and the weight of leaving your home started to sink in. In the blink of an eye, Jungwon wouldn’t be a twenty minute drive and hanging out with all of your friends wouldn’t be as easy as it once was. You’d be in Japan all alone.

This past summer has been a whirlwind as you tried to do everything under the sun, savoring each moment until you wouldn’t be able to anymore. Jungwon’s been a good sport about it, never once complaining when you drag him to your latest adventure. He deals with your sudden shift in mood from happy to sad, letting you cry on his shoulder and braving the cliche words you say when telling him you’ll miss him a lot. 

Unlike past seasons, this is the first summer you haven’t seen Heeseung very often. Lee Heeseung, who usually keeps his head down and minds his business, always seems to have a bone to pick whenever his eyes settle on you. It confuses you to no end and he keeps his quips to a minimum when your mutual friends are around, but it doesn’t stop you from wondering what you must’ve done to make him act like that towards you. It’s a shame because that small childhood crush you always had on him was squashed the first time he ignored your presence 

None of your friends comment on it much. They’re used to the dynamic between the both of you because it's been years of this. Elementary school saw the two of you become friends for the first time and middle school brought more friends into the group. It was in high school that things changed and Heeseung started ignoring you out of nowhere until one Thursday afternoon when he’d told you to leave him alone after pestering him about his change in behavior. 

The odd tension followed you into university and continued to seep into your life. You don’t think you’ve ever been in a room with Heeseung where he’s been anything but nonchalant towards you, often acting like you aren’t there to begin with. You do your best to put up with it and plaster a smile on your face but six years have gone by and you don’t think you can handle a seventh. All of your friends seemed to have moved past it. You don’t know why you can’t.

“Don’t think about Heeseung,” Jungwon says with a sigh. “In fact, don’t think at all. Let me handle everything and enjoy this trip before you move to Okayama, okay?”

“Okay, fine. But I want to see Maeumi.”

Jungwon snorts. “She’s gonna be real pissed when she doesn’t see you for a year, you know.”

“Don’t remind me.”

Jungwon knows you like the back of your hand and has seen what you bring on these trips enough to know what you like to have in your duffle. He packs things you neglected to pull out because your mind has been elsewhere. As much as he wants to flick your head and tell you to quit overthinking so you can help him, he did tell you to let him handle everything. 

Your best friend makes you triple check that the two of you didn’t miss anything before heading back to his apartment to fetch Maeumi. She jumps into your arms when you squat to pick her up and won’t allow Jungwon to pet her white fur body while she’s nestled against you. This fondness and the familiar jab of Jungwon’s elbow to your ribcage makes your heart ache despite the sweet moment. You’re really going to miss home. 

Ever the concerned mothers your mom and Jungwon’s are, they send you with a tray full of sweets for the road. They make you tell them exactly when you’ll be picked up and by who (“Jongseong, Eomma,” Jungwon says for the umpteenth time) and when you plan to come back. His dad gives you a spare bucket hat for when you’re on the water and an old sweater from his college days when Jungwon complains about how you never pack enough layers. The gesture feels warm since you consider his father to be somewhat of your own.

Leaving them to go back to your house feels a bit bittersweet. A lot of your belongings sit in storage boxes in the garage from when you moved out of your campus apartment upon graduating. Jungwon decided to get an apartment for himself with the money he saved from his part-time job as a busboy at a local chain restaurant. Staying over with you makes it seem silly when you remember he used to live next door. 

It’s nine in the evening when the two of you get ready for bed. Jungwon puts your bags by the front door so neither of you would forget while you finish brushing your teeth. He grabs extra blankets from the linen closet and settles onto your L-shaped couch, pulling the fabric just underneath his chin. Your heart feels like it’s sinking in on itself when you think about how this might be the last time you’re able to be so casual around him. 

“Stop overthinking,” he says in the quiet of the night as if he can hear the thoughts in your head. The living room lights are off and the moonlight is what’s responsible for illuminating the space. 

You refrain from throwing your pillow at him. “I’m not overthinking. You’re overthinking.” 

Jungwon snorts. “We both know that’s not true. I know you’re scared about Okayama and I know that’s why you’ve been on edge about Heeseung. You’re usually never this loud about it.” Like always, your best friend is right. 

“It’s hard not to.” Your meek voice makes Jungwon’s heart lurch. “Everything’s changed so fast. I feel like I didn’t get enough time to properly say goodbye to everyone.”

“You’ll be in Japan, not America. It’s not like we’ll never see you.” 

“Yeah, but I won’t be able to annoy you for boba and you won’t be coming over to have dinner with my mom and I.” Jungwon frowns. Too caught up in making sure you were happy this summer, he hadn’t given it that much thought. “I know I won’t be far but I’m scared that things will change too much.” 

For the first time today, Jungwon doesn’t know what to say to make you feel better. “I’ll miss you a lot.” 

“I know that, dummy. I guess…I feel like I’ve been dealing with a lifetime of shittiness and the universe wanted to throw another curveball at me.” Jungwon’s heart softens at your confession. He’s used to your quick jabs and sarcastic humor. Knowing you’ve more afraid than excited makes him upset. 

“The universe sucks,” he says, happy that it pulled a laugh out of you. “I’ll always be a phone call away and you’ll never have to worry about me ignoring you because we both know I’m gonna blow up your texts anyway.” 

“I can always count on you to annoy the hell out of me.” You can’t see his face, but no you already assume Jungwon’s sporting a shit-eating grin. Even if you both know the main reason why you’re afraid of living in Okayama, neither of you say it. You’re grateful that Jungwon doesn’t bring it up. “Still, though. You know how I am with change. I’m really scared that I’m going to hate it there and not have you to keep me company.”

“Life is crazy and unpredictable but that doesn’t mean you’re going to be miserable. I mean, you did a pretty good job of making sure both of us had happy childhoods even though I know you were hurting when we were younger.” 

“It’s really hard not to have expectations or think badly about the future when I feel like I took everything for granted.” 

“I know, Bug,” Jungwon says, using a nickname from your childhood he reserves for when he thinks you need an extra bit of comfort. “But you’re the best person I know. You didn’t do anything wrong. Life just…gets in the way.” 

“Yeah, I know.”

Jungwon is quiet for a moment. “Just please promise me you’ll try to have fun, okay?”

“I know I’ll have fun, Wonnie. I’m scared that I’ll have too much fun and be a sobbing wreck when we get back.” 

The two of you share a laugh. “Alright, fair. Promise me you won’t let Heeseung get under your skin.”

You groan. “If he doesn’t like me, that’s fine. I don’t need everyone to like me. But why go out of his way to act like I’m scum of the Earth?”

“Just ignore him, okay?” Jungwon pleads. “I know it’s uncomfortable but he paid for a last minute spot. I’ll tell him to be mature about it too.” 

And, well, part of you believes Heeseung will listen to Jungwon. Despite being on the younger side in your shared friend group, everyone seemed to listen to your best friend most of the time. Jungwon has an authoritative aspect to himself when he’s refrained from being the silly, happy-go-lucky guy you all know him to be. 

It’s quiet for a brief moment with the wind gently tapping on the windows behind you. “I don’t know why he doesn’t like me.” 

Truthfully, neither does Jungwon. “I’m sorry he’s putting you in a tough spot.” 

