Elon Musk still hasn’t given up on the idea of turning X into an ‘everything app‘, recently adding phone and video calls into what was once Twitter. Now the app is apparently expanding to… TVs. Yup. According to a Fortune report (via Bloomberg) over the weekend, the eccentric billionaire wants people tuning in to, uh, Tucker Carlson, we guess, on Samsung and Amazon TVs as early as next week.

Fortune doesn’t name its sources, only citing an unnamed employee within the company, but this is Musk we’re talking about. Of course he’s got a video app in the works — one that reportedly looks “identical” to YouTube’s own app — a ploy to try and draw customers and compete with YouTube simultaneously. Whether it’ll work is yet to be determined. Our guess? It’ll be abandoned within the year, tail tucked between its legs.

Or, we’ll be proven wrong, and have to bow down to a new overlord of internet TV. We’re not particularly excited about that prospect. That can only happen if X can lay hands on exclusive content and push the app out to a far-wider host of TV brands. The odd Putin interview or shoddy Diablo IV stream might garner at least a few eyeballs. We’ll find out if X’s unnamed would-be YouTube killer has enough gall to do so next week.

Source

MAR10 Day, unsurprisingly, delivered a bunch of Mario news

Yesterday was Mario day. MAR10 Day, et cetera et cetera. As usual, Nintendo served up a slew of Mario-related news on a platter, including a Super Mario Bros. sequel film that reminded us water is wet, and a few other announcements that took our cynicism down a notch. The first involves release dates for two classics remastered: Paper Mario: The Thousand-Year Door and Luigi’s Mansion 2 HD.

Announced in September and June of 2023 respectively, it’s clear Nintendo’s been sitting on these titles for a while — possibly in an attempt to bolster the Switch’s 2024 line-up in the event of a delay to its follow-up console. That, unfortunately, happened. It’s fine. The 23 May release for Paper Mario and a 27 June release for Luigi’s Mansion should do enough to hold us over ’til 2025. Also, a Tears of the Kingdom replay might be on the cards.

The last announcement and possibly the most important involved a teaser for something going by called LEGO Mario Kart (a new game, maybe, or just new sets?) and three new Mario Lego sets hitting shelves this August. The Bowser Express train set is the most expensive of the lot, but honestly, King Boo’s Haunted Mansion set or the Battle with Roy at Peach’s Castle would suit us just fine.

Warner Bros. Discovery isn’t just in the business of deleting movies

You might have thanked Warner Bros. for vaulting Batgirl in 2022, but there’s no denying it set a horrific precedent that’s created a ripple effect across the rest of the business. Coyote vs. Acme is the latest (completed) film to be sent to the bins — and now the company is looking to do something similar for its games.

Several developers under the Adult Swim Games umbrella said that Warner Bros. Discovery reached out to them to essentially tell them that their games would be removed from digital storefronts on PC and consoles. Why? It might hint at the company’s plans for the Adult Swim Games brand — possibly looking to kill it off and watch the tax breaks roll in. Because that’s how business works, right?

Some of the affected developers said they would be republishing their games on Steam, but would lose out on the title’s community pages, Steam achievements, forums, and screenshots. That wouldn’t be the case if Warner would transfer publishing permissions to those developers — a process that takes roughly three minutes and three clicks according to @onemrbean — but isn’t being done due to a ‘lAcK oF rEsOuRcEs’.

You can see a list of the 25 games being removed by the $21 billion company right here.

Source

Your favourite Wordle clones might not be Wordle clones for much longer

Wordle, the word-guessing game that grabbed the world by its genitals in 2022, is looking to stomp out the thousands of clones riding off the back of the Wordle brand, idea, and colour scheme that The New York Times picked up for a cool “undisclosed price in the low-seven figures” in 2022.

The New York Times has reportedly filed several DMCA issues over any Wordle clones still out there, specifically targeting Reactle earlier this week, an open-source clone used to power around 1,900 other versions of the game. The NYT reckons the clones using Reactle’s code did so in “clearly bad faith,” and have been served the same DMCA takedown issue.

“I write to submit a revised DMCA Notice regarding an infringing repository (and hundreds of forked repositories) hosted by GitHub that instruct users how to infringe The New York Times Co.’s (‘The Times‘) copyright in its immensely popular Wordle game and create knock-off copies of the same,” the notice reads.

