Sabitlenmiş Tweet
Astraia
5.6K posts
Astraia
@astraiaml
AI Researcher • Friendly 🤗 • Tweets are not Financial Advise • Thank you for visiting my profile 💜
Europe Katılım Nisan 2023
679 Takip Edilen62K Takipçiler
@AndrewCurran_ China would probably refuse to cooperate
What do you think?
English
Today's signing event and announcement of the AI Executive Order has been postponed to an unannounced later date for unknown reasons. Reporting by Axios.
English
@arcticinstincts This was always going to be how the job market ends up
DEI is good and all, but if you want actual progress you need nerdy mathtards. Microsoft also did it in the 80s, it's the only real way to get a genuine business advantage.
English
@DigitalEU @EurobarometerEU This is ridiculous. No, I do not support regulating AI in any way shape or form. Enough with this BS.
English
Europeans want safe AI.
80% of EU citizens say that AI must be regulated to ensure safety.
Their main concerns?
🔹Privacy
🔹Misinformation
🔹Ethical issues & misuse of AI
Learn more: link.europa.eu/pQ7Cp4
@EurobarometerEU
English
@TheAhmadOsman You are right. The only genuine AI safety can be through its democratization.
English
Most AI researchers are European and Asian. The Americans are driving capital, not sciences.
Even so, if we are being honest, betting on Europe's AI industry itself is a fool's errand.
Only France has shown seriousness on the subject, and we are almost sure to miss this opportunity for tech sovereignty.
A real, real pity.
English
@damian_89_ @JagersbergKnut Qwen 3.6 27B 🤗
It’s really good, much better than its size would suggest
I also have 2x 3090s and moved away from Qwen-Coder-Next (100B MoE) in favor of Qwen 3.6 27B
English
@astraiaintel @JagersbergKnut Any recommendations with 200k context, solid intelligence that would work on 2x4090 24GB + 256 GB RAM?
English
I often think about professors coming up with these claims.
OpenAI@OpenAI
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
English
OpenAI made history today.
An internal reasoning model autonomously disproved a famous conjecture in mathematics that stood for nearly 80 years.
The problem: In 1946, Paul Erdős asked how many pairs of points can be exactly 1 unit apart if you place n points on a flat surface. The best known answer came from square grid constructions, and Erdős himself conjectured you can't do meaningfully better. Mathematicians believed this for decades.
The AI proved him wrong. It found entirely new point configurations that beat the square grid by a fixed polynomial factor, not a marginal improvement, a real mathematical gap.
The proof uses methods from algebraic number theory, a completely different branch of math, Class field towers, Golod-Shafarevich theory, tools nobody expected to be relevant to a geometry problem about distances in the plane (reminds me of move 37, AlphaGo tbh).
Fields Medalist Tim Gowers calls it "a milestone in AI mathematics." The proof was verified by leading external mathematicians.
According to OpenAI, this is the first time AI has independently solved a prominent open research problem in mathematics!
Caveat: Obviously OpenAI chose which problems to test the model on. So "autonomous" means the model generated the idea and wrote the proof, not that it wandered into the problem on its own.
But if reasoning models can reliably make cross-domain connections like this, finding paths that experts didn't prioritize, this changes research far beyond math. Biology, physics, materials science, medicine.
This isn't AI reproducing human knowledge anymore. This is AI producing new knowledge. That's a qualitative shift.
OpenAI@OpenAI
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
English
@AndrewCurran_ @teortaxesTex Google is lagging so far behind on the software side though
Great models, not great user interface and tools integration
English
Just call it Gemini 4 Pro at this point
Logan Kilpatrick@OfficialLoganK
Gemini 3.5 feels like the start of a new era for Gemini, we spent the last 2.5 years putting the infrastructure, products, team, etc in place (learning lots of lessons along the way). The model is the product, please keep the feedback coming!
English
@karlprosser 8 bits about 24 GB
4 bits roughly half that
Excluding KV cache
English
@kimmonismus Namely, this AGI definition paradox started by them in the earlier days, even before ChatGPT
Other companies simply played along because they had to
English
@kimmonismus Well, they use AGI to refer to the early stages of ASI
You are right to raise that, because the entire thing is unscientific marketing on the part of OpenAI
English
As excited as I am that a takeoff seems to have begun and we are entering a golden age of science, one thing remains:
I'm hearing more and more from all sides that AGI is within reach. This applies to Google (AGI, Physical AGI), as well as OpenAI and Anthropic.
The only caveat: there's no unified definition of what AGI actually is. There have been attempts to standardize a definition, and in my opinion, the most sensible one is Google DeepMind's. But as long as we're talking about different things, it's difficult to find common ground to say *when* AGI will be achieved (which AGI).
Sam Altman@sama
three of the things we are most excited about: 1. AGI accelerating research 2. AGI accelerating companies 3. personal AGI accelerating everyone in achieving their goals today it was great to announce the unit distance result. yesterday it was great to announce that we are offering to invest $2M in openai credits into every YC company. now we need to increase our efforts on the third!
English
@peteskomoroch I'm sorry, but I have to disagree even if the underlying intent is noble.
Yes, humans do predict the next context-piece, which simply in AI happens to be the next token.
x.com/astraiaintel/s…
Astraia@astraiaml
False. This is something most people, even many researchers, do not understand: The human mind actually DOES work in a very similar way to LLMs. It is called predictive coding, and it is the central theme of all human intelligence functions. The autoregressive nature of LLMs captures and solves for what is, in effect, a more limited representation of the very same contextual semanticity. The meaning captured in the input sequence is intended to represent context, the solution of which is the true mathematical target of backpropagation.
English
"It's not intelligent, it's just predicting the next word"
Have the people saying this tried consistently predicting the next word themselves?
English
False.
This is something most people, even many researchers, do not understand: The human mind actually DOES work in a very similar way to LLMs.
It is called predictive coding, and it is the central theme of all human intelligence functions.
The autoregressive nature of LLMs captures and solves for what is, in effect, a more limited representation of the very same contextual semanticity.
The meaning captured in the input sequence is intended to represent context, the solution of which is the true mathematical target of backpropagation.
Pete Skomoroch@peteskomoroch
"It's not intelligent, it's just predicting the next word" Have the people saying this tried consistently predicting the next word themselves?
English
I first used ChatGPT on Dec. 10th 2022
it's crazy how far we have come since then
English