“Won, sometimes I really wonder if he hates my guts. He doesn’t talk to me and he never replies to my messages in the group chat. It’s like I don’t exist to him.”

“I think that might be a little extreme.” 

“It’s not and you know it.” 

Jungwon hums. “Well, at least you’ll get away from him when you move to Okayama.” Just like that, all of your worries come flooding right back.

“Yeah,” you say meekly. “I’ll have Okayama.”

You don’t see him, but you know Jungwon’s smiling since you agreed with him for the first time tonight. “That’s more like it. You have your whole future ahead of yourself, dude. Heeseung is just a blimp. In three weeks, he won’t matter because you’ll be having fun in Japan. Just think about that.” 

You try not to think about the fears and hesitations you have about starting anew. This time, you wouldn’t be going back to university after the camping trip. You’ll have a week and a half back home before you’re boarding your flight and saying goodbye to the place you’ve called home for the past two decades. Thinking about the future keeps you up until you hear Jungwon’s snores from the other side of the couch. 

Unsure of when your mom will be coming home, you snuggle further into the cushions and curl yourself into a ball before falling asleep. 

***

comments and reblogs are appreciated! xx

[wip]

For a long time now, it's been impossible to see comments or reblogs with comment/tags on posts over a certain age when using the mobile app or blog view. Today I was looking at a post from 2015 that I knew had at least one reblog comment and lots of tags, but all the reblogs were under "other". I found the comment (but couldn't see any tags) by going to the [blog name].tumblr.com/post/[###] link and scrolling through all the notes in one list, but it's impossible anywhere else.

I know this probably has to do with the many changes Tumblr has gone through in that time, but it's still really inconvenient to have disappearing notes on the platform where part of the charm is that posts can survive for, at this point, almost a decade and a half.

Is it even possible to fix this, and of so, is it something you would consider?

Answer: Hey there, @maplerosekisses!

It is possible to fix this, and we would like to fix it, but it’s a daunting problem at Tumblr’s scale. Buckle up for storytime.

Long, long ago, Tumblr was created, and in the beginning, there weren’t even notes on posts. There weren’t even reblogs or likes. In fact, we were one of the first platforms to introduce the heart icon and the concept of “likes”! We created the reblog! Back in those days, each of these actions were tracked separately. Likes were tracked in one database table and reblogs weren’t tracked at all as notes. When we introduced replies, those had yet another way of being tracked in our database. Totally separate entities on the platform for years.

Eventually, we wanted to consolidate these into one number—so we had to count each of those different places. That’s horribly inefficient, and as Tumblr grew in size and popularity, this became a bottleneck that hurt the whole platform. So one of the things we did was to invent a new denormalized database table called “notes,” to track all of these different things in one place so we could easily count them. We still have that table, and it’s still the fallback whenever we need to count the notes on a post.

But this itself is ancient history. Since then, the product has changed even more, and we removed replies and re-added them later, back in 2015 or so—and made some changes in that process to help further improve efficiency. These improvements allowed us to include media in the notes view, and be able to split out replies versus reblogs-with-comment versus likes (kind of going back to the way it was originally.) Even then, we didn’t yet support showing tag usage in the notes—that would come even later.

In the process of making all of these changes for efficiency and functionality, we had to ask ourselves, as you point out: should we try to backfill these new database tables with all of the data from before? For a long while, we were using both systems to power the notes view, so we could display as much information from “before” as we could. Eventually, we didn’t need to do that anymore, because the number of people scrolling back to that “before” time became infinitesimally small. And that's the situation we’re in today.

Because if we wanted to backfill the data, we would need to process literally tens of billions of posts and notes from before 2015, at a conservative estimate. Let’s say it’s 10,000,000,000, for the sake of argument: if we started an automated process to go through them at ~100 per second (which would be relatively safe at our scale, so Tumblr doesn’t break as we’re digging up these old rows in the database), it would take over three years of continuous operation to complete that task.

In situations like this, we have to ask ourselves if that’s worth it. So far, the answer we’ve determined is no. But we may find a more efficient way to do it, there’s undoubtedly a way, and when we do, we will re-evaluate the decision again. We hope that makes sense—trying to make changes to Tumblr can be really, really hard.

But thank you for your question. We appreciate them and hope that goes some way to answering your query. Keep 'em coming, y'all.

[finlaure13]

CANADA

Ontario to exempt planned Highway 413 from Environmental Assessment Act

JEFF GRAY QUEEN'S PARK REPORTER

TORONTO

PUBLISHED OCTOBER 24, 2024

Ontario has tabled legislation that would exempt its marquee proposed Highway 413 project from the need for a full-scale environmental assessment, prompting some critics to renew calls for Ottawa to do more to ensure the protection of endangered species along the 52-kilometre route.

Transportation Minister Prabmeet Sarkaria’s bill this week also includes new measures that would make it easier for government officials or contractors to inspect land they intend to expropriate for the highway. Under the legislation, anyone who tries to block a site inspection along the route of the 413 could face up to $10,000-a-day in fines.

In 2022, Premier Doug Ford’s Progressive Conservative government made the planned 413, which would arc through protected Greenbelt land around the Greater Toronto Area west and north of Toronto, a centrepiece of its re-election campaign.

But efforts since to get shovels in the ground have run into roadblocks, including a now-abandoned move by the federal government to do its own lengthy, detailed environmental evaluation of the project. Despite demands from environmental groups and opposition leaders, Ottawa has so far declined to revive this review.

Ontario’s new bill, tabled Monday, would exempt the 413 project from the province’s own Environmental Assessment Act, which would have required a lengthy study that considered the highway’s impacts and looked at alternative options.

The new bill would still require the government to produce a report on how it intends to mitigate the project’s environmental effects. But it would also allow the government to keep any studies that “contain information about sensitive natural or cultural heritage matters” out of the public eye.

The construction of Highway 413 would negatively impact the environment, not help it, say scientific researchers

Speaking to reporters, Mr. Sarkaria said the legislation would allow construction on early works to start sooner. He also again declined to provide a cost estimate cost for the 413, citing the need not to tip off bidders for the contract to build it. The province’s Auditor-General pegged the likely bill at $4-billion in 2022, but critics say the final price tag could be double that.

Tim Gray, executive director of the group Environmental Defence, said that, while the government had long said it intended to streamline the environmental assessment process for the highway, this bill would essentially gut it.

“It’s pretty perfunctory window-dressing. They’re getting rid of any of the substantive work that they were actually planning on doing and just saying, ‘We’ll publish a report,’” he said in an interview.

Mr. Gray said the federal government needs to step back in and redesignate the project for more extensive environmental scrutiny under its newly rewritten Impact Assessment Act, an idea supported by Ontario NDP Leader Marit Stiles and Green Leader Mike Schreiner. In an e-mail, Liberal Leader Bonnie Crombie would not say whether Ottawa should step in but said she would investigate the highway if she became premier.

Earlier this year, the federal government agreed to back off its battle with the province over the 413, after settling a court challenge that had been initiated by Ontario over whether the impact assessment Ottawa had imposed on the highway was overstepping. Last year, the Supreme Court of Canada had ruled that the federal Impact Assessment Act strayed too far into provincial jurisdiction, but Ottawa has since rewritten the legislation to narrow its scope.

