miki.dragonscave.space.ap.brid.gy
blind coder / comp-sci student, working in automatic speech recognition for CLARIN. Polish. Libertarian leaning. Feel free to get in touch.
[bridged from https://dragonscave.space/@miki on the fediverse by https://fed.brid.gy/ ]
71 posts
1 followers
0 following
Prolific Poster
Conversation Starter
comment in response to
post
@simon Oh this'll be useful.
I'm doing a lot of work with structured outputs professionally, and being able to just try things out in the cli for quick prototyping and testing will be extremely useful.
This was definitely one feature I felt LLM was missing.
Thank you for such a great and […]
comment in response to
post
@briankrebs Apple urgently needs to deploy their location detection technology, which they're now using to make sure Americans can't install third party browsers, a lot more widely.
If banks knew where the fraudster was, and *how sure Apple was of that info*, the fraudsters' job would be a lot […]
comment in response to
post
@Unixbigot It's even better when the author actually means a *temperature difference*, something like "the temperature increased by 5 degrees C", but some stupid automated rule converts that using the traditional conversion formula, getting 45 degrees (where the correct value is 9).
comment in response to
post
@Unixbigot It's even better when the author actually means a *temperature difference*, something like "the temperature increased by 5 degrees C", but some stupid automated rule converts that using the traditional conversion formula, getting 45 degrees (where the correct value is 9).
comment in response to
post
@jscholes What's the current situation with KFX DRM removal? I've heard conflicting opinions.
comment in response to
post
Muskwatch.com analyzed URL traffic for an image sharing service run by Coristine (tesla dot sexy), and found between April 2021 and September 2021 that numerous noxious URLs redirected to Coristine’s site, including, “children-sex.party,” “child-porn.store,” “kkk-is-cool.club,” “nigga.rentals,” […]
comment in response to
post
@jscholes Yeah, that hint used to also be present on mobile, though I don't think it's the case any more.
comment in response to
post
@jscholes Enter departure station name - Mandatory field - The autocomplete information will be displayed after entering the first character. Use the UP and DOWN arrows to move between the subsequent results. Use the ENTER button to confirm the station selection. menu pop up edit text
comment in response to
post
The author of this model claims that it took $1000 to train. Meanwhile, $6m is considered cheap for *a single training run for a decent LLM*, and that doesn't even include ablations and experiments.
I really don't understand why this didn't happen earlier.
I thought Eleven Labs had some […]
comment in response to
post
@rdrozd Wystarczyły by podpisy cyfrowe na PDFach dyplomów (my w Polsce już mamy do tego infrastrukturę, wiele elektronicznych dokumentów urzędowych jest teraz autentykowanych w ten sposób).
Alternatywą jest jakiś QR kod i appka do weryfikacji, à la eDowód.
comment in response to
post
@mapache You need Windows for that nowadays.
Oh how the times have changed.
comment in response to
post
@rdrozd Rozkłady to jedno, dane o lokalizacji pojazdów to drugie.
Wrocław ma do tego wewnętrznie bardzo fajne API (jest pięknie opisane w dokumentacji przetargowej na pojazdy), ale to co wystawiają na zewnątrz to jest jakaś kpina.
comment in response to
post
@futurebird I haven't thought of this before, but this will cause so much IT chaos.
Many people in the US have non-US-issued documents that have a gender on them, and the US government needs to have accurate knowledge of what those documents say (regardless of whether it considers what they […]
comment in response to
post
It's kind of sad and extremely funny at the same time that Elon now owns the only platform that actually has a decent fact checking setup that has explicitly been designed to be unbiased from the start (and to be clear, the start was pre-acquisition).
comment in response to
post
@ProjectFearlessness How do grammatical errors and spelling mistakes indicate that the orders were produced using Chat GPT?
Language models don't usually make these.
Extremely blant writing, repetitive sentence and paragraph structure, and frequent use of idioms that are characteristic of […]
comment in response to
post
@ProjectFearlessness How do grammatical errors and spelling mistakes indicate that the orders were produced using Chat GPT?
Language models don't usually make these.
Extremely blant writing, repetitive sentence and paragraph structure, and frequent use of idioms that are characteristic of […]
comment in response to
post
@GossiTheDog Um, isn't that just a consequence of US people preparing for the Tiktok shutdown?