nmstoker

22 ms·

301.
by nmstoker 2y ago
The deduplication discussion shows they don't filter out ads as part of their cleaning - I appreciate this could be risky and perhaps a huge processing step given dataset sizes, but intuitively it feels like it would cut the noise dram
302.
by nmstoker 2y ago
I'm guessing sometime between now and next few weeks. With things like this there's usually a few last recordings with some audio to clean up etc etc. A substantial number from PyCon US 2023 went out on 02-Jun-2023 (ie exactly a y
303.
PyCon US 2024 Recap(katherinemichel.github.io)
186 points by nmstoker 2y ago|91 comments
304.
by nmstoker 2y ago
Am continuing to reflect on this, so may have more to say later, but what struck me first was the commitment to openness with FineWeb, it is impressive and thorough (eg all the scripts they actually used are available and linked to, not jus
305.
FineWeb: Decanting the web for the finest text data at scale(huggingface.co)
127 points by nmstoker 2y ago|14 comments
306.
by nmstoker 2y ago
There's no mention of handling with regard to costs inappropriately incurred - wouldn't access to the secrets let people call APIs and run up costs? Or is this purely about theft of data/code?
307.
Unprojecting text with ellipses (2016)(mzucker.github.io)
151 points by nmstoker 2y ago|21 comments
308.
by nmstoker 2y ago
Yes, this is a big weakness. Seems tied in large part to not tackling versions as well as they could have done.
309.
by nmstoker 2y ago
"This decision is the result of a violation of Company policy"?
310.
by nmstoker 2y ago
Some related material here too: https://twitter.com/bfspector/status/1789749117104894179?t=k...
311.
GPUs Go Brrr(hazyresearch.stanford.edu)
1104 points by nmstoker 2y ago|263 comments
312.
by nmstoker 2y ago
Totally agree. Well before AI issues the platform was turning a blind eye to low quality questions and answers, ones obviously written by people with limited experience and judgement, focused mainly on chasing completion targets over correc
313.
by nmstoker 2y ago
It's open source ( https://github.com/cq2-co/cq2 ) and not part of Salesforce, both appealling features. But deeper than that I have no experience of Quip (and precious little of CQ2 beyond the blog post) so can&#x2
314.
by nmstoker 2y ago
I always feel sympathy for the devs on this project as they get so many issues raised by people that are largely lazy (since the solution is documented and/or they left out obvious detail) or plain wrong. I suspect it's a side eff
315.
by nmstoker 2y ago
Definitely no LLM! Espeak dates from at least 10 years before LLMs appeared and was based on the approach used on Acorn computers in the 80s and 90s.
316.
Smart gadgets: Tougher rules for sellers of internet-enabled devices in the UK(bbc.co.uk)
4 points by nmstoker 2y ago|0 comments
317.
by nmstoker 2y ago
I think the underlying platform (built by that team) was posted here only a week or so ago... I'll see if I can find it EDIT: found it! LTK: https://github.com/pyscript/ltk
318.
by nmstoker 2y ago
Depressing. Would be interesting to see if once he'd established a record with Google for attracting users, if the site owner were to slowly drop the ruinous features, whether Google would turn a blind eye and in time it would be back
319.
by nmstoker 2y ago
Reminds me of the shudder of fear that "stranger danger" induces in parents, even though evidence suggests massively more risk to children is from those known to them. Stranger danger also ties in with other recall aspects reinfor
320.
by nmstoker 2y ago
Yes, I'd be curious to know how much better it is than them - from my limited understanding, they also share many of the advantages that Nimble boasts of, thus I can appreciate they'd both be better than legacy formats but it&#x27
321.
by nmstoker 2y ago
A friend told me about how his work had some senior IT mgrs, who'd clearly been playing with their iPhones too long, decide that the firm shouldn't use Ids at all any more, and started pushing this without consulting the business,
322.
by nmstoker 2y ago
Fantastic if you do, as it's a nice elegant yet simple design you've created. I totally thought it meant computer language too, but that's likely the context of viewing it here! Anyway, great job with this tool, whether or no
323.
by nmstoker 2y ago
2001 - it was on TV one afternoon on a rainy bank holiday in the early 80s, when I was probably far too young to fully appreciate it, and what I saw was simply awe-inspiring.
324.
by nmstoker 2y ago
I like and pay for Kagi but for me, outside of search, the rest of largely ignored fluff (aeema to chime with several other customers commenting here, eg on AI features). The search is generally good enough, often really good, and the kille
325.
by nmstoker 2y ago
I would suggest first removing the reference to LLM in the description: "build fast LLMs from scratch!"
326.
by nmstoker 3y ago
Should probably have "(2019)" appended to the title as per HN conventions, given the date in the citation paper links (1x 2019 and 2x 2020) and the fact that this site has been around for quite a few years...
327.
by nmstoker 3y ago
That's cool - I had been thinking with the OP site that quite a number looked like they were scowling/angry so a way to vary that is interesting to explore.
328.
by nmstoker 3y ago
Will depend on which browser you want to use (on mobile). Assuming the core editing code runs in mobile Mozilla, the sync aspect should work and I've used bookmarklets in mobile browsers before.
329.
by nmstoker 3y ago
The key part is you cannot change your birthday - that aspect makes it clear identity theft and not just use of a different given name (legal in many jurisdictions).
330.
by nmstoker 3y ago
Yes, this is correct. Provided you are not doing it to deceive it's perfectly legal in the UK - however you can't go changing other details like your date of birth, which Kierans did here. The problem in the UK is many people in p

More ›