1
point
statusfailed
Joined 91 karma
https://statusfailed.com/about.html
- Which repos worked well? I've had the same experience as op- unhelpful diagrams and bad information hierarchy. But I'm curious to see examples of where it's produced good output!
- Really nice! Had a quick read, here's my quick summary:
- Arrays are typed `S : D` with shape S and strides D
- Each of `S` and `D` is a nested tuple (instead of the flat tuples one typically sees in a tensor framework)
- Together `S` and `D` define the layout of a tensor
- Not every layout is "tractable", but the tractable ones form a nice category
A really good exposition, my only criticism is that it's quite front-heavy- it would be nice to see a detailed example like in 2.3.8 earlier in the document; there is a lot of detail presented early that doesn't seem necessary to understand the core ideas.
Last comment: I suspect there is a connection to strictification[0], would love to know more if the authors see this!
[0]: in the sense i mean here: https://arxiv.org/pdf/2201.11738v3
- Hellas.AI | Remote | Full Time or Contract | https://hellas.ai
Hellas.AI is building a compiler for categorical deep learning to power decentralised, serverless AI.
We're looking for an experienced Rust engineer. You must have demonstrated:
* Experience in designing and implementing BFT consensus protocols * Ability to engage with academic literature on distributed systems * Understanding of ZK proofs and proof aggregation (you don't need to design these, but you should know how to use them)
Nice-to-have: interest or experience in:
* Deep Learning / ML / AI, GPU programming * Compilers and array programming languages * Category Theory and string diagrams
If you have project(s) demonstrating your experience in one or more of these areas, get in touch!
Our process is simple: we read code you've written and research you've published, then interview you about it. We don't do tedious coding tests.
Email us at hello@hellas.ai
- I've been hoping for a nice concrete example of braided monoidal categories for ages, who knew that the best one was string diagrams that represent actual string! Great post!
@the author - I assume you're aware that morphisms in symmetric monoidal categories can be represented using cospans of hypergraphs - do you know if there's a similar combinatorial representation for braided monoidal categories?
- (author here) the goals of catgrad are a bit different to JAX - first of all, the "autodiff" algorithm is really a general approach to composing optics, of which autodiff is just a special case. Among other things, this means you can "plug your own language" into the syntax library to get a "free" autodiff algorithm. Second, catgrad itself is more like an IR right now - we're using it at hellas.ai to serve as a serverless runtime for models.
More philosophically, the motto is "write programs as morphisms directly". Rather than writing a term in some type theory which you then (maybe) give a categorical semantics, why not just work directly in a category?
Long term, the goal is to have a compiler which is a stack of categories with functors as compiler passes. The idea being that in contrast to typical compilers where you are "stuck" at a given abstraction level, this would allow you to view your code at various levels of abstractions. So for example, you could write a program, then write an x86-specific optimization for one function which you can then prove correct with respect to the more abstract program specification.
- Hellas.AI | Remote | Full Time or Contract | https://hellas.ai
Hellas.AI is building a compiler for categorical deep learning to power decentralised, serverless AI. If you're interested in any of the following, get in touch:
* Category Theory and string diagrams
* Compilers and array programming languages
* Deep Learning/ML/AI and/or GPU programming (e.g. CUDA/ WGSL)
* Building and deploying distributed systems
* Peer-to-peer software and blockchain
If you have project(s) demonstrating your experience in one or more of these areas, get in touch!
Our process is simple: we read code you've written and research you've published, then interview you about it. We don't do tedious coding tests.
Some experience of Rust or a functional language (Haskell, OCaml, any Lisp) is preferred.
Email us at hello@hellas.ai
- 2 points
- Seems bad. "An attacker was able to achieve code execution in the content process by exploiting a use-after-free in Animation timelines. We have had reports of this vulnerability being exploited in the wild."
See:
- NVD page for CVE-2024-9680: https://nvd.nist.gov/vuln/detail/CVE-2024-9680
- Mozilla security advisory: https://www.mozilla.org/en-US/security/advisories/mfsa2024-5...
- I only had a quick look, but it looks like they tweaked the state update so the model can be run with parallel scan instead of having to do it sequentially.