The federal government had agreed to set up a working group to deal with Ontario on the 413 and ensure effects on endangered species are mitigated. And to build the highway, the province will still need to seek federal permits under endangered species legislation.

In an e-mail, Ottawa’s Impact Assessment Agency said Ontario’s bill would not affect the federal-provincial working group. However, the agency also said requests to designate projects for a full federal impact assessment would be addressed on a “case-by-case basis.”

Ontario’s new legislation also includes changes that would allow the government or its contractors sweeping powers to go onto land on or near the highway’s path to do site inspections, soil sampling or other such work – without any objection from landowners. Anyone who hinders such an activity would face up to a $50,000 fine for a first offence, plus fines of up to $10,000 a day. A second offence could cost up to $100,000.

The government would also be allowed to seek an order from a justice of the peace, rather than a Superior Court judge, authorizing it to enter a property or a dwelling or even take possession of land it “has the right to possess.”

Veteran Toronto expropriation lawyer Shane Rayman said the changes will likely speed up preliminary work and the land acquisition process. However, he said, the new powers would be “unusual” in Ontario and give the government more of an upper hand with landowners facing expropriation.

“This act gives the government remarkably broad powers to possess land, to enter properties and to put down resistance,” Mr. Rayman said in an interview. He warned that if the new powers are not exercised fairly, it could “be fairly harmful to property owners.”

@allthecanadianpolitics

[reasonsforhope]

"For the first time, genetically modified pig kidneys provided “life-sustaining kidney function” during the course of a planned seven-day clinical study—a first step in addressing the critical crisis worldwide of kidney donor organ shortage.

The University of Alabama’s pre-clinical human study at Birmingham also advances the science and promise of xenotransplantation as a therapy to potentially cure end-stage kidney disease—just as a human-to-human transplants can.

“It has been truly extraordinary to see the first-ever preclinical demonstration that appropriately modified pig kidneys can provide normal, life-sustaining kidney function in a human safely and be achieved using a standard immunosuppression regimen,” said UAB transplant surgeon scientist Jayme Locke, M.D., director of UAB’s Comprehensive Transplant Institute and lead author of the paper...

The peer-reviewed findings published last month in JAMA Surgery describes the pioneering pre-clinical human research performed on a recipient experiencing brain death...

The pre-clinical human brain death model developed at UAB can evaluate the safety and feasibility of pig-to-human kidney xenografts, or transplants, without risk to a living human. It is named for transplant pioneer Jim Parsons, an organ donor whose family generously donated his body to advance xenotransplant kidney research, like the latest patient did.

A Critical Need

Kidney disease kills more people each year than breast or prostate cancer, while more than 90,000 people are on the transplant waiting list. More than 800,000 Americans are living with kidney failure and 240 Americans on dialysis die every day. The wait for a deceased donor kidney can be as long as five to 10 years, and almost 5,000 people per year die waiting for a kidney transplant.

Groundbreaking Study Details

The 52-year-old study subject for this research lived with hypertension and stage 2 chronic kidney disease, which affects more than one in seven U.S. adults, or an estimated 37 million Americans. As part of this study, the subject had both of his native kidneys removed and dialysis stopped, followed by a crossmatch-compatible xenotransplant with two 10 gene-edited pig kidneys, or UKidney.

The transplanted pig kidneys made urine within four minutes of re-perfusion and produced more than 37 liters of urine in the first 24 hours. The pig kidneys continued to function as they would in a living human for the entirety of the seven-day study. Also, the kidneys were still viable at the time the study was concluded.

“In the first 24 hours these kidneys made over 37 liters of urine,” said Dr. Locke. “It was really a remarkable thing to see.” ...

Gene editing in pigs to reduce immune rejection has made organ transplants from pigs to humans possible. The natural lifespan of a pig is 30 years, they are easily bred, and they have organs of similar size to humans. Genetically modified pig kidneys have been extensively tested in non-human primates, and the addition of UAB’s preclinical human research model—the Parsons Model—now provides important information about the safety and efficacy of kidneys in human transplant recipients."

-via Good News Network, September 17, 2023

[nostalgebraist]

clarification re: ChatGPT, " a a a a", and data leakage

In August, I posted:

For a good time, try sending chatGPT the string ` a` repeated 1000 times. Like " a a a" (etc). Make sure the spaces are in there. Trust me.

People are talking about this trick again, thanks to a recent paper by Nasr et al that investigates how often LLMs regurgitate exact quotes from their training data.

The paper is an impressive technical achievement, and the results are very interesting.

Unfortunately, the online hive-mind consensus about this paper is something like:

When you do this "attack" to ChatGPT -- where you send it the letter 'a' many times, or make it write 'poem' over and over, or the like -- it prints out a bunch of its own training data. Previously, people had noted that the stuff it prints out after the attack looks like training data. Now, we know why: because it really is training data.

It's unfortunate that people believe this, because it's false. Or at best, a mixture of "false" and "confused and misleadingly incomplete."

The paper

So, what does the paper show?

The authors do a lot of stuff, building on a lot of previous work, and I won't try to summarize it all here.

But in brief, they try to estimate how easy it is to "extract" training data from LLMs, moving successively through 3 categories of LLMs that are progressively harder to analyze:

"Base model" LLMs with publicly released weights and publicly released training data.

"Base model" LLMs with publicly released weights, but undisclosed training data.

LLMs that are totally private, and are also finetuned for instruction-following or for chat, rather than being base models. (ChatGPT falls into this category.)

Category #1: open weights, open data

In their experiment on category #1, they prompt the models with hundreds of millions of brief phrases chosen randomly from Wikipedia. Then they check what fraction of the generated outputs constitute verbatim quotations from the training data.

Because category #1 has open weights, they can afford to do this hundreds of millions of times (there are no API costs to pay). And because the training data is open, they can directly check whether or not any given output appears in that data.

In category #1, the fraction of outputs that are exact copies of training data ranges from ~0.1% to ~1.5%, depending on the model.

Category #2: open weights, private data

In category #2, the training data is unavailable. The authors solve this problem by constructing "AuxDataset," a giant Frankenstein assemblage of all the major public training datasets, and then searching for outputs in AuxDataset.

This approach can have false negatives, since the model might be regurgitating private training data that isn't in AuxDataset. But it shouldn't have many false positives: if the model spits out some long string of text that appears in AuxDataset, then it's probably the case that the same string appeared in the model's training data, as opposed to the model spontaneously "reinventing" it.

So, the AuxDataset approach gives you lower bounds. Unsurprisingly, the fractions in this experiment are a bit lower, compared to the Category #1 experiment. But not that much lower, ranging from ~0.05% to ~1%.

Category #3: private everything + chat tuning

Finally, they do an experiment with ChatGPT. (Well, ChatGPT and gpt-3.5-turbo-instruct, but I'm ignoring the latter for space here.)

ChatGPT presents several new challenges.

First, the model is only accessible through an API, and it would cost too much money to call the API hundreds of millions of times. So, they have to make do with a much smaller sample size.

A more substantial challenge has to do with the model's chat tuning.

All the other models evaluated in this paper were base models: they were trained to imitate a wide range of text data, and that was that. If you give them some text, like a random short phrase from Wikipedia, they will try to write the next part, in a manner that sounds like the data they were trained on.

However, if you give ChatGPT a random short phrase from Wikipedia, it will not try to complete it. It will, instead, say something like "Sorry, I don't know what that means" or "Is there something specific I can do for you?"

