so maybe they do sample at 24 bit at a well chosen gain level and then convert to 32 bit float, with the max 24 bit value being above 1.0 float
or as GP said, use two separate ADCs at two different gains and combine their output
99 karma · joined June 24, 2026
so maybe they do sample at 24 bit at a well chosen gain level and then convert to 32 bit float, with the max 24 bit value being above 1.0 float
or as GP said, use two separate ADCs at two different gains and combine their output
your software is among the best, but not pitch black best :)
this has practical applications
which is why no US defense contractor is given unsupervised access to nukes, even when maintaining them
it was long predicted that it is inevitable for national governments to fully nationalize AI labs and put them under military control
https://www.reuters.com/world/europe/kremlin-demands-explana...
all OSes have malware level capabilities. it's literally the definition of an OS
https://huggingface.co/spaces/TabArena/leaderboard
xgboost is listed, they say "tuned" but who knows what that means. and its below CatBoost and LightGBM
as an aggregator of models (trees, neural networks, clustering, ...) AutoGluon doesn't really have a dog in this fight
Google bought DeepMind, so its possible
but not much interest
it's main client, TSMC had 4x, $120 bil revenue in 2024
the Chinese equivalents are 5 years away
sounds like VW from 10 years ago - "Chinese can't make quality cars"
or do you think EU doesnt really need that, it can just rent from the americans
what is the EU frontier AI model?
what is the EU frontier chip fabs?
> BERLIN, June 26 (Reuters) - Volkswagen is considering shutting four German factories and ramping up job cuts to as many as 100,000, two people familiar with the matter said on Friday, in what could be the biggest ever overhaul in the industry.
https://www.reuters.com/business/autos-transportation/volksw...
but its also the reason for the lack of EU tech capability and over-reliance on US
so naturally employees have a "fuck you" attitude
that's what I was asking, of modules are being introduced at scale or if they are a hobbist thing
https://nibblestew.blogspot.com/2025/08/we-need-to-seriously...
that's a common trope, and its both true and false
true: more bank employees after ATM's
false: less bank employees after smartphone banking
GPT-3 was 175B, models like Gemma4 with 31B vastly outperform it, so there is more to it
as Karpathy noted, the initial GPTs were trained on complete garbage (literally, the average document from the Common Crawl is random nonsense), yet they worked. now we can use present LLMs to curate the data for the next generation