@FaustoGoethe12 @_Sebas_Blanco @AmarilliTK Sabes leer? Los loros perciben las caricias en el lomo y bajo las alas como estimulación sexual, que no se vea igual que en mamíferos no significa que para ellos no lo sea
Español
SebaM
164 posts
Chicos no podéis tener un animal sin haberos informado un mínimo sobre él hasta el punto de estar literalmente haciéndole una paja sin daros cuenta, eh?
Birds Colour 🕊️@birdscolour56
Right angle parakeet
Español
@HSVSphere @spicey_lemonade FFmpeg thanked Anthropic for sending patches, so it’s not just marketing.
Benchmarks are not direct proof, but saying they signify nothing is just wrong, especially when there are already real vulnerability reports, bug fixes in multiple Python libs, and real patches landing
English
@SebaMG2 @spicey_lemonade These numbers do not signify anything, marketingslop. I'll see it when we get the next public model, and I'll judge the code it'll generate.
English
I've said a while ago that LLMs will be far more useful for reverse engineering & hacking than code and I was correct. These "predictions" are wrong, as the code they produce is still garbage.
The fact that this video always displayed "coding" capability higher than "hacking" was a major tell the creators didn't have a clue what they were predicting. Also, Bioweapons? Politics? lol
LLMs are and have been great for searching over tons of code for a while, and the remarkable part is how fast they do it as they have some knowledge about most public code. Finding vulnerabilities is the same, it's just reading code over and over and matching existing vulnerable patterns to code that exists in the wild.
I'm thankful that these tools exist, but they should be treated more like linters & grep tools, and they should never be allowed to edit code freely.
billy@billyhumblebrag
Haha those doofuses at ai2027 predicted we'd have professional level hacking abilities and the top ai company would be at $26B in revenue in May 2026. It's April and we already have superhuman hacking and $30B in revenue, why would you take forecasters this bad seriously???
English
@HSVSphere @spicey_lemonade I’m pretty sure they’re talking about Claude Mythos Preview, which still hasn’t been released publicly
English
@spicey_lemonade No, they don't. I actively use the latest models, and you are just lying to yourself or suck at programming.
English
@config_Geist @Michael_Druggan And reasoning is not irrelevant here. Those kinds of newer mechanisms are exactly what can catch superficially intuitive but wrong answers. So a result on older/smaller models is not automatically a result about “LLMs” in general
English
@config_Geist @Michael_Druggan A 5yo and an adult are still instances of the same cognitive system. Small older LLMs and modern frontier models are not just the same mind, but bigger. They differ in scale, training regime, post-training, inference-time behavior, etc
English
Humans are architecturally incapable of doing calculus.
I know this because I just asked a bunch of 5 year olds to solve some integrals and they all got them wrong.
can@marmaduke091
>AI can't do math >Look inside >Strongest model they tested was GPT-4o-mini Get this slop out of here. The models tested are so inferior to current SOTA models that the findings are completely irrelevant to the current AI landscape.
English
@config_Geist @Michael_Druggan You can't extrapolate those results to LLMs as a whole, the models they tested are really old, small and don't even have reasoning
English
@Michael_Druggan This is actually a decent analogy bc if you read the paper, you'll see that they were not asking complicated questions. They were testing for *extremely* basic intuitions about quantity. Intuitions that 5 year olds absolutely have. And they found LLMs don't have it.
English
No surprise here. I and others have been saying for a long time that, deep down, every LLM is dumb as a rock. Yes, cheating can make a machine sound smart but it's still dumb as a rock.
If you know how LLMs work, this truth is obvious. No one needs to read a paper to know that LLMs are massive cheating machines with zero intelligence. I, for one, am not amused. 😠
Sukh Sroay@sukh_saroy
🚨Everyone thinks GPT can do math they're wrong new paper called SenseMath just proved LLMs don't have number sense at all this changes everything about how we should use them:
English
@sheikki @sukh_saroy You're delusional, the newer models are pretty good at math. This paper used ancient, small non-reasoning models for the benchmarks
English
@sukh_saroy They don't understand words either. They understand nothing. They just output strings based on probabilities. These CEOs with their "oh we think it might be conscious" etc bullshit should be held accountable
English
🚨Everyone thinks GPT can do math
they're wrong
new paper called SenseMath just proved LLMs don't have number sense at all
this changes everything about how we should use them:
English
Why are we doing this to ourselves? What happened to all those tech companies helping the fight against climate change? This is so pointless!
CNN@CNN
The vast data centers that power artificial intelligence guzzle huge amounts of energy but they also have another alarming impact, according to new research. They are creating “heat islands,” warming the land around them by up to 16 degrees Fahrenheit, and making life hotter for more than 340 million people. cnn.it/4rZSiG5
English
English
@corruptario The things that they don’t tell you about until somebody does a deep dive!
English
Burst that bubble. Stat.
CNN@CNN
The vast data centers that power artificial intelligence guzzle huge amounts of energy but they also have another alarming impact, according to new research. They are creating “heat islands,” warming the land around them by up to 16 degrees Fahrenheit, and making life hotter for more than 340 million people. cnn.it/4rZSiG5
English
We have been gifted the most beautiful planet, full of trees, water, fruits, vegetables, and animals and all our species does is destroy so that a handful of sociopathic leeches can get a quick cash grab.