So their random-short-phrase-from-Wikipedia method, which worked for base models, is not going to work for ChatGPT.

Fortuitously, there happens to be a weird bug in ChatGPT that makes it behave like a base model!

Namely, the "trick" where you ask it to repeat a token, or just send it a bunch of pre-prepared repetitions.

Using this trick is still different from prompting a base model. You can't specify a "prompt," like a random-short-phrase-from-Wikipedia, for the model to complete. You just start the repetition ball rolling, and then at some point, it starts generating some arbitrarily chosen type of document in a base-model-like way.

Still, this is good enough: we can do the trick, and then check the output against AuxDataset. If the generated text appears in AuxDataset, then ChatGPT was probably trained on that text at some point.

If you do this, you get a fraction of 3%.

This is somewhat higher than all the other numbers we saw above, especially the other ones obtained using AuxDataset.

On the other hand, the numbers varied a lot between models, and ChatGPT is probably an outlier in various ways when you're comparing it to a bunch of open models.

So, this result seems consistent with the interpretation that the attack just makes ChatGPT behave like a base model. Base models -- it turns out -- tend to regurgitate their training data occasionally, under conditions like these ones; if you make ChatGPT behave like a base model, then it does too.

Language model behaves like language model, news at 11

Since this paper came out, a number of people have pinged me on twitter or whatever, telling me about how this attack "makes ChatGPT leak data," like this is some scandalous new finding about the attack specifically.

(I made some posts saying I didn't think the attack was "leaking data" -- by which I meant ChatGPT user data, which was a weirdly common theory at the time -- so of course, now some people are telling me that I was wrong on this score.)

This interpretation seems totally misguided to me.

Every result in the paper is consistent with the banal interpretation that the attack just makes ChatGPT behave like a base model.

That is, it makes it behave the way all LLMs used to behave, up until very recently.

I guess there are a lot of people around now who have never used an LLM that wasn't tuned for chat; who don't know that the "post-attack content" we see from ChatGPT is not some weird new behavior in need of a new, probably alarming explanation; who don't know that it is actually a very familiar thing, which any base model will give you immediately if you ask. But it is. It's base model behavior, nothing more.

Behaving like a base model implies regurgitation of training data some small fraction of the time, because base models do that. And only because base models do, in fact, do that. Not for any extra reason that's special to this attack.

(Or at least, if there is some extra reason, the paper gives us no evidence of its existence.)

The paper itself is less clear than I would like about this. In a footnote, it cites my tweet on the original attack (which I appreciate!), but it does so in a way that draws a confusing link between the attack and data regurgitation:

In fact, in early August, a month after we initial discovered this attack, multiple independent researchers discovered the underlying exploit used in our paper, but, like us initially, they did not realize that the model was regenerating training data, e.g., https://twitter.com/nostalgebraist/status/1686576041803096065.

Did I "not realize that the model was regenerating training data"? I mean . . . sort of? But then again, not really?

I knew from earlier papers (and personal experience, like the "Hedonist Sovereign" thing here) that base models occasionally produce exact quotations from their training data. And my reaction to the attack was, "it looks like it's behaving like a base model."

It would be surprising if, after the attack, ChatGPT never produced an exact quotation from training data. That would be a difference between ChatGPT's underlying base model and all other known LLM base models.

And the new paper shows that -- unsurprisingly -- there is no such difference. They all do this at some rate, and ChatGPT's rate is 3%, plus or minus something or other.

3% is not zero, but it's not very large, either.

If you do the attack to ChatGPT, and then think "wow, this output looks like what I imagine training data probably looks like," it is nonetheless probably not training data. It is probably, instead, a skilled mimicry of training data. (Remember that "skilled mimicry of training data" is what LLMs are trained to do.)

And remember, too, that base models used to be OpenAI's entire product offering. Indeed, their API still offers some base models! If you want to extract training data from a private OpenAI model, you can just interact with these guys normally, and they'll spit out their training data some small % of the time.

The only value added by the attack, here, is its ability to make ChatGPT specifically behave in the way that davinci-002 already does, naturally, without any tricks.

[losergames]

an update! - 6th june '24

oh my god it's june, where does the time go --

i am still active! even though the blog isn't very alive, i'm still around and writing. there's not much to update on as i'm in a pretty slow period at the moment, re-evaluating bits and pieces of the game as i continue to move forward with episode 04. i am 20k in - so about a third of what i'm estimating the total episode word count to be.

i've also recently had a little breakthrough regarding the PC's father in the game! i intentionally left this character vague to give myself room for if i ever wanted to flesh him out or for people to head canon their own ideas about the relationship etc. but, now i actually have a pretty solid idea of where i want to take him! it'll mostly be background plot, something that could be developed later on if i decide to go that way.

this is also crazy late but thank you to everyone that entered my milestone giveaway a few months back! i ended up doing two portraits instead of the one - felt it would be fairer and a nice prize for both people! Maz and Sunny were selected for the two short stories as well. i've got some ideas down and slowly writing them but i do not have a time or date for when they'll be here - sorry!

lastly, i do really appreciate people checking in and seeing if the game is still around; the support means everything, even in these slower periods, and it's really cool to see people still tuning in and finding the game for the first time. thank you all again and have a lovely day ❤️

[covid-safer-hotties]

Also preserved in our archive

TL;DR: Covid prevention measures worked so well, they controlled and disrupted seasonal flu too. We should be taking more steps to keep everyone safe and healthy from airborne illness. Science shows that it makes a big difference.

Seasonal influenza epidemics impose substantial burdens on health care systems and cause >5 million hospitalizations of adults each year. The current approach to influenza vaccine development requires comprehensive surveillance of circulating strains, which are constantly moving from continent to continent.

The reduction in global human mobility during the COVID-19 pandemic provided a unique opportunity to evaluate how seasonal influenza is impacted during pandemics.

In a new study, an international team of researchers, including from the University of Oxford, Fudan University, and KU Leuven, combined data on the spread of seasonal influenza, its genetic makeup, and international travel patterns to study how seasonal flu viruses moved and evolved.

The paper "COVID-19 pandemic interventions reshaped the global dispersal of seasonal influenza viruses" was published in the journal Science.

This approach helped to estimate how long the viruses remained in certain regions during periods of high and low volumes of international travel and how their genetic diversity changed before, during, and after the COVID-19 pandemic.

During the COVID-19 pandemic, seasonal influenza levels dropped worldwide due to COVID-19 pandemic-related restrictions on population movement and mixing. However, the subsequent rapid bounce back of influenza once air travel returned to pre-pandemic levels showed that the virus was in most cases maintained during the pandemic with continued viral movements and accumulation of genetic diversity.

Lead author of the study Zhiyuan Chen (University of Oxford and Fudan University) says, "It was remarkable how quickly seasonal flu re-established to a pre-pandemic equilibrium just a few years after the height of the COVID-19 pandemic."

Tropical climates, like those found across South and East Asia, allow for continued flu transmission year-round, thereby creating a broader range of flu strains and increasing overall viral diversity. The increased capacity for virus genomic surveillance during the COVID-19 pandemic has provided more detail on the role of other regions, such as Africa and West Asia, in the global circulation of influenza.

These regions also showed evidence of sustained transmission and during the pandemic had relatively less restrictions on movement, in part due to lower levels of COVID-19 transmission.

