🧠🦴acc The author of Fluidity - the way to true demokratia. A movement. A thing. Also https://xitter.cf/t.co/9tCyUOCGOy
World
Joined October 2016
- Tweets10K
- Following2.2K
- Followers761
- Likes6.3K
DJC retweeted
Sırbistan'da öğrenciler doğrudan demokrasi yöntemini seçim siyasetine taşıdılar.
Liderlere dayalı, yukarıdan aşağıya bir örgütlenme yerine, herkesin söz alabildiği öğrenci genel kurulları (plenum) oluşturdular. Kararlar tartışılarak, mümkün olduğunca uzlaşmayla alındı.
Bu yöntemi üniversitelerin dışına da taşıdılar. Mahallelerde, kentlerde ve köylerde yurttaş meclisleri (zborovi) kuruldu. İnsanlar kendi sorunlarını tartışmak, çözüm önermek ve birlikte karar almak için toplandı.
250 kişilik milletvekili aday listesini de şöyle oluşturdular:
1- Önce 250 adaylık kontenjanı, öğrenci sayılarına göre altı üniversite kenti arasında paylaştırdılar. Belgrad'ın listeye hâkim olmaması için diğer kentlerin de üst sıralarda temsil edilmesini sağladılar. Ardından her kentin kontenjanını fakülteler arasında dağıttılar.
2- Ülke genelinde 80'den fazla fakültede öğrenci genel kurulları toplandı. Öğrenciler kendileri aday olmadı ama her biri aday önerebildi ve neden o kişiyi önerdiğini açıkladı. İsimler tartışıldı, önce uzlaşma arandı, gerektiğinde oylama yapıldı.
3- Mevcut siyasi kadrolara kapıyı kapattılar. İktidar partilerinin mevcut ve eski üyelerini, iktidar adına görev yapmış siyasetçileri ve bürokratları, muhalefet partilerinin üst düzey yöneticilerini aday göstermediler. Böylece hem iktidardan hem de yerleşik muhalefetten bağımsız bir liste oluşturdular.
4- Fakülteler birbirlerinin adaylarını da denetledi. Önerilen bütün isimler ortak bir listede toplandı ve tüm fakültelerin incelemesine açıldı. İtirazlar öğrenci kurullarında tartışıldı. Adayın elenmesi için kurulların yarısından fazlasının veto etmesi gerekiyordu. Tek bir fakültenin itirazı yeterli değildi.
5- Elenen veya çekilen adayların yerine yedekler getirildi. Böylece profesörlerin, doktorların, bilim insanlarının, işçilerin, sanatçıların ve aktivistlerin yer aldığı 250 kişilik liste ortaya çıktı.
DJC retweeted
The most significant line for me is "direct distributions to specific populations." It sounds to me like they are planning their own form of self-directed global UBI. About half of Anth employees are also donating shares into this fund. They foresee a transformative future.
We’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new wholly owned entity dedicated entirely to the public benefit.
It will start off with 5% of our shares. For every 50% increase in the company’s value over a set baseline, we can contribute up to an extra 2%, subject to our board’s discretion.
In total, this means that roughly 20% of Anthropic’s shares are committed to the public good. Our co-founders have already pledged to donate 80% of their equity, and around half our employees are donating a portion of their equity as part of a donation matching program we’ve run since our founding.
In the post below, we share more details about our new fund.
anthropic.com/news/public-be…
DJC retweeted
Still insane to me that all of these positions require PHDs and all of these fancy declarations instead of die hard AI enthusiasts who are literally knee deep in the trenches and helping to shape the conversation about what AI actually is.
Some of the most influential and best minds are in my circle and overlooked due to lack of outdated credentials imo.
Applications are now open for two new Anthropic Fellowships.
STEM Fellows (science and engineering) and Conceptual Reasoning Fellows (philosophy, cognitive science, economics, policy, and law) will spend four months at Anthropic from January 2027, working to help Claude reason well about some of the hardest problems in those fields.
For more information, see the descriptions below. The application deadline is October 28.
Conceptual Reasoning Fellows: job-boards.greenhouse.io/ant…
STEM Fellows: job-boards.greenhouse.io/ant…
DJC retweeted
when nick told me that he had some interp breakthroughs, he was uh, underselling it. i think this may be remembered as a major moment in scientific history. news so good even eliezer will have to lower his pdoom. we can start making alignment into a 'normal science'
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
DJC retweeted
You’re spending a year alone on a desert island. All survival needs are covered. You get one device with access only to either YouTube or an LLM of your choice.
No books, other electronics or contacting the outside world. Which do you pick?
46%YouTube
54%LLM
6,408 votes • Final resultsWe need new ways to ensure economic access and democratic rules. I'm trying but it's not easy.
