HackerTrans
TopNewTrendsCommentsPastAskShowJobs

jxmorris12

5,016 karmajoined 10 yıl önce
personal website: jxmo.io

Submissions

The Meridian of Her Greatness (2017)

samzdat.com
1 points·by jxmorris12·13 saat önce·0 comments

The One-Step Trap (In AI Research)

incompleteideas.net
53 points·by jxmorris12·dün·10 comments

A Treatise on How to Use the Internet Without Committing Philosophical Suicide

pastebin.com
3 points·by jxmorris12·5 gün önce·0 comments

Are all evils caused by insufficient knowledge?

bretthall.org
3 points·by jxmorris12·6 gün önce·1 comments

Leverage Research 1.0

lydialaurenson.substack.com
7 points·by jxmorris12·11 gün önce·0 comments

TreeSheets (Hierarchical Spreadsheet)

strlen.com
2 points·by jxmorris12·15 gün önce·0 comments

Why eval startups fail (2025)

thomasliao.com
110 points·by jxmorris12·21 gün önce·56 comments

Are You in the Weights?

intheweights.com
5 points·by jxmorris12·25 gün önce·1 comments

Cursor and SpaceX: In search of a complete loop

kwokchain.com
5 points·by jxmorris12·26 gün önce·0 comments

Zen and the Art of Machine Learning Research

blog.jxmo.io
289 points·by jxmorris12·28 gün önce·105 comments

Demystifying Noise Contrastive Estimation

jxmo.io
5 points·by jxmorris12·29 gün önce·0 comments

Joan Didion: Staking Out California (1979)

nytimes.com
1 points·by jxmorris12·geçen ay·0 comments

Gram Newton-Schulz: A Fast, Hardware-Aware Newton-Schulz Algorithm for Muon

tridao.me
30 points·by jxmorris12·geçen ay·5 comments

The Decline of Token-Level Purchasing Power

bigspin.ai
2 points·by jxmorris12·geçen ay·0 comments

A one-parameter model that gets 100% on ARC-AGI-2

eitanturok.github.io
3 points·by jxmorris12·geçen ay·0 comments

The Effective Sample Size

alex.smola.org
25 points·by jxmorris12·geçen ay·1 comments

Can you go 82-0?

82-0.com
5 points·by jxmorris12·geçen ay·0 comments

Automating Plain-Text Location Updates with Apple Shortcuts and Redis

nanjiangwill.com
2 points·by jxmorris12·geçen ay·0 comments

Bias Compounds, Variance Washes Out

convergentthinking.sh
29 points·by jxmorris12·2 ay önce·32 comments

Rothko for your current weather conditions

rothko.joonas.wtf
151 points·by jxmorris12·2 ay önce·15 comments

comments

jxmorris12
·18 gün önce·discuss
Lo and behold, a nice arithmetic coding implementation that wasn't written by an LLM! A sight for sore eyes – a treat, even. Looks like it was written by someone else though.

Check it out: https://github.com/samyak112/pym-particles/blob/main/arithme...
jxmorris12
·23 gün önce·discuss
Haha. Unfortunately is my regular voice, since long before I started using Codex. You can check through some of my old writing. It definitely could've gotten worse though. Not sure if I'm training on Codex, or Codex is training on me...
jxmorris12
·24 gün önce·discuss
This is certainly part of it! My point was that focusing on problems proposed by others is one very specific and pretty short-term mode of thinking. Good researchers improve benchmark scores. Great researchers think about what problem they're solving.
jxmorris12
·25 gün önce·discuss
Incredible concept and a very well-crafted site. I scored very low, but then very high with my legal name. It seems DeepSeek knows a lot of arxiv papers (or at least, about the authors).
jxmorris12
·29 gün önce·discuss
There’s nothing to read.

Model A: A_1, …, A_n Model B: B_1, …, B_n

C_i = A_i * p + B_i * (1 - p)

In other words, it’s just a linear combination of the other models’ weights, per position.
jxmorris12
·geçen ay·discuss
> The app does absolutely no work in the background. It works by simply existing as a running process, thanks to having the same bundle identifier as the Music app.

I love clever, low-or-no-code engineering solutions like this. You typically need to understand a systems very deeply to reach this level of elegance. In this case, one has to understand exactly what happens when the play button is pressed in Mac OS, how bundle identifiers work, etc. And the outcome is an app with almost no code at all – just a collision – it's beautiful.

(As an aside, coding agents are terrible at this kind of thing; I'd guess Codex as of right now would write some overpowered application that polls in a loop looking for Music App starts and killing them)
jxmorris12
·geçen ay·discuss
For some reason I can't get past the start page; I press all the buttons but nothing happens. Rats.
jxmorris12
·2 ay önce·discuss
Which one? San Francisco right now is "Untitled Brown and Gray"...
jxmorris12
·7 ay önce·discuss
I also wonder why this article was flagged. The article is a highly-respected researcher and professor at CMU. His thoughts are worth reading.
jxmorris12
·8 ay önce·discuss
Keep typing.
jxmorris12
·9 ay önce·discuss
It's great that people are starting to take continual learning seriously, and it seems like Jessy has been thinking about LLMs and continual learning longer than almost anyone.

I especially like this taxonomy

> I think of continual learning as two subproblems:

> Generalization: given a piece of data (user feedback, a piece of experience, etc.), what update should we do to learn the “important bits” from that data?

> Forgetting/Integration: given a piece of data, how do we integrate it with what we already know?

My personal feeling is that generalization is a data issue: given a datapoint x, what are all the examples in the distribution of things that can be inferred from x? Maybe we can solve this with synthetic datagen. And forgetting might be solvable architecturally, e.g. with Cartridges (https://arxiv.org/abs/2506.06266) or something of that nature.
jxmorris12
·11 ay önce·discuss
Matryoshka embeddings are not sparse. And SPLADE can scale to tens or hundreds of thousands of dimensions.
jxmorris12
·11 ay önce·discuss
It seems to be an error with the classifier. Sorry everyone. I probably shouldn't have posted that graph; I knew it was buggy, I just thought that the Perl part might be interesting to people.

Here's a link to the model if you want to dive deeper: https://huggingface.co/philomath-1209/programming-language-i...
jxmorris12
·11 ay önce·discuss
Hi again. I had already written about this later in my blog post (which is unrelated to this thread), but the point was that RLHF hadn't been applied to language models at scale until InstructGPT. I edited the post just now to clarify this. Thanks for the feedback!
jxmorris12
·12 ay önce·discuss
Whoops. I hope you can overlook this minor logical error.
jxmorris12
·12 ay önce·discuss
I don't think architecture matters. It seems to be more a function of the data somehow.

I once saw a LessWrong post claiming that the Platonic Representation Hypothesis doesn't hold when you only embed random noise, as opposed to natural images: http://lesswrong.com/posts/Su2pg7iwBM55yjQdt/exploring-the-p...
jxmorris12
·geçen yıl·discuss
I recently wrote a post about scaling RL that has some similar ideas:

> How to Scale RL to 10^26 FLOPs (blog.jxmo.io/p/how-to-scale-rl-to-1026-flops)

The basic premise behind both essays is that for AI to make another big jump in capabilities, we need to find new data to train on.

My proposal was reusing text from the Internet and doing RL on next-token prediction. The linked post here instead suggests doing 'replication training', which they define as "tasking AIs with duplicating existing software products, or specific features within them".
jxmorris12
·geçen yıl·discuss
My bad. Do they make this for Chrome?