Charles Feinn

learning & occasionally shouting thoughts into the void

[AK @_akhaliq]

AudioX

Diffusion Transformer for Anything-to-Audio Generation

[Charles Feinn @dynamics]

@simonw Happy to contribute these features to your tool if you're interested. I've found them really useful for providing context to Claude.

[Charles Feinn @dynamics]

@simonw Some things that I've been having good results with: 1) including a tree diagram of the project and 2) including database schema when relevant. https://t.co/OHLvVI2JoL

[Charles Feinn @dynamics]

The thought that math is akin to a game based on binary, reproducible 'legos' which describe increasingly more complexity rings true to me. It's our imagination trying to make more and more sense about the messy universe around us.

[David Bessis @davidbessis]

The post is here: https://t.co/MZbU5EcZEL
Please share and subscribe if you like it🥰!

[Charles Feinn @dynamics]

@prlz77 @Apple Are any of these roles open to non-PhD candidates? I have an MS in AI/ML + extensive SWE experience and would love to learn more.

[Nathan Lambert @natolambert]

Everyone I know that goes deep into language model evaluations comes back😵‍💫😵‍💫😵‍💫 Here's an overview of why
1. Why its hard to trust big lab evals
2. Why apples to apples comparisons of even open weight models is so hard
3. Synthetic data contamination
4. The market for harder evals

[François Chollet @fchollet]

A common misconception about Transformers is to believe that they're a sequence-processing architecture. They're not.

They're a *set-processing* architecture. Transformers are 100% order-agnostic (which was the big innovation compared to RNNs, back in late 2016 -- you compute the full matrix of pairwise token interactions instead of processing one token at a time).

The way you add order awareness in a Transformer is at the *feature* level. You literally add to your token embeddings a position embedding / encoding that corresponds to its place in a sequence. The architecture itself just treats the input tokens as a set.

There's nothing that prevents you from applying the same trick to 2D grid of tokens or a 3D cube of tokens instead! Instead of adding an embedding for (i * N + j), the position of (i, j) in a flattened grid, just embed i and j and concat the result.

Now you have a 2D position embedding that makes your Transformer natively process grids of tokens! This can be useful for ARC-AGI, since ARC-AGI grids are of course 2D grids of discrete tokens, not sequences.

[Charles Feinn @dynamics]

@yaroslavvb What does this suggest one would see if you did train a deep learning model to its condition number? The loss would go to near zero meaning a complete overfit of its data? Or different function and therefore different result?

[Charles Feinn @dynamics]

New learnings!

Who knew evaluations would be so complicated?

https://t.co/5sf3Blz7aa

[Andrej Karpathy @karpathy]

I love calculator
https://t.co/mDVX1AiBiI

A short post on philosophy of product and technology. What is beauty in technology and how can we get more aesthetically pleasing products that spark joy?

[MMitchell @mmitchell_ai]

"the most important consequence of these longer processing times is...branding. “Reasoning” models with “chains of thought” that need “more time”..OpenAI is communicating, plainly and forcefully..software that more closely approximates our minds..Anthropic has described its leading model, Claude, as having “character” and a “mind”; Google touts its AI’s “reasoning” capabilities..."
https://t.co/KTwRpHhcYd

[Jeremy Howard @jeremyphoward]

It's easy to write a LOT of words and a LOT of code with AI.

But that's unfair to the person who has to read it.

Be concise.

[Charles Feinn @dynamics]

Macro being the average across subjects whereas micro is the aggregate total across subjects.

I learned this while investigating Meta's reported Llama 3 MMLU scores: https://t.co/iB3KDTMJGl

[Charles Feinn @dynamics]

TIL the difference between macro and micro averages.

[Charles Feinn @dynamics]

I just published my first blog post, Creating an Evaluation Framework! Excited to finally be applying what I learned in @dan_s_becker and @HamelHusain's Mastering LLMs: A Conference For Developers & Data Scientists.

https://t.co/nTqBgpAjiZ

[Jeremy Howard @jeremyphoward]

For those that hope (or worry) that LLMs will do breakthrough scientific research, I've got good (or bad) news:

LLMs are particularly, exceedingly, marvellously ill-suited to this task. (if you're a researcher, you'll have noticed this already)

Here's why🧵

[Charles Feinn @dynamics]

@natolambert I think the point is they are trying to pull all the folks not already using off the sidelines (and into their ecosystem).

[François Chollet @fchollet]

One thing that even relatively senior ML people often fail to grasp is that deep learning models are curves fitted to a data distribution. You cannot expect them to solve tasks outside of their training distribution (which is the sort of thing that you need intelligence for).

"Emergent learning" is an incorrect label -- if a model demonstrates performance on task A that it wasn't trained on, that simply means that there is significant overlap between A and all the data that you did train on. Competence doesn't magically emerge out of nowhere.

[Charles Feinn @dynamics]

@boazbaraktcs @RylanSchaeffer @BrandoHablando @sanmikoyejo seems like the mirage is being replaced by a higher fidelity understanding of what emergent abilities are and how they arise