- 2 points
- I'd love to know what your use case is that makes those things important to you - and what kind of benchmarks and cleaning tasks do you need to run?
Also, what kind of evaluations for quality of reasoning do you use?
- This lets other family members send emails too, and has an interface everyone is familiar with.
- At least for chinese there are off-the-shelf word segmenters you can use like jieba[0]- I used it in gptlingo and it Just Works(TM).
The "show pronounciations" setting just turns on pinyin above characters - what I want is to type pinyin and enter chinese characters. Actually showing the pinyin above characters is quite distracting!
- > Duolingo is good at going from nothing to something
Seconding this. I got noticeably better at Chinese after using duolingo every day. I feel like I hit a ceiling now and it's not helping too much, but it definitely worked.
- OK I played with this some more, it's so good - exactly what I dreamed of!!
A couple more bits of feedback:
(1) The "suggestion" / "I'm unsure" etc. feedback is fantastic
(2) Word segmentation doesn't seem to be working properly, and so the context lookup doesn't work right. Example:
中国 should be parsed as a single word ("china"), but it's parsed as individual characters ("middle", "kingdom").
This means I have to tab out to a dictionary to look up words, and it's a bit annoying to select the right text.
- Nice! I built the same thing to learn chinese[0], it even has the in-context word lookup!
I have a feature request: if I don't have a pinyin IME installed, it's very hard to use - it would be nice to have an in-browser one!
- I used postgrest (without htmx) on an old project; it's impressive how far you can push it. HTMX seems like a perfect fit for it too, although I'm not sure how much I really want to maintain htmx templates inside SQL functions...
- Location: London, UK (able to work in the US without visa) Remote: Preferred Willing to relocate: No Technologies: Python, Haskell, Rust, HTML/CSS, Javascript, AWS, Terraform, Postgres, & more. Résumé/CV: https://statusfailed.com/files/cv.pdf Email: paul@statusfailed.com Website: https://statusfailed.com GitHub: https://github.com/statusfailed
I'm a technical generalist with 10 years experience and a Ph.D in CS focusing on category theory and machine learning. I've worked in a variety of roles both full-time and contract: Full stack developer, data scientist, pentester, and machine learning researcher. Most recently I was awarded a grant to develop a differentiable IR for zero-knowledge machine learning[0].
[0]: https://yarrow.id
- 1 point
- I'm in the same boat- I just want a simple headset that overlays a big, high-resolution monitor on my normal vision. The closest thing I know of is the [Simula One](https://www.simulavr.com/), which seems to be a VR "virtual monitor".
VR instead of AR is a dealbreaker for me though- I want AR virtual monitors that overlay the displays on my actual vision. Even VR which uses a camera to pass through to the display isn't really what I want.
My ideal hardware would just be a "dumb" AR display with wide FOV and an HDMI input that just overlays the video data onto my normal vision.
From what I understand a big challenge is the wide FOV though, but I don't know enough about optics to really understand why this is the case.
- What kind of applications do you see for training on mobile devices? Is anyone using this in industry?
- I (as a Ph.D student) paid open access fees with my own money. Not grant money, my own salary.
- For anyone wondering, the parent is quoting "The Dispossessed"[0] by Ursula K. Le Guin, which is one of my all-time favorite books. Definitely read it if you like science fiction.
- If this interests you, you will probably enjoy my favourite Wired article of all time: "Mother Earth Mother Board"[0] (By Neal Stephenson! in 1996!)
- It mostly funds projects:
> Approximately 94% of the EU budget funds programmes and projects both within member states and outside the EU. Less than 7% of the budget is used for administrative costs, and less than 3% is spent on EU civil servants' salaries.
From [0], see also [1]
[0]: https://en.wikipedia.org/wiki/Budget_of_the_European_Union#P...
[1]: https://www.europarl.europa.eu/about-parliament/en/organisat...
This is a little outside my area, but I think the relevant part of that abstract is "Gradient-based optimization follows horizontal lifts across low-dimensional subspaces in the Grassmannian Gr(r, p), where r p is the rank of the Hessian at the optimum"
I think this question is super interesting though: why can massively overparametrised models can still generalise?
[0]: https://opt-ml.org/papers/2025/paper90.pdf