Peter Barnett @csgosmorf

Joined April 2021

Tweets

462
Followers

35
Following

223
Likes

1K

Peter @Phrases1439078

5 months ago

Major update to my free app “Lexi: Vocabulary Crosswords” for expanding vocab via crosswords with spaced repetition! Now featuring: — fill-in-the-blank clues testing in-context application — pronunciation — improved definitions 50% of profits go to @QualiaRI. #Vocabulary #QRI

0 2 3 221 0

Download Image

Peter Barnett @csgosmorf

5 months ago

This app is awesome!! (that’s my alt haha)

Peter @Phrases1439078

5 months ago

This app is awesome!! (that’s my alt haha)

0 2 3 221 0

Download Image

0 0 2 44 0

AK @_akhaliq

8 months ago

Microsoft presents rStar-Math Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking On the MATH benchmark, it improves Qwen2.5-Math-7B from 58.8% to 90.0% and Phi3-mini-3.8B from 41.4% to 86.4%, surpassing o1-preview by +4.5% and +0.9%. On the USA Math Olympiad…

33 186 1K 313K 743

Download Image

Peter Barnett @csgosmorf

9 months ago

Is it just me or does o1-pro perform worse when source code files are uploaded than when they’re each copy-pasted into the chat window?

0 0 0 55 0

Peter Barnett @csgosmorf

9 months ago

Hold your horses? No. My horses will not be held. Horses, attack!!!

0 0 1 41 0

Peter Barnett @csgosmorf

10 months ago

“Error in message stream” almost every time I prompt o1, after it thinks for a couple minutes and writes almost the entire answer. Can only imagine how much compute is being wasted by this

0 0 2 119 0

Peter Barnett @csgosmorf

11 months ago

System 2 thinking was a step up. Maybe the next step up is introspection via meta-representations. It could not only know things, but also know what it knows

0 0 1 53 0

Peter Barnett @csgosmorf

12 months ago

o1 casually thinking about Elon Musk when I'm trying to get it to fix some code that's completely irrelevant xD

0 0 2 78 0

Download Image

Alex Albert @alexalbert__

2 years ago

Fun story from our internal testing on Claude 3 Opus. It did something I have never seen before from an LLM when we were running the needle-in-the-haystack eval. For background, this tests a model’s recall ability by inserting a target sentence (the "needle") into a corpus of…

561 2K 12K 3.4M 4K

Download Image

Peter Barnett @csgosmorf

2 years ago

(vision, seeing) (audition, hearing) (olfaction, smelling) (tactician, touching) (gustation, tasting)

0 0 0 61 0

Peter Barnett @csgosmorf

2 years ago

0 0 0 70 0

Download Image

Peter Barnett @csgosmorf

2 years ago

If the entirety of Earth’s surface had a tiling like this and you stepped on 2 new tiles per second, 8 hours per day, everyday, it would take you around 100 million years to step on every last tile. If u could observe 1M new tiles per second you still wouldn’t live to see it all

0 0 0 70 0

Download Image

cts🌸 @gf_256

2 years ago

"Why does everyone keep telling me to use c++?"

74 588 7K 607K 979

Download Image

Peter Barnett @csgosmorf

2 years ago

How is there still not an “edit prompt” button here on mobile?

0 0 1 74 0

Download Image

Peter Barnett @csgosmorf

2 years ago

@alexeyguzey **Explaining Induction** 1/? To grok it we need predicates. A predicate is a function that maps each input to a truth value. e.g. Let n%2 denote the remainder after dividing n by 2 (so it is 0 for evens and 1 for odds) Let isEven(n) = (n%2 = 0) Let isOdd(n) = (n%2 = 1)

1 1 1 469 1

Peter Barnett @csgosmorf

2 years ago

Incredible video youtu.be/FQ9l4v7zB3I?si…

0 0 1 56 0

Peter Barnett @csgosmorf

2 years ago

Shouldn't [x,y] = xy - yx be called the "anti-commutator", since [x,y] = -[y,x]? And since it "measures anti-commutativity"? Then {x,y} = xy + yx would be called the "commutator" since {x,y} = {y,x}, and since it measures commutativity Or am I missing something