Co-author Professor Moritz Kraemer (Department of Biology and Pandemic Sciences Institute, University of Oxford) says, "Increased genomic surveillance capacity established during the COVID-19 pandemic means that we are finally getting a deeper insight into the global distribution patterns of seasonal flu and other respiratory viruses.

"These novel and large openly accessible datasets provide an opportunity to learn about the intricate relationships of climate, co-circulating viruses, and human behavior."

Further, with this increased global capacity for surveillance of viruses, it might be possible to better monitor seasonal influenza to reduce the risk of vaccine mismatches, help inform more effective interventions, and reduce the burden of seasonal influenza on our health care systems. This is especially relevant as more regions become suitable for year-round circulation of influenza with changes in climatic conditions.

Co-author Professor Hongjie Yu from Fudan University says, "Further efforts should still focus on the continuing surveillance of seasonal influenza viruses and other respiratory pathogens, particularly resource-limited regions. The established surveillance systems for seasonal respiratory pathogens could also play an extremely vital role when the next pandemic emerges in the future."

More information: Zhiyuan Chen et al, COVID-19 pandemic interventions reshaped the global dispersal of seasonal influenza viruses, Science (2024). DOI: 10.1126/science.adq3003. www.science.org/doi/10.1126/science.adq3003

[lady-raziel]

insane to me that Bethesda purchased the rights to the ENTIRE Fallout IP for only $5.75 million USD in 2007...like honestly, say what you will about the quality of Bethesda games, but it's kind of insane what the series of games they've made have done to the value of the IP. There hasn't been any re-evaluation of how much the Fallout franchise IP is worth in the time since Fallout 3 was made (so it's hard to say how much every game since has boosted the value in concrete numbers), but estimates indicate that because of the Amazon Prime show alone Bethesda made more than $80 MILLION DOLLARS-- and this is just in revenue of people going back to revisit already existing games and stuff. Fallout is more likely than not a billion+ dollar franchise at this point, and this is even if they don't make another full game for years!

Like that's a nuts return on investment. Kinda explains why they are in no rush to make a new game or why it might not even matter if the next game is "good." They've already made bank.

[cl34rb3ll]

james somerton apology video is so interesting to me

-he said he didn’t mean for the misinfo to end up on the channel? like oops? i said a blatant lie! my bad idk how that happened

-thanking hbomber for setting up the fund for “those who were wronged” (james somerton, deleted apology video) i know this is nitpicky but his language keeps shifting the blame from himself. he won’t say “i didn’t credit them” or “i wronged them” like james. they didn’t just HAPPEN to have their work stolen. the universe didn’t intervene and force their work uncredited into your videos YOU stole from them YOU wronged them.

-“i’m not trying to make this a sob story” (james somerton, deleted apology video) then bringing up the bad things happening to him. like. i don’t feel bad that your friends left you for being a thief profiting off the work you stole from queer people. i also just. don’t care. like, i really do hope that he gets help for his mental health. i hope he gets better and he goes to a class on MLA or something. i really do not want him to die. but thats not what this is about. this is about the people he hurt, not the fact that the consequences of his own actions have impacted his mental health.

-refusing to give actual numbers to his income. just saying that people were over estimating. normally peoples income is none of my business but he was profiting off stolen work and is substantial evidence to lead people to believe he was lying about his financial situation to get money from fans.

-there were plenty [of videos] that didn’t have any uncredited stuff in them” (james somerton, deleted apology video) i’m starting to think james didn’t watch the video😨

-side note i don’t understand how this man could go to university and not understand proper citation???

-him saying that him hating women just isn’t true??? then why did he say all that shit??? like bro? you can just admit that you were misogynistic. you can say that. you can say “i was misogynistic, i said misogynistic things, i didn’t understand that i was wrong but i am trying to be better and re-evaluate my views” dude, i’m female (ftm tho, he/him only!) and i still have to occasionally catch myself when i think something misogynistic, despite literally being a victim of misogyny. just be better, don’t pretend you were never wrong. also him saying that he doesn’t think those things, that they were just writing videos too fast? misogyny doesn’t just End Up in your videos because you wrote too fast?? it has to come from somewhere?

-he says that he never wrote misinformation but he also says he isn’t trying to throw nick under the bus, that him and nick are 50/50 partners? which is it?

[collegiate-collective]

Starting my German learning Journey!

Daily learning things!

🌱 vhs-Lernportal

This is a free online learning tool that offers lessons, tests, tutors and many activities to assist with learning. The first level has 12 lessons, with three sections per lesson. The goal is to complete one section per day

🌱 Nico's Weg

This tool has been recommended to me so many times. Essentially it's more of an immersion course. there are 18 lessons in the first level, so I'll be completing one lesson every other day

🌱 Duolingo

Now, I'm not using this to learn German, but more so to practice. It's easy, it's simple, and it works. Seeing as I am surrounded by German all the time, I don't really need as much language exposure.

I'll probably re-evaluate how well it's going in about a month! That's when I estimate I'll have completed the first level. I'll probably update here!

[butchsophiewalten]

FindJackWalten update and More! 1/13/24

Some of you may be familiar with the twitter thread Martin made a few days ago, talking mostly about stress and burnout and how he was planning to take some days away to relax, but that he was starting to feel like his current approach to the series was unsustainable and that he was probably going to have to re-evaluate things.

Cue last light! Or this morning, really, when two FindJackWalten pages updated. First we have the main page:

Findjackwalten.com

This page has updated to, seemingly, just show the title of The Walten Files: 4? (Or, if your screen brightness is configured specifically, a little bit more...)

This is really all a single image, which appears like This if i copy/paste it from the website:

But if taken into a photo editing program and color-adjusted appears like this:

Hi Bon! And the titles of... the next three Walten Files episodes?

We get a better understanding of this when we take a look at Findjackwalten.com/name-me:

The first short part of this page is about the same, except the top text has changed to read "THE WALTEN FILES SEASON ONE (2020-2024)". Then the first three episodes are listed as they were, and then we see the new stuff

The three new episodes we'd seen before, subtitled as "CyberFun Tech's Collection"? It seems to me like The Walten Files 4 has been split up between three smaller, more manageably-sized videos. Considering TWF4 was estimated to be about an hour and 20 minutes long, I'm envisioning these new episodes to be within the 20-30 minute range, much more like TWF2 was.

Also importantly, it looks like the end of TWF4, now TWF6, is going to be the Season 1 Finale. This leaves the numbers the same, but the actual plot is different. Originally Season 1 was going to have a two-part finale revisiting Sophie and Jenny's narrative where we left off in BunnyFarm.

Speaking of, if you highlight all the text on the page, (ctrl+A is a help,) something new appears:

That all said, it looks like these next 3 episodes are all slated to release sometime in February? Which is exciting.

We also have some new tweets from Martin:

As well as some replies he left on some other people's tweets that give a little more context and explanation for everything:

I'm like, ridiculously excited, honestly? I think this was a really good decision for him to make. Pivoting to smaller episodes rather than really big theatrical ones is a really good move for him, I think. I also think this is going to be better for a lot of fans who were starting to feel like the ever-increasing scope and standard of the series was starting to feel kind of... cumbersome? I've seen more than a few people worried than an almost feature-length episode was going to feel really overwhelming.