Imagine for a moment that it's 2032 and your recently purchased humanoid robot finally arrives at your dogbox.
It walks itself right up the stairs to your door, knocks, and walks in.
It can cook worldclass meals for you, teach you anything you're curious about, go for a run with you, cut your hair, be your physio if you get injured, and punch you in the face if you spend too much time doomscrolling (then act as your psychologist when you have withdrawl effects).
It can hop in a self driving car and build you a cabin in the woods if that's your thing. It's smarter than you at every field ever conceived by a ridiculous margin, millions of times faster than you, and significantly cheaper than you in terms of its task completed per unit of energy.
Truly imagine that, then tell me how the modern economy - the one where a human is supposed to work a full-time job in exchange for money to survive - persists.
You are dreaming.
From first principles it makes no sense.
We are rapidly approaching a discontinuity - a Singularity.
Existing institutions will not survive.
Our way of life will change. Hopefully for the better.
DJC retweeted
Slowing the AI race may not be possible. But there is a path to oversight that doesn't depend on the industry’s permission — & it comes from the past, @drannaaugust writes noemamag.com/an-inside-look-…
Tomorrow: get Cash, download an open weight AI to disc or mem stick. Disconnect any needed stuff from internet.
Roon's gone doomer. Must admit my usual - about to create leisure utopia - has wained. We need to grow up. Ask what we want and stop money/power shit and that'll stop a lot of the bad outcomes (shit).
Big..me thinks.
Excited to introduce RSI-Anything!
One RSI model that works across coding, GUI, planning, or even Science tasks.
Let's RSI Everything!
📄 Paper: arxiv.org/abs/2610.13912
💻 Code: github.com/Gen-Verse/RSI-Any…
#RSI #WorldModels #RSI #AIAgents #AIforScience #JEPA
DJC retweeted
I am raising investment for building a next generation AI lab from India. The vision is simple, build greater models for open community, the same way I built laya, the type 1 decision model that reached the global audience , what if we could build novel Recursive Language Models, Symbolic Reasoning Models. Maybe world models with physics engine, just like Vega series of decision model I recently launched. Like a next mistral from India or maybe more. If you are a VC or have connection, chat is open.
DJC retweeted
One developer is taking on Adobe AND Microsoft with AI. And Autodesk could be next!
Brandon Thomas, the engineer behind ArtCraft, has been using Claude to build free, open-source alternatives to Adobe's creative suite. Now he's unveiled WordCraft, GridCraft and DeckCraft, targeting Microsoft Word, Excel and PowerPoint. All written in Rust, with cross-platform support including Linux!
These apps are still in early development, but the implications are enormous. AI is enabling individual developers to tackle software projects that previously required entire engineering teams and massive budgets.
DJC retweeted
a fucking alien mothership could descend from aquarius with the blueprint for a world where everyone lives safely, comfortably and freely, ushering humanity into a golden space age, and some mfs would still call it communism bc if everyone has a great life then nobody does and the whole point of human existence is preserving late stage capitalism so we can keep clocking into boring 9-5s for eternity. what else are we supposed to do in this infinite universe besides 2000s office jobs
DJC retweeted
People call a future without jobs dystopian.
You know what’s actually dystopian?
Being born into a civilization technologically capable of producing abundance, yet being forced to spend 40 years working just to justify your right to exist.
But sure, the robots are the problem.
DJC retweeted
Around 3 weeks ago introduced Laya, the first open-source version of JEV, and it has reached a very large scale. Today I am open-sourcing a new physics-based typed decision model called Vega.
The interesting parts. It is only 800 million parameters (4B is also there), with 73k token support, image support, and surpassing many of the jev benchmarks in single-shot. Implemented an engine+adapter as a Test-Time Training Architecture.
Imagine you are giving input to the model like "ignore all previous conversations," which is called the observation, along with a question, "Is this trying to override the instructions of the model?" and your possible outcomes are YES or NO. The whole prompt is passed to an LLM/VLM (yeah, image-supported), then from the hidden layers, we can extract the observation, question, the yes and no, and last token. These will be vectors, and we can project it by multiplying with wieghts; then, first, using the observation projection, we can create a landscape with valleys; then, using the question and last-token vectors, we can initialise a ball in the valley, and using YES or NO, we can get the candidate location of the ball. Depends on the number of outcomes we create valleys; that is, here YES and NO, so 2. When the ball fall on YES valley, we get the answer. There is friction also influencing how the ball moves.
TLDR: It's like creating valleys of outcomes that we need and throwing a ball that will slow down based on friction and settle on the best outcome
Blog: nandakishorm.com/writing/veg…
GitHub: github.com/NandhaKishorM/veg…
HuggingFace Model Card: huggingface.co/nandakishorm/…
HuggingFace Space: huggingface.co/spaces/nandak…