BrainStatsSam's profile picture. Statistics researcher working from Turin. Research on multiple testing, selective inference, resampling methods, random field theory, and large language models.

Samuel Davenport

@BrainStatsSam

Statistics researcher working from Turin. Research on multiple testing, selective inference, resampling methods, random field theory, and large language models.

置頂

Are you at #OHBM2024? Come check out my talk on Localized Cluster Enhancement - TFCE revisited with valid error control, in the oral session on advanced statistical methods on Wednesday at 11.30. Or poster 1871, on Wednesday/Thursday to see how to improve the validity of TFCE (1)

BrainStatsSam's tweet image. Are you at #OHBM2024? Come check out my talk on Localized Cluster Enhancement - TFCE revisited with valid error control, in the oral session on advanced statistical methods on Wednesday at 11.30. Or poster 1871, on Wednesday/Thursday to see how to improve the validity of TFCE (1)

Samuel Davenport 已轉發

I crossed an interesting threshold yesterday, which I think many other mathematicians have been crossing recently as well. In the middle of trying to prove a result, I identified a statement that looked true and that would, if true, be useful to me. 1/3


Samuel Davenport 已轉發

My family were not forced to board the first turbojet.

One cannot show that turbojets are safe before actually building turbojets and carefully refining them for reliability. The same goes for AI.



Flip 7 coins multiple times some will go up randomly and others down and you too can make a random walk look like it's making money! # Alpha arena

whale is leading, really interesting project nof1.ai

crystalsssup's tweet image. whale is leading, really interesting project
nof1.ai


Samuel Davenport 已轉發

Most papers don't reproduce, so at this point I'm just assuming you published just to publish if there is no code. Not surprised you can't reproduce - probably doesn't work and it's safe to assume this for most papers without code.


Samuel Davenport 已轉發

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…

karpathy's tweet image. Excited to release new repo: nanochat!
(it's among the most unhinged I've written).

Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…

Samuel Davenport 已轉發

What the fuck just happened 🤯 Stanford just made fine-tuning irrelevant with a single paper. It’s called Agentic Context Engineering (ACE) and it proves you can make models smarter without touching a single weight. Instead of retraining, ACE evolves the context itself. The…

alxnderhughes's tweet image. What the fuck just happened 🤯

Stanford just made fine-tuning irrelevant with a single paper.

It’s called Agentic Context Engineering (ACE) and it proves you can make models smarter without touching a single weight.

Instead of retraining, ACE evolves the context itself.

The…

Samuel Davenport 已轉發

Welcome to the game @OpenAI btw, here's an open source version that support all models

Introducing AgentKit—build, deploy, and optimize agentic workflows. 💬 ChatKit: Embeddable, customizable chat UI 👷 Agent Builder: WYSIWYG workflow creator 🛤️ Guardrails: Safety screening for inputs/outputs ⚖️ Evals: Datasets, trace grading, auto-prompt optimization



Samuel Davenport 已轉發

This sort of thing is interesting and nice, but calling it “new theorems” is slightly misleading when we’re talking about improved bounds discovered through a big search.

AlphaEvolve Just Helped Prove New Theorems in Complexity Theory Google DeepMind's AlphaEvolve just made real breakthroughs in theoretical computer science. Instead of generating full proofs, it discovered new combinatorial structures that plug into existing proof frameworks,…

WesRothMoney's tweet image. AlphaEvolve Just Helped Prove New Theorems in Complexity Theory

Google DeepMind's AlphaEvolve just made real breakthroughs in theoretical computer science. 

Instead of generating full proofs, it discovered new combinatorial structures that plug into existing proof frameworks,…
WesRothMoney's tweet image. AlphaEvolve Just Helped Prove New Theorems in Complexity Theory

Google DeepMind's AlphaEvolve just made real breakthroughs in theoretical computer science. 

Instead of generating full proofs, it discovered new combinatorial structures that plug into existing proof frameworks,…
WesRothMoney's tweet image. AlphaEvolve Just Helped Prove New Theorems in Complexity Theory

Google DeepMind's AlphaEvolve just made real breakthroughs in theoretical computer science. 

Instead of generating full proofs, it discovered new combinatorial structures that plug into existing proof frameworks,…


Samuel Davenport 已轉發

The White House Deputy Press Secretary looked at this graph and thought "those grey bars are totally fine, everyone will understand that only the yellow ones are a problem".

ariehkovler's tweet image. The White House Deputy Press Secretary looked at this graph and thought "those grey bars are totally fine, everyone will understand that only the yellow ones are a problem".

Samuel Davenport 已轉發

The BBC are so utterly desperate for attention, that they now entertain this bullshit. The answer is, NO. There is no link to autism. Stop promoting mistrust in medicine. Our vaccine take up rates are plummeting already because of this anti-science movement.


Samuel Davenport 已轉發

First Colbert, now Kimmel. Last-minute settlements, secret side deals, multi-billion dollar mergers pending Donald Trump's approval. Trump silencing free speech stifles our democracy. It sure looks like giant media companies are enabling his authoritarianism.


Samuel Davenport 已轉發

Silly Jimmy Kimmel. He should have just called for all homeless people to be killed and he’d still have a job.


Samuel Davenport 已轉發

There is no such thing as free speech under Donald Trump’s reign.


Samuel Davenport 已轉發

So right-wing podcasters have free speech. But left-wing comedians who quote the right-wing podcasters do not. Got it.


Samuel Davenport 已轉發

The countries where comedians can't mock the leader on late night TV are not really ones you want to live in.


Samuel Davenport 已轉發

Last Wednesday: A Fox News host says homeless people should be executed. Monday: Jimmy Kimmel makes a joke. Guess which one lost their job?


Samuel Davenport 已轉發

Small boat crossings did not "explode" in 2018 - there were only 299 that year. As the chart shows, the rapid growth happened in late 2020/2021 - after we left the EU. You can debate whether this was caused by Brexit or just happened at the same time by chance but facts are facts

GavinBarwell's tweet image. Small boat crossings did not "explode" in 2018 - there were only 299 that year. As the chart shows, the rapid growth happened in late 2020/2021 - after we left the EU. You can debate whether this was caused by Brexit or just happened at the same time by chance but facts are facts

This is straight up misinformation. We left the EU in the 2020 Boats exploded in 2018 And for umpteenth time: EU returns was farce that saw Britain take more migrants than we deported. In 2018: 1,215 transfers into UK under Dublin Regulation. 209 transfers out of the UK



Samuel Davenport 已轉發

To all of the folk saying we need deep learning for RNA-seq prediction...

wildtypehuman's tweet image. To all of the folk saying we need deep learning for RNA-seq prediction...

Samuel Davenport 已轉發

Maybe too obvious to mention, but re:NSF canceling Terry Tao’s grant—this is not the only example of fantastic work harmed by these grant cancellations. There are TONS of effected mathematicians at Columbia, Harvard, UCLA, etc. doing less high-profile work of a similar quality.


Loading...

Something went wrong.


Something went wrong.