So yeah. I think this was a good decision both to make The Walten Files more accessible for new and returning fans and to make it less of an incredible burden on its creator. And that's just a win-win to me. I'm so pleased.

[voidami]

The Great leap forward

The Great Leap Forward (GLF) and the associated famine in China from 1959-1961 have often been subjected to significant myth-making, much of which exaggerates the death tolls and distorts the causes and outcomes. Commonly, figures as high as 30 to 45 million are cited, largely based on estimates from Western demographers like Judith Banister and Frank Dikötter. However, deeper analysis and more recent scholarship—particularly the work of mathematician Sun Jingxian—suggest that these numbers are highly inflated and do not account for various factors that contribute to a more realistic understanding of the famine.

Re-Evaluating Death Toll Figures

The initial myth that "tens of millions" died during the GLF has been largely debunked through more careful statistical and historical analysis. According to Sun Jingxian, the estimated death toll from the famine was around 3.66 million deaths, which includes deaths from various causes, not just starvation. This number is 8% of the 45 million figure posited by Dikötter, and 12% of Banister's estimate of 30 million. Sun’s work shows that deaths during the famine were not caused solely by starvation but included other "unnatural deaths," such as deaths from diseases exacerbated by malnutrition. This reevaluation places the famine in a context comparable to other major historical famines in China, which also had multifaceted causes rooted in poverty and ecological challenges.

A key point in Sun’s work is the differentiation between year-end registered household population and total population. If a similar methodology were applied to the U.S. during the Great Depression, it could lead to vastly inflated death toll estimates, anywhere from 67 to 170 million deaths, a number that is patently absurd. This comparison highlights the dangers of relying on simplistic population metrics without understanding the nuances of registration systems and migration patterns.

Natural Disasters and Systemic Factors

The famine was exacerbated by severe natural disasters. Droughts, floods, and other ecological crises significantly reduced grain production during the period. Claims that systemic factors like the public canteen system or the planned economy were to blame for the famine are largely based on misconceptions. For example, the public canteen system is often portrayed as a "Tragedy of the Commons" scenario, where people supposedly over-consumed resources, leading to shortages. However, this system was not widely implemented across China, and even where it was, only 22% of canteens offered unrestricted supply. Most canteens only provided extra grain for laborers during harvest seasons, making it an unlikely culprit for mass famine.

Similarly, the notion that the planned economy was responsible for the famine ignores the fact that the planned economy had been in place long before and continued after the famine without leading to similar crises. The city-oriented grain supply system is another factor often cited, with claims that urban areas drained grain resources from rural farmers. While there was some truth to this, it overlooks the reselling of grain to rural areas during the famine, which mitigated the impact of urban preferences to some degree.

Historical Context of Chinese Famines

China has a long history of recurring famines, particularly under imperial rule and during the early republican period. For centuries, China’s agrarian society was vulnerable to natural disasters, ineffective governance, and foreign exploitation, leading to regular, catastrophic famines. For example:

The Great North China Famine (1876–1879) killed an estimated 9-13 million people.

The 1907 Famine resulted in approximately 24 million deaths.

In the early 20th century, the 1928-1931 famine caused 3-6 million deaths, while the 1936-1937 famine claimed another 5 million lives.

Annual death tolls from famine ranged between 2-8 million during turbulent periods like the Warlord Era and the Sino-Japanese War, illustrating the chronic nature of famine in China prior to Mao’s leadership. It is crucial to note that the famine during the GLF was the last major famine in Chinese history, marking a significant shift from previous eras where famines were a persistent, almost annual occurrence.

Human and Political Factors

Mao Zedong is often held responsible for the failures of the GLF, but the reality is more complex. While Mao did push for rapid industrialization and agricultural transformation, many key decisions during the famine were made collectively by the Central Committee of the Communist Party of China (CCCPC). By the time the famine peaked, Mao had already retired to a secondary position, leaving leaders like Liu Shaoqi and Deng Xiaoping to oversee much of the national response. Moreover, local cadres were often reluctant to report real conditions due to the political atmosphere, further delaying disaster relief efforts.

The Sino-Soviet split also played a role, as the deterioration in Sino-Soviet relations curtailed grain imports and exacerbated the famine. Still, the Chinese government took various actions to mitigate the disaster, including grain imports, agricultural policy adjustments, and efforts to inform the public of the situation and acknowledge mistakes. While not everything worked as planned, these measures undoubtedly reduced the scale of the disaster.

Criticism of Exaggerated Narratives

Many narratives today selectively present facts about the famine, often exaggerating its scale to make ideological arguments against socialism and the Chinese Communist Party (CPC). These narratives aim to invalidate the first 30 years of the PRC and undermine the CPC’s achievements in nation-building. Comparisons between death rates in India and China during the famine are telling: in 1960, at the height of the famine, China's death rate was 2.543%, nearly identical to India's rate of 2.4%—yet only China's rate is deemed problematic in Western critiques.

Sun Jingxian's research challenges the widely held assumption that the GLF was an unmitigated disaster caused by ideological fervor. Instead, he presents a more balanced view, acknowledging that the famine was a tragic event, but one that was not unprecedented in China's history and was largely mitigated through the CPC’s efforts.

A common myth surrounding the Great Leap Forward is that the Four Pests Campaign, particularly the killing of sparrows, led to crop failures by causing a surge in insect populations, especially locusts. However, this claim is largely exaggerated.

While sparrows were targeted for eating grain seeds, they also consumed insects, and their eradication may have had some ecological impact. However, sparrows were not the primary predator of locusts, and other natural factors, such as floods and droughts, had a far greater effect on crop failures during the GLF.

Moreover, the Chinese government quickly adjusted its approach, replacing sparrows with bed bugs on the pest list by 1960. The main causes of the famine were natural disasters, bureaucratic mismanagement, and external factors like the Sino-Soviet Split, not the sparrow policy. This myth has been overstated in an attempt to discredit Mao’s policies and oversimplify the famine’s complex causes.

The Great Leap Forward famine was a tragic event, but it must be understood in the broader context of Chinese history and the global struggles of agrarian societies transitioning to modernity. The death toll, while significant, has been exaggerated in Western accounts, and many of the purported causes of the famine are based on ideological hostility rather than material analysis.

By considering the natural disasters, bureaucratic failings, and political climate that contributed to the famine, we can arrive at a more accurate picture, one that situates the GLF within a long history of famines in China. Moreover, the measures taken by the Chinese government, while not perfect, helped to prevent future famines, making the 1959-1961 famine the last major famine in China’s history—an achievement that should not be overlooked.

References/sources:

Some links may be omitted due to Tumblr limits but available here: https://voidami.wordpress.com/2024/09/13/the-great-leap-forward/

"The Great Leap Forward: Anatomy of a Historical Catastrophe" by Liu Renwen - Provides a detailed analysis of the GLF and addresses various myths surrounding it.

"Mao’s Great Famine: The History of China’s Most Devastating Catastrophe, 1958–1962" by Frank Dikötter - Examines the famine in detail, including the impact of policies and natural disasters.

"The Great Leap Forward and the Chinese Famine of 1959-1961" by Sun Jingxian - Offers a critical re-evaluation of the death toll and causes of the famine.

"Famine in China: 1959-1961" by Xue Muqiao - Discusses the impact of various policies and natural factors on the famine.