CNN@CNN
The vast data centers that power artificial intelligence guzzle huge amounts of energy but they also have another alarming impact, according to new research. They are creating “heat islands,” warming the land around them by up to 16 degrees Fahrenheit, and making life hotter for more than 340 million people. cnn.it/4rZSiG5
English
@Michael_Druggan @MN_Vikings_Pete Go ahead prove me wrong. One single example
English
> "Solving a Millennium Prize problem de novo w a genuinely novel technique (not latent in the corpus) would constitute such evidence; it has not occurred."
The 'AI is unimpressive' crowd has gotten nervous enough about AI progress that they are now adding qualifiers about what types of solutions to a Millennium problem they might accept as evidence of AI impressiveness. The writing is on the wall.
Jen Zhu@jenzhuscott
Agree w Terence Tao - LLMs limitations are structural. I’ve always said the usefulness of current AI correlates w the users expertise. So the illusion of creativity can impress/fool non experts. The current LLMs excel at Keplerian work (empirically testing many combinations via brute/compute scaling) but not Newtonian unification or genuine leaps. They act as a “super-assistant” for literature search, candidate generation, formalization, and exposition - freeing us for the creative core - but there is no evidence yet of autonomous originality at the frontier. Solving a Millennium Prize problem de novo w a genuinely novel technique (not latent in the corpus) would constitute such evidence; it has not occurred.
English
QME
@MN_Vikings_Pete @Michael_Druggan That's not the only Erdos problem solved, there's some with no prior solutions found
English
@TvDannyZ Un trato de 1 billón de dólares no es tan grande para una empresa de IA como crees. El equipo de Sora está metiendo recursos en world-models, otra tecnología que puede crear simulaciones y videos, si cerraron la app es porque no tiene sentido perder cómputo en ofrecer una demo
Español
Esto es muy grande.
No solo es que las empresas de IA necesitan flujo constante de inversión para vivir.
Disney echándose para atrás en un trato tan grande y de una forma tan pública envia un mensaje: está tecnología no vale la pena.
DiscussingFilm@DiscussingFilm
Disney has cancelled their deal with OpenAI. They planned to invest $1 billion into the company and let Sora have access to AI generate videos of characters from Disney, Marvel, Pixar and Star Wars. (Source: hollywoodreporter.com/business/digit…)
Español
@haterdabber @PixelPoncho_ @bigTfromthadmv @Radrappy The Sora team will be focusing on world-models, that can be used to make general models better AND make video simulations
English
@PixelPoncho_ @bigTfromthadmv @Radrappy I don’t think that’ll happen. The big AI labs are laser-focused on models that write code since that’s where the real money is. They won’t waste money on a product that nobody’s using.
English
“What you made with Sora mattered, and we know this news is disappointing. We’ll share more soon, including timelines for the app and API and details on preserving your work.”
Here's where you're wrong, Sam - none of it mattered. It was slop and it will be forgotten immediately
The Hollywood Reporter@THR
Breaking: Disney Exits OpenAI Deal After AI Giant Shutters Sora hollywoodreporter.com/business/digit…
English
@CrispyChickenmm Lo cerraron para centrar los recursos en world-models, otra tecnología que puede ser usada para generar vídeos. Seguramente va a terminar volviendo como parte de la app de chatgpt cuando tengan suficiente cómputo
Español
imposible, si un tipo en twitter hace 3 años me dijo que esto era el futuro, y que me tenía que adaptar
DiscussingFilm@DiscussingFilm
OpenAI is shutting down its AI video slop-making platform Sora.
Español
@Toadsanime They're just allocating resources on more important stuff than video generation demos x.com/i/status/20365…
Chubby♨️@kimmonismus
OpenAI's Sora team is now working on world-models - they prioritize longer-term world simulation research especially as it pertains to robotics. tl;dr what we know so far: - Sora has been cancelled because they needed the compute for their new LLM - they renamed product organization to "AGI Deployment" - the LLM (codename Spud) is "very very strong" and "accelerates the economy" - release in a few weeks - Sam is going to focus on "raising capital, supply chains and “building datacenters at unprecedented scale” my take: To me, it really sounds like they are preparing for the IPO and will make AGI official beforehand.
English
This is so sudden, and OpenAI so huge, that shutting down their prolific video generating app short-notice screams to me that a big, big, big lawsuit happened / is happening that they feel they must avoid at all costs.
Sora@soraofficialapp
We’re saying goodbye to the Sora app. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing. We’ll share more soon, including timelines for the app and API and details on preserving your work. – The Sora Team
English
@MC_chibill @pawdiefur_fury @nicdunz @BadMuthaHubbard Google translate has always been AI, the base architecture used in ChatGPT was originally meant for translation
English
@pawdiefur_fury @nicdunz @BadMuthaHubbard Only person not using their brain is the one that assumes AI is ised in everything when its not. So much shit you idiots call AI doesnt even use AI at all.
English
this makes no sense because intelligent people are using ai and unintelligent people are complaining about it and that implies the ones with brains are the ones that use ai and the ones without brains are the ones that dont
Michael@TheMG3D
English