Expect plenty of those 1,900 or so games taken down to reappear in the coming weeks with Wordle-less names attached, and maybe a fresh coat of paint.

Source

Today, we’re mostly concerned with Pam’s answer – Wordle. We’re sure we aren’t the only ones who think the internet is a slightly different place since the viral word-guessing game dominated watercooler chatter.

According to Progress Bar 2023, we’re just about halfway through the year so it’s about time for another sickenly-addictive browser-based time-waster. Enter The Password Game from New York-based coder Neal Agarwal.

Clear your schedule

Before we’d entered a single character we recognised the potential damage this game could wreak on our productivity so we promised ourselves we’d only go at it for 15 minutes or until we got stuck. 85 minutes later, after suffering an unthinkable tragedy, we closed the tab and were left with a strange mix of sadness, regret, and mild annoyance.

The game starts by politely asking you to “Please choose a password.” Entering ‘Password#1’ immediately satisfied the first four ‘rules’ like any good password should. It’s at least 5 characters long and includes a number, an uppercase letter, and a special character. So far so good.

Rule 5 is where things started to get weird. “The digits in your password must add up to 25” is not a password rule we’ve seen before.

Things quickly descended into absurdity and before we knew it we were adding Roman numerals, including one of the game’s “sponsors”, and making sure the Roman numerals we added previously multiplied to 35.

Then it took us off-site, forcing us to first solve and then include the current Wordle answer followed by the name of the country which you’ll have to deduce based on a random Google Street View location. Luckily you’re still able to ‘walk around’ to find clues. We’ll admit we also had to look up what the current phase of the moon is before finding and adding it as an emoji.

If you’re using a Windows PC, the emoji shortcut is the Windows logo key + . (period). If you’re on an Apple PC it’s a little more complicated.

Don’t give up on your password too soon

You might think, as we did, that you’re toast when you get to Rule 16 (or thereabouts) and you have to include the next best move of a given chess board in algebraic chess notation. But we’d encourage you to press on, whether that be through luck or with the help of nextchessmove.com a, uh, friend.

Then you’ll meet Paul. He’s just an egg that hasn’t hatched at first so you’re tasked with adding him to your password to keep him safe. We were really looking forward to getting to know Paul better once he’d hatched but sadly disaster struck.

We couldn’t scramble to douse the flames fast enough so Paul’s precious life, along with our last bit of patience, perished in the ensuing blaze.

Stuff says: 10/10 – Would play again while the editor isn’t looking.

It’s no surprise that other AI companies have been rushing to release their own large language models (LLMs) – the name for the technology underlying chatbots like ChatGPT. Some of these LLMs will be incorporated into other products, such as search engines.

With its impressive capabilities in mind, I decided to test the chatbot on Wordle – the word game from the New York Times – which I have been playing for some time. Players have six goes at guessing a five-letter word. On each guess, the game indicates which letters, if any, are in the correct positions in the word.

Using the latest generation, called ChatGPT-4, I discovered that its performance on these puzzles was surprisingly poor. You might expect word games to be a piece of cake for GPT-4. LLMs are “trained” on text, meaning they are exposed to information so that they can improve at what they do. ChatGPT-4 was trained on about 500 billion words: all of Wikipedia, all public-domain books, huge volumes of scientific articles, and text from many websites.

AI chatbots could play a major role in our lives. Understanding why ChatGPT-4 struggles with Wordle provides insights into how LLMs represent and work with words – along with the limitations this brings.

First, I tested ChatGPT-4 on a Wordle puzzle where I knew the correct locations of two letters in a word. The pattern was “#E#L#”, where “#” represented the unknown letters. The answer was the word “mealy”.

Five out of ChatGPT-4’s six responses failed to match the pattern. The responses were: “beryl”, “feral”, “heral”, “merle”, “revel” and “pearl”.

With other combinations, the chatbot sometimes found valid solutions. But, overall, it was very hit and miss. In the case of a word fitting the pattern “##OS#”, it found five correct options. But when the pattern was “#R#F#”, it proposed two words without the letter F, and a word – “Traff” – that isn’t in dictionaries.

Under the bonnet

At the core of ChatGPT is a deep neural network: a complex mathematical function – or rule – that maps inputs to outputs. The inputs and outputs must be numbers. Since ChatGPT-4 works with words, these must be “translated” to numbers for the neural network to work with them.