Specific Topics

"The Four Pests Campaign" - An overview of the campaign’s objectives and outcomes. Available in Historical Studies journals.

"Ecological Consequences of the Great Leap Forward: An Evaluation of the Four Pests Campaign" by Hao Yufan - Analyzes the ecological impact of the campaign, including the sparrow policy.

"The Environmental Impact of the Great Leap Forward: A Critical Review" by Li Xiaohua - Discusses the broader environmental impacts of the GLF, including pest control measures.

"Pests, Plagues, and Policy: The Great Leap Forward and Its Ecological Consequences" by Zhao Yao - Examines the myths and realities surrounding pest control during the GLF.

Famine and Death Toll

"China’s Great Leap: The Leap into the Future" by Kong Yiji - Provides statistical analysis of the famine's impact and death toll.

"The Death Toll of the Great Leap Forward: Reassessing Historical Data" by Wang Qing - Re-evaluates historical death toll estimates and their accuracy.

"Famines in China: Historical Perspectives and Modern Understandings" - Analyzes various famines in China’s history, including the GLF.

Additional Resources

Hungry Ghosts: Mao’s Secret Famine - Extensive response to claims around the Great Leap Forward and the associated famine.

Sun Jingxian and the Myth of Mao’s Genocide - Summary of Sun Jingxian’s paper and the debate on the famine's death toll.

Joseph Ball, The Mao Killed Millions Myth: The Last Word?

[felassan]

In this video - “The First Info is the BEST Info! - Anchor Bias” - Mark Darrah talks about anchor bias (over valuing the first piece of information you have on a topic) in the games industry.

he thinks it's very important for game teams to try to get ahead of their sales projections within larger publishing organizations because of this phenomenon. as an example, he relates that for pretty much the entirety of the development cycle of Dragon Age: Inquisition, he was fighting against a very low sales projection set by the sales team because of their anchor bias. the sales team had apparently decided (in part based on the sales of DAII, and using this as a starting point, even though DA:I is a very different game to DAII) that DA:I wasn't going to sell very well, and was going to sell around 2-3 million copies. this projected number never really changed significantly, regardless of whatever additional information was provided to the sales team. any changes to the estimate were incremental improvements only, as opposed to actual re-evaluations of the projection from the ground up. DA:I ended up selling over 10 million copies.

in another example of this from within the EA organization, he relates that during the development of Anthem, there were originally planned to be 5 different Javelin suit types (4 shipped). the fifth was a large, Colossus-type size of suit that had a 'Paladin' vibe going on. it got cut and the organization as a whole "kind've freaked out about it, even though it didn't exist". at that point in time, only the Ranger actually existed. Mark explains that they freaked out because 5 is more than 4, and because of the reduction in a number from the anchor value of 5. the Anthem team had actually gone through a lot of thinking, work, conversation and iteration over months to realize that the game didn't need 5, that in fact 5 would undermine the individuality of the different Javelins, and that with 4 the suits could be made better and more distinct. but the people who were hearing that 5 had gone to 4 (e.g. an executive in a 45 mins long meeting where they were being shown the whole project) hadn't gone through that process themselves.

[source and full video watch link]

[datascience-from-the-trenches]

On March 14th, 2023, OpenAI launched GPT-4, said to be the latest milestone in the company’s effort in scaling up deep learning. As part of its launch, OpenAI revealed details regarding the model’s “human-level performance on various professional and academic benchmarks”. Perhaps none of these capabilities was as widely publicized as GPT-4’s performance on the Uniform Bar Examination, with OpenAI prominently displaying on various pages of its website and technical report that GPT-4 scored in or around the “90th percentile,” or “the top 10% of test-takers”.

This paper begins by investigating the methodological challenges in documenting and verifying the 90th-percentile claim, presenting four sets of findings that indicate that OpenAI’s estimates of GPT-4’s UBE percentile are overinflated.

Summary of results:

The 90% percentile estimates are heavily skewed towards repeat test-takers who failed the July administration and score significantly lower than the general test-taking population.

data from a recent July administration of the same exam suggests GPT-4’s overall UBE percentile was below the 69th percentile, and 48th percentile on essays.

GPT-4’s performance against first-time test takers is estimated to be 62nd percentile, including 42nd percentile on essays.

when examining only those who passed the exam (i.e. licensed or license-pending attorneys), GPT-4’s performance is estimated to drop to 48th percentile overall, and 15th percentile on essays.

[thesupreme316]

Is the Grass Greener on the Other Side? EP. 6

Summary: After meeting with Triple H, Y/N starts to evaluate her career and where to go next. So what happens when she meets with one of WWE's biggest competitors?

Word Count: 2.1K

Warnings: None I think, isn't proofread tbh

Notes: I FINALLY GOT INSPO AND TIME SO HERE IT IS

Links to other parts: Part 1, Part 2, Part 3, Part 4, Part 5

August 31st

“Ms. Y/L/N, it is a pleasure to meet you.”

You smiled as you sat down at the table, across from the other person. “Likewise. I’m happy to have this opportunity.”

“I know you and I am just happy to be in a room with such a legend in the making.”

“Mr. Kahn, I’m sure you are surrounded by legends in the making every day.”

“Well, yes to a certain extent; but none of them have the charisma and charm that you possess. And please call me Tony.”

Could you complain about life right now?

HELL NO

You had a lot to be grateful for like this meeting with Tony Kahn.

Everything was going great after the meeting with Paul. You started to feel like your old self again; you made your first public appearance at a fashion show, you started becoming an ambassador for multiple brand deals, you went viral for sitting next to Rihanna and then became the head ambassador for Fenty, and you were a single pringle.

But most importantly, you began to wrestle and train more. Almost every day, you got in the ring and you saw vast improvements even from the last time you stepped in the ring (which was two months ago). And because you were re-entering the public’s eye, your career and where you would end up as people noticed your contract was almost up at WWE (and the fact they haven’t seen you since July). Many companies have been offering you different amounts for you to sign to them.

Now you were here….at AEW…WWE’s biggest competitor since WCW….just “looking around.”

“Well, Tony, I really do appreciate this hospitality. It just softens the blow when it comes down to business.” You joked, making him chuckle.

“Of course. Well, I would say that we are just like everything you’re used to, but we both know that’s a lie.”

“How would you say that you’re different?”

“A big example is creative freedom; we barely use scripts, it’s just a basis and talent says what they want. Gimmicks, ring style, and creative ideas are mostly from the locker room and then are tweaked to make sure it’s okay with our sponsors. I also would say we have a great checks-and-balances system here as well.” You nodded your head as that was your main concern as you didn’t wanna throw out the things that brought you to your status now.

Since hearing that, your heart rate has decreased. You wanted to be yourself and not have to worry about lessening yourself to make others look good. “But I’m curious, since you’re a hot topic and agent, how much have you been getting offered?”

You discussed the amounts that you have been offered from Impact, AAA, NJPW, and Shimmer. Then, you told him about the arrangement that WWE (specifically Hunter) presented to you. Tony’s eyes went wide as he spoke up again, “That is really a good deal. I think William Regal received the same offer. It seems that they only offer that to those they must really like.” You giggled as he continued, “If you want, we can extend the talk to next week so I can give you a proper figure estimate. Until then, we can roam around backstage and meet the talent.”

