30 karma · joined July 20, 2023
poke to wake up
apart from that: the word "mark" comes from a root for "boundary" or "border", and really it doesn't need to be about that; we're all in this together.
From what I can tell the bulk of the work for it was done in 2018[1], but it needs updating to consider SIMD, and for legwork to be done on moving it along as a proper spec extension.
Until someone picks up this valuable work and lands this much-needed feature in Wasm, we're extremely vulnerable to timing attacks in all Wasm crypto.
[0] https://github.com/WebAssembly/proposals/blob/9fc7a85e/inact...
Checks out I guess.
https://news.ycombinator.com/item?id=43121383
It would have to be from unsupervised tool usage or accepting backdoored code, not traditional remote execution from merely inferencing the weights.
The package manager just offers a common interface for interop, you can still build without dependencies.
It's interesting and a headline worthy result, but I think the research they were doing where they accidentally found that line of questioning is slightly more interesting: does an LLM trained on a behavior have self-awareness of that behavior. https://x.com/betleyjan/status/1894481241136607412
I guess adding memory to some cards is a matter of completely reworking the PCB, not just swapping DRAM chips. From what I can find it has been done, both chip swaps and PCB reworks, it's just not easy to buy.
Software support is of course another consideration.
How does "clamshelling" get around the 32-bits per module requirement? Do the two 2GB modules act as one 4GB module when clamshelled?
GB10, or DIGITS, is $3,000 for 1 PFLOP (@4-bit) and 128GB unified memory. Storage configurable up to 4TB.
Can be paired to run 405B (4-bit), probably not very fast though (memory bandwidth is slower than a typical GPU's, and is the main bottleneck for LLM inference).
My card: https://files.catbox.moe/t8rzlb.png
No thread up though, and the archive doesn't support the tags.
https://support.google.com/youtube/answer/132596
>Ads appear on brand safe sites: YouTube works diligently so that our advertisers' brands appear on sites that reflect our respective core values. Our systems closely evaluate websites and their content against various factors when finding out whether to turn on In-stream ads on YouTube embeds. These factors include a strict set of guidelines on content like adult imagery, violence, inappropriate and hateful language, and sites that promote infringement.
The web version was mostly replete last I used it half a decade ago, including realtime collaboration with mac and iOS.
I revived this simple loop the other weekend and it was a breeze to extract the media;
https://fwoofy.moe/ (sound warning)
Original https://dagobah.net/flash/fwoofy.swf