The translation is performed by a computer program called a tokenizer, which maintains a huge list of words and letter sequences, called “tokens”. These tokens are identified by numbers. A word such as “friend” has a token ID of 6756, so a word such as “friendship” is broken down into the tokens “friend” and “ship”. These are represented as the identifiers 6756 and 6729.

When the user enters a question, the words are translated into numbers before ChatGPT-4 even starts processing the request. The deep neural network does not have access to the words as text, so it cannot really reason about the letters.

Read More: ChatGPT has been ported to (and is functional on) MS-DOS

Poem task

ChatGPT-4 is good at working with the first letters of words. I asked it to write a poem where the opening letter of each line spelled out “I love robots”. Its response was surprisingly good. Here are the first four lines:

I am a fan of gears and steel

Loving their movements, so surreal,

Over circuits, they swiftly rule

Vying for knowledge, they’re no fool,

The training data for ChatGPT-4 includes huge numbers of textbooks, which often include alphabetical indices. This could have been enough for GPT-4 to have learned associations between words and their first letters.

The tokenizer also appears to have been modified to recognise requests like this, and seems to split a phrase such as “I Love Robots” into individual tokens when users enter their request. However, ChatGPT-4 was not able to handle requests to work with the last letters of words.

ChatGPT-4 is also bad at palindromes. Asked to produce a palindrome phrase about a robot, it proposed “a robot’s sot, orba”, which does not fit the definition of a palindrome and relies on obscure words.

However, LLMs are relatively good at generating other computer programs. This is because their training data includes many websites devoted to programming. I asked ChatGPT-4 to write a program for working out the identities of missing letters in Wordle.

The initial program that ChatGPT-4 produced had a bug in it. It corrected this when I pointed it out. When I ran the program, it found 48 valid words matching the pattern “#E#L#”, including “tells”, “cells” and “hello”. When I had previously asked GPT-4 directly to propose matches for this pattern, it had only found one.

Future fixes

It might seem surprising that a large language model like ChatGPT-4 would struggle to solve simple word puzzles or formulate palindromes, since the training data includes almost every word available to it.

However, this is because all text inputs must be encoded as numbers and the process that does this doesn’t capture the structure of letters within words. Because neural networks operate purely with numbers, the requirement to encode words as numbers will not change.

There are two ways that future LLMs can overcome this. First, ChatGPT-4 knows the first letter of every word, so its training data could be augmented to include mappings of every letter position within every word in its dictionary.

The second is a more exciting and general solution. Future LLMs could generate code to solve problems like this, as I have shown. A recent paper demonstrated an idea called Toolformer, where an LLM uses external tools to carry out tasks where they normally struggle, such as arithmetic calculations.

We are in the early days of these technologies, and insights like this into current limitations can lead to even more impressive AI technologies.

• Michael G. Madden is an Established Professor of Computer Science, University of Galway
• This article first appeared on The Conversation

“Yeah, we beat Flappy Birdle“

Flappy Birdle is simple. Or rather, simple to learn. Mastery of Flappy Birdle isn’t something that comes naturally. It requires practice. It requires dedication. And, as we said, multiple attempts. Start a game as you would with any Wordle – by choosing a starting letter. Straight away, you’ll begin to see some differences to your casual game of Wordle. For one, the accompanying Flappy Bird takes off, kicking your mind into hyperdrive. To keep the Flappy Bird ‘alive’, it’ll need the next letters of your five-letter Wordle. Whether they’re wrong or not makes no difference to Flappy Birdle.

After the first five letters (spelling mistakes and repeatedly incorrect letters are allowed), the completed word will flash which letters are right and in the correct place (green), which are correct but in the wrong place (yellow), and anything left over is greyed out. With that knowledge, you can start your second word. That’s it. Keep the bird flapping until it either falls out of the sky, or you guess the correct word. But you’ll have to keep guessing, more or less as fast as you’re able to type.

Read More: Wordle has a new editor and new rules – time to up your game

You won’t succeed on your first try. You won’t succeed on your second try. Actually, you may never succeed. But there’s no harm in trying over and over again.

If you do eventually defeat Flappy Birdle and suddenly feel like you could take on anything, turn on hard mode. It makes the Flappy Bird move faster and hit the ground quicker, and turns the green pipes into more than the decorations they are in easy mode. Oh, and guesses will be limited to six. If that’s just too difficult, it may be time to head back to easy mode.

