Samuel Burbulla

[levent @__alpoge__]

hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)

[Logan Kilpatrick @OfficialLoganK]

Lots of people want Gemma 4!

Google AI Edge is #8 on the iOS App Store for productivity apps.

[Demis Hassabis @demishassabis]

It’s 2am again, my favourite time, and as always reality is still staring me in the face 🧐

[Konstantin Kisin @KonstantinKisin]

It's been an intense week and there is much anger, hatred and resentment. Some want to use this moment to make more money, blame the people they always wanted to blame and amplify hate, so it's important to say this.

To my enemies: I hate your ideas but I don't hate you. You are not, actually, my enemies. You are my fellow citizens, my brothers and my sisters.

The ideas you have been seduced by are dangerous and destructive. You have been tricked into believing that your country and your civilisation are bad. That words are violence. That words you disagree with must be suppressed and that the people who utter them deserve to be killed. That the best way to deal with anti-black racism is anti-white racism. That men can become women. That losing a war you started is genocide.

These ideas are wrong, bad and dangerous. I hate them. But I don't hate you. You have been tricked into believing them as many of us once were. Yes, it's true - some of you will continue to believe them to your dying day and, because of that, I will keep fighting you with argument, persuasion and debate to my dying day. But many of you will come to a moment of clarity as many of us have before you. It might be having a child, it might be seeing the hate you receive from your own side for daring to stray from the plantation, it might be something else that affects you personally.

And what I want you to know is that when that moment comes, you will be welcomed with open arms. Come on over.

[Demis Hassabis @demishassabis]

Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! https://t.co/pp9bXF7rVj

[Simon Willison @simonw]

It's been 2.5 years with little progress finding mitigations for prompt injection attacks LLM apps... but that may finally have changed!

Google DeepMind published a paper describing CaMeL, an ingenious system that could, maybe, lead to secure digital assistants

[Simon Willison @simonw]

Model Context Protocol has prompt injection security problems ... and it's not a problem with the protocol itself, this comes up any time you provide tools to an LLM that can potentially be exposed to untrusted inputs

[Demis Hassabis @demishassabis]

Gemini 2.5 Pro is an awesome state-of-the-art model, no.1 on LMArena by a whopping +39 ELO points, with significant improvements across the board in multimodal reasoning, coding & STEM. You can try it out now in AI Studio https://t.co/lLpF8ToTVJ & @GeminiApp with Gemini Advanced

[Yann LeCun @ylecun]

Hey Europe, you want a vibrant tech industry, right?
Do this immediately (it's genius!)
https://t.co/h48w0fTKDJ

[Samuel Burbulla @samuelburbulla]

@sama Wow, OpenAI is confident to know how to build AGI now

[François Chollet @fchollet]

One very important thing to understand about the future: the economics of AI are about to change completely. We'll soon be in a world where you can turn test-time compute into competence -- for the first time in the history of software, marginal cost will become critical.

[Samuel Burbulla @samuelburbulla]

„All intuition about AI capabilities will need to get updated for o3.“

https://t.co/hPMnh7rhyh

[François Chollet @fchollet]

Today OpenAI announced o3, its next-gen reasoning model. We've worked with OpenAI to test it on ARC-AGI, and we believe it represents a significant breakthrough in getting AI to adapt to novel tasks.

It scores 75.7% on the semi-private eval in low-compute mode (for $20 per task in compute ) and 87.5% in high-compute mode (thousands of $ per task). It's very expensive, but it's not just brute -- these capabilities are new territory and they demand serious scientific attention.

[François Chollet @fchollet]

The real problem with EU cookie banners is that they're still around after several years of achieving nothing but being an annoyance. Mistakes happen. The key is to identify them and correct them quickly. If you can't do that, you're on the wrong path.

[Samuel Burbulla @samuelburbulla]

We've just published #continuiti 0.2.0! The new version features an improved documentation page (https://t.co/JRZHon5krK), some attention features, and a surprisingly effective #neuraloperator architecture we have termed #DeepCatOperator (DCO):

pip install -U continuiti

[Yann LeCun @ylecun]

Pause AI....legislation.

[Fei-Fei Li @drfeifei]

I’m deeply concerned about California’s SB-1047, Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. While well intended, this bill will not solve what it is meant to and will deeply harm #AI academia, little tech and the open-source community. https://t.co/ycPGVuTbz0

[Delip Rao e/σ @deliprao]

companies "talking" about AI for public benefit // HuggingFace

[François Chollet @fchollet]

Many people have been recognizing that AGI won't be coming from mere scaling of current tech, and that generative AI has been severely overhyped. This is all true. But those people often conclude, "therefore AI is not going to be transformative -- this is a nothingburger". Absolutely not.

AI (both current and near-future tech) *will* transform nearly every industry, and we're still only in the very first steps of that process. Generative AI may be a bubble, but AI is going to be bigger in the long run than what almost all observers currently anticipate.

[Yann LeCun @ylecun]

Meta will *not* release the multimodal versions of its AI products and models in the EU because of an unpredictable regulatory environment.
This means that EU users of Ray-Ban Meta won't be able to use the image understanding features.
It also means that the EU industry will not have access to future multimodal versions of Llama-3.

https://t.co/sRWyEKyV3D

[Thuerey Group at TUM @thuereyGroup]

Interestingly, even for larger "foundation-model" networks there seems to be a similar scaling for accuracy over parameters: here's a comparison with the Poseidon paper (very interesting work: https://t.co/N6dkpODbYN) , compared to our experiments (from https://t.co/p2noWrgOJl):