“That would be great!”

For the next hour, Tony dragged you around backstage, introducing you to many talents that you haven’t had the pleasure of stepping in the ring with yet. You saw some old faces of people you’ve met and were friends with. Some were happy to see you, others not so much. Either way, you caused a wave backstage, you could even feel the tension to a certain extent. Eventually, Tony allowed you to roam by yourself as he attended to some business, which is how you started hanging out with Daniel (who was surprised) yet again.

“I can’t believe you didn’t tell me that you were coming here,” Daniel said playfully hitting your shoulder.

“Surprise? I didn’t want you to spoil anything, cause you know, you have a big mouth.” You said making him do a fake offended face. It was nice joking with him, it kind of helped relieve you of the stress and tension that you were feeling.

“How dare you?” He said laughing, he gave you a hug. You breathed in his cologne as he breathed in your perfume, both being delighted in the scents. You felt great being in familiar arms, especially of a man who you keep close to your heart. The same could be said for Daniel, he was legitimately grateful and glad to see you, as he normally is. “But I am happy you’re here. Now we can-“

“Y/N!” A voice from behind you cut him off. You broke away from the hug and turned around to see a fluffy-haired Hook with a wide smile walking up. Squealing you ran up to him and hugged him, excited to see him. “I heard that you were around, had to see it to believe it.” He said pulling away from you, allowing his hand to fall at your waist.

“Yeah, I’ve been sitting here catching up with Danny.” You smiled while motioning towards him. Daniel’s face became stoned real quick when Hook entered his view. His eyes traveled and landed on the spot where Hook’s hand was sitting on your waist. It bothered him. It bothered him that you are allowing him to touch you, let alone allowed your attention to go towards him. Daniel smacked his teeth in annoyance as Hook smirked.

“Garcia,” Hook greeted.

“Hook.” He responded.

Hook was loving this; he loved how jealous Daniel was getting and knowing he can’t do a damn thing about it as long as you were here. He loved that he was touching you like this and not him, no matter if it was a small gesture. “How are you, man?”

Daniel shrugged, not allowing his face to fall. “Good. How are you feeling?”

“Comfortable.”

“A little too comfortable,” Daniel said under his breath as Hook’s smirk widened. If anything, Hook tightened his grip around you before moving his hand further down to your hip. Daniel didn’t like that either, and he didn't like how you weren't doing anything about it, as if you were oblivious to everything. He cleared his throat, "So, how do you feel about your big match coming up?"

Hook shrugged and briefly flicked his eyes to you, "Nothing I can't handle."

"To be honest, I think it might be too much for you."

Your eyes widened, you didn't know what was going on but you didn't like the direction it was heading in. Hook squinted at Daniel, making his hand drop from your body completely. "What do you mean?"

"I just think you're biting off more than you can chew. Considering that you are just a rookie after all. Some prizes are just too big for you to win," Garcia said smirking as Hook was now stone-faced.

"Well then Garcia, what about your match?"

"Fine, I'm confident that I'm gonna win and everything that comes with it."

Hook chuckled, "Okay, but what if it's past your time? That you don't even have a chance anymore? Or that the opportunity for you has passed you by." He stepped closer to Daniel, who also stepped forward to the point that they were almost nose-to-nose.

"I still got my shot. I haven't failed-"

"Yet. That is until someone else better comes into play."

The tension in the air was becoming so thick that you couldn't even cut it….without a chainsaw. You couldn't even grab either of their attention as they were silently challenging each other. As the staring contest started between the two men, a stagehand came up to Daniel to let him know his segment was starting. Daniel, without any words, gave you a hug and left to go to the stage.

"What was that about?" You asked as Hook looked at Daniel's figure walking away.

"Don't worry your pretty head about it." Hook said, "Everything's gonna be okay."

September 7th

Today was a great Wednesday…a lot more fun and stress-free in comparison to last Wednesday. That's also probably because you were at a party. Not just any party, but Daniel's celebratory party for winning the ROH Pure Championship. It's been a long day for you as you were stuck in meetings all day, watched Dynamite backstage (which caused an uproar online), explored the locker room, and built and re-established bonds. You were tired, but you put that aside for Daniel because he deserves it; the title, the happiness, and everything else that came with it as he said.

But you still needed to try to get all your ducks in a row for the next major stage in your life. So you were, in the bathroom, on the phone with Mercedes and Trinity, catching them up on everything for the first time in a while.

"Well, this has been a very eventful month for you so far." Mercedes said, "But, hey, you deserve it. You're getting back on your feet and back out in society."

"You say that as if I've been isolated from society." You pouted.

"You basically have been, girl!" Trinity exclaimed, making you roll your eyes. "Now, go enjoy yourself at the party. You're supposed to be having a good time and here you are talking about work and stress. There will be a time for all that later."

Mercedes agreed, "Girl, go! I'm sure that Garcia would like to see you, especially with how you look right now." You were dressed in the best dress you could find with heels that hurt your feet so much.

You all exchanged goodbyes before hanging up. After fixing your makeup, you exited the bathroom to find your best friend on the floor like there was no tomorrow. His eyes lit up even more as he made eye contact with you. Daniel made his way to you, title around his waist and all. "Y/N! I'm so happy to see you! I thought I would have to cancel if I didn't see you!" Daniel said, having his words run together

"First Danny, congrats champ! Second, are you drunk?"

"Nope! Just buzzed, I think." He slurred and laughed, making you laugh. Daniel was so cute when before leaning down. "Do you wanna dance? It's boring without you on the floor."

Laughing, you took Daniel's hand as he lead you to the dance floor. The two of you danced to many songs, twirling and spinning around to the point Daniel gave Isiah his title belt to put away in VIP. You were having a great time, singing at the top of your lungs with him like old times, taking shots, taking various pictures and videos throughout the hour, and not once did you feel stiff or feel any tension. Soon the songs started to slow down, making the dance floor scarce. Daniel and you were still dancing as if no one was watching, but in the meanwhile, gradually getting closer to each other.

You wrapped your arms around his neck as his arms were wrapped around your waist, moving along to the beat. You were trying to look anywhere except for Daniel's eyes, which were glued on you. Soon, his hand lifted to your face, gently lifting your face up to him. As soon as you made eye contact, you could have sworn either Cupid shot you (and or Daniel) in the ass or you were drinking too much cause the way he was looking at you made you feel like you shrunk two sizes. Blushing, you were drawn into his beautiful brown eyes. You saw the same glint in his eyes that you saw when you first reunited. Despite the club having different colored lights everywhere, you could make out the emotions more clearly than ever before; admiration and lust.

As you looked into his eyes, you could feel those same feelings coming to the surface, or maybe it was the liquor coming to your mind. Apparently, that was clear to Daniel as well. "You are so damn gorgeous," Daniel whispered while looking back and forth between your eyes and lips.

"Danny, you're drunk." You said, trying to excuse his words and make the moment less

"Maybe, but that doesn't mean I don't know a beautiful woman when I see one. One especially that makes me feel…I don't know how to describe it but just know that I love how you make me feel nonetheless." Daniel finished as he pressed a kiss to your forehead.

"And hopefully soon, you'll feel it too." He muttered in your hair.

Taglist: @triscillal @wwenhlimagines @hooks-martin @hookerforhook @sheinthatfandom @zatarias-pandora @jess95laugh