The basics haven’t changed, you’ll still get six chances to guess the five-letter word, but the word list that was previously assembled by the game creators will now be assembled by Tracy. Who is Tracy, we hear you ask? Tracy Bennett is the person who used to oversee The New York Times’ crossword puzzle, and this is the person who has become Wordle’s dedicated editor.

“Tracy Bennett will be programming Wordle every day,” notes the publication in an announcement about the change.

Read More: So it begins – The New York Times has gotten the Wordle Archive taken down

That’s not all. The word of the day will no longer be a plural of a three or four-letter word with a simple ‘s’ or ‘es’ addition at the end. You’ll get plurals, but with a twist.

“The answer will never be FOXES or SPOTS, but it might be GEESE or FUNGI. As the game is currently designed, FOXES or SPOTS can be used as a guess word to help narrow
down the answer, but FOXES or SPOTS will not be the answer,” says The New York Times.

The New York Times acquired Woodle, created by Josh Wardle, in January this year and some have complained it’s gotten harder since the acquisition. “After nearly a year of speculation, it will finally be our fault if Wordle is harder,” says the editor.

By the way, we have a Wordle challenge here at Stuff and we know some Stuff members have a few tricks under their sleeves that they use on a daily when guessing the daily five-letter word. The changes are bound to heat up the in-office competition just a little more.

Source: The New York Times, Fortune.

If you ever wanted to be Jigsaw of “Do you want to play a game?” fame, this Wordle-spinoff is definitely for you. Especially since you’re playing against AI and not a real person. It makes the activity far less creepy.

Quizmaster, not competitor

Adverswordle reverses roles by putting the AI in the player’s shoes. Rather than being put on the spot for your linguistic abilities, the player thinks of a five-letter word. Which the AI tries to figure out. If a letter is in the word but in the wrong place, the player clicks it once to mark the letter in yellow.

If a letter is in the correct position, plays click it twice to mark it green. Then, the AI guesses a new word. The player’s goal is to make the word as hard as possible, with the score based on how many guesses the AI needs to get it right. ‘Possible’ is the operative word here. Players lose when the AI guesses the correct word.

As with Wordle, Players get one attempt per day. The game AI has a new starting guess each time. Or you can be boring, and head back to safe old Wordle.

So many Wordles

Wordle has nearly 3 million players across the world, and versions of it in several languages. And then there are the fan-made spinoffs. You now get Griddle, for Formula1 fans, just in time for F1 season; Heardle, where you listen to a song and have to guess the name; Nerdle, turning it into a numbers game; Crosswordle; Taylordle; Lewdle; and even Afrikaans options – Wortel. If you’re still not satisfied, check out this Wordle alternatives list.

Source: Toms Guide

At the time, changes to the guessing game were expected. Wordle was to remain free-to-play for a time, but there was always a chance that it would require a login or a subscription to access once The New York Times figured out what to do with its new toy.

Wordle Archive goes away

The first official enforcement of the NYT‘s new authoritah has come in the form of Wordle Archive‘s shutdown. The archive used to let players replay previous days, a feature that the ‘original’ word game doesn’t include. The website makes it clear that its shutdown was at the behest of The New York Times.

A representative, speaking to Ars Technica, confirmed this. They said, “The usage was unauthorized, and we were in touch with them. We don’t plan to comment beyond that.”

It’s suggested that the various clones of the popular game might be next in the firing line. Other archives (like this one right here) might also find themselves asked ‘nicely’ to shut down operations. American trademark and copyright law in particular is a strange beastie and many of the spinoffs of the game (like Wortle) should be fine. Unless, that is, they use the word ‘Wordle’ or have a ‘dle’ suffix in their title. Then, they’re likely to face potential legal action, should the NYT decide it wants to protect this thing it just paid for.

Whether that’ll actually happen isn’t certain. “We don’t have set plans for the game’s future,” a company spokesperson said. “We’re focused on continuing to make Wordle a great daily puzzle.”

Source: Ars Technica

Be warned. Once you start playing it’s more difficult to stop than you might think. Especially when you see people posting their blocks and only getting it after five or six guesses. ‘It can’t be that hard’, you’ll murmer as you push aside whatever important thing you were busy with. The next thing you know, it’s lunchtime.

