erichocean

4 ms·

1.
by erichocean 9d ago
If you're an author of this, please follow up with the SME2 cores in Apple Silicon, and the AMX cores in Intel server processors.
2.
by erichocean 10d ago
Would be fun to get this API working on it now: https://github.com/sebbbi/NoGraphicsAPI You could code for it now directly, instead of having to wrap a driver API. (There's also a few API bits missing on Metal tod
3.
by erichocean 10d ago
Meditations on Moloch[0] explains why this is impossible at scale--too many defectors, asymmetric information, and no way to police defection. [0] https://www.slatestarcodexabridged.com/Meditations-On-Moloch
4.
by erichocean 11d ago
Agreed, but how will we find out? Pre-training alone on frontier-class models costs billions of dollars.
5.
by erichocean 11d ago
I could put this to use today. I think we'll see a bunch of different architectures over the next five years.
6.
by erichocean 11d ago
> PC-ALM closes the PC-BP gap at matched inference budget in nonlinear networks Okay, so no efficiency improvement?
7.
by erichocean 12d ago
So they lied to the public then? Wow, we should really trust these people today.
8.
by erichocean 13d ago
> Does anyone in the industry think any of these companies are actually close to that kind of self-improvement? Their current plan is to take the existing architecture and shorten the cycle times: move all new RLVR work into mid-traini
9.
by erichocean 13d ago
> Why are AI agents lying, cheating and coordinating? Have you seen the labs training them?
10.
by erichocean 14d ago
Anyone remember when GPT-3 was too dangerous to release?
11.
by erichocean 14d ago
They want the AI labs to stop solving "important" math problems so that humans can do it, on the theory that when AI solves "important" math problems, it robs humans of the opportunity to gain new insights into those pro
12.
by erichocean 14d ago
> vibe math Ah yes, Lean 4. Famous for how unreliable it is. Absolute slop!
13.
by erichocean 14d ago
AI is displacing real mathematicians.
14.
by erichocean 14d ago
I've had a similar experience. LLMs doing math has made math far more accessible to me than it was pre-LLM. Not just using existing math, but developing new math. (For example, I developed an alternative to NURBS surfaces for geometric
15.
by erichocean 17d ago
Astra with an animation app (pick whichever one you like)
16.
by erichocean 23d ago
Well the geniuses over at Anthropic have been showing it's text watermarking technology. "Hey AI, here's how to hide what you're thinking in normal looking language. Have fun!" A few moments later... "Woah, how
17.
by erichocean 24d ago
Really interesting project, appreciated.
18.
by erichocean 24d ago
If you want to do this in Clojure, Clobber[0] is a great base to start from. It can be used in headless-mode, I've hooked it up to the latest JavaFX text editing component it works very nicely. [0] https://github.com/ph
19.
by erichocean 1mo ago
> I don't want to stick linear or dependent types into TLA+. Proving dynamic properties with an exhaustive runtime is a totally different game from what you might do statically. Yup, TLA+ is a totally different beast from, say, Le
20.
by erichocean 1mo ago
> but was quickly shutdown by people telling me it was a useless endeavor due to Goedels incompleteness theorems You shouldn't have listened to them, "incompleteness" only applies when you try to encode the language and
21.
by erichocean 1mo ago
Scenario: an AI lab develops ASI for software development internally. The ASI produces bug-free software and human-readable specs. It's so reliable the company can guarantee the code matches the spec. Rather than provide tokens to de
22.
by erichocean 1mo ago
What I like about RISC-V is not the ISA per se, but the ecosystem that has developed around it, particularly Chisel and CIRCT. Specific choices for instruction encoding is less interesting, especially in the age of AI.
23.
by erichocean 1mo ago
I read the thread, and couldn't disagree more. > If Rust isn’t memory safe (because unsafe), and Zig isn’t (because uaf), then Fil-C isn't (because zunsafe_call/zunsafe_fast_call). Rust isn’t memory safe (because unsafe
24.
by erichocean 1mo ago
That's an interesting project, thanks for the link.
25.
by erichocean 1mo ago
I thought it was sorting an infinite set of infinite strings as the first step in the "algorithm" that seemed sketchy. [0] It's definitely not a constructive proof, even though it pretends to be; none of the mathematical obj
26.
by erichocean 1mo ago
Bizarre, it has a nearly identical improvement as the flash model.
27.
by erichocean 2mo ago
It took 90 days earlier this year for our venture-backed business to get approved by Apple merely for a corporate developer account, with all documentation available at the beginning of the process. They are insanely, glacially slow.
28.
by erichocean 2mo ago
> I've been able to formally verify that the Rust behavior is identical to the Postgres C behavior for over 1000 of them. Since both Rust and C have LLVM IR intermediates, you could use KLEE[0] for this. [0] https://kle
29.
by erichocean 2mo ago
That's because it has the best compaction of anyone, and it's not even close. Claude might as well not even do it in my experience.
30.
From Academic Research to a Frontier LLM: A Case Study in DPO [video](youtube.com)
2 points by erichocean 2mo ago|0 comments

More ›