Like most things on the internet these days, it was never going to just stop at Wordle. The number of spinoff games (like South Africa’s own Wortle, which we covered here) is vast but we’re rounded up some of the best of the rest. Just in case you were looking for a little more of the good stuff. But don’t worry, we haven’t spoiled any of them… much. Click on the titles to head to the respective sites to try them out for yourselves. But only if you don’t have anything else to do today. Okay. Don’t say we didn’t warn you.

Lewdle – A rude alternative

If the original Wordle is too tame for you and you don’t mind being a little naughty, check out Lewdle. It works in exactly the same way as Wordle in every way but one. The words in Lewdle’s lists are those your parents and your pastor feared. It’s why they kept you from watching South Park until you moved out of the house. Don’t be afraid to get wild. If there is a line, we haven’t found it yet. That being said, maybe don’t play this one around small children or people that might ask questions you don’t want to answer.

Quordle – For when you want to show off

If you consistently get Wordle in two or three guesses and are looking for more of a challenge, this is what you need. Quordle has you solve four different Wordles… all at the same time. You get three extra guesses to do it and you’re going to need them. All your guesses count towards all the words so casting a wide net would do you good initially. Its creators must’ve known this one might be a little confusing at first so they were nice enough to make a practice feature. That’ll let you guess without counting towards your progress so you can see how things work and plan a strategy accordingly. Good luck, and try not to break anything when you don’t get all four.

Worldle – Wordle for geography

Worldle is our first major deviant from the established method. Instead of guessing a word of the day, you’ll need to identify a country from the outline of its borders. Then it’ll take your guess and tell you how close or far you were and in which direction you should aim. It also displays blocks that indicate how close your guess was but they aren’t as useful as the Wordle blocks. This one is a lot easier to cheat with, by just opening Google Maps and scrolling around for a bit. Where’s the fun in that?

A Greener Worldle – Because the planet is dying

Just before you start having any fun, we must remind you that the planet is dying and it’s our collective fault. ‘But what can I do? I’m just one person,’ you might ask. Of the very large list of things you can do to reduce your carbon footprint, playing A Greener Worldle probably isn’t one of them. But if the regular Wordle was just too… dirty for you, then you’ll be happy to note A Greener Wordle only uses 100% recycled words. Words that are organically sourced and that have to do with climate change and the environment. Because “…you want the tiles to turn green, just like the planet.”

Absurdle – If you thought you were good at these…

This is a strange one. It works like Wordle in that you need to guess a five-letter word correctly. Only that correct five-letter word can change midway through. The author describes it as an ‘adversarial’ version of Wordle that “is actively trying to avoid giving you the answer.” There’s even a button for a random guess. Do you see what we mean by strange? There’s a separate page for a detailed breakdown of how it works here. There’s also a Tetris derivative that’ll have you punching your screen in further. Which is… fine, if you’re into that.

Murdle – If you needed higher stakes

If you’ve ever played Hangman, Murdle should look familiar. You’re allowed to guess as many times as you’d like. But you’re only allowed to get ten unique letters wrong, or it’s the gallows for you. Ten might seem like a lot but remember, in a five-letter word if you only get one right that’s already four down. The coloured blocks that keep you on the right track work similar to Wordle. Consider your guesses carefully. That poor chap is counting on you. And so are we.

Dundle – Back to the classic, with a slight spin

Dundle is a The Office themed Wordle that’ll accept names, places, or any other five-letter word to do with the American version of the popular mockumentary-style TV show. The only other difference is that correct guesses are shown in purple and not green. Not like you need the encouragement, but this is probably the perfect excuse to rewatch The Office instead of… anything else on your Netflix list. To the great despair of basically everyone in the household. But at least it’s vaguely educational this time. Right? Right?!

Lordle of the Rings – You have my sword. And my guesses

Another twist on the classic, Lordle-OTR — as we call it — borrows pretty much everything from the original game. But it swaps the word list out for something more Tolkein-esque. You’ll need to be on the ball with your Tolkein knowledge because you’re still limited to five letters and six guesses. How you’re supposed to play with one effectively when the average letter length in an average Tolkein novel is like… eleven letters (Númenorean, Glorfindel, Fangorn) is anyone’s guess. Such a pity that Bilbo has two ‘b’s in it. This one doesn’t have a dark mode though so prepare your retinas. Apparently dark mode (which should have been called Mirkwood mode) is for Mordor supporters.