The result from Gemini 3 Pro using the default media resolution (the medium one): "(Заголовок / Header): Арсеньев (Фамилия / Surname - likely "Arsenyev")
Состояние удовл-
t N, кожные
покровы чистые,
[л/у не увел.]
В зеве умерен. [умеренная]
гипер. [гиперемия]
В легких дыха-
ние жесткое, хрипов
нет. Тоны серд-
[ца] [ритм]ичные.
Живот мяг-
кий, б/б [безболезненный].
мочеисп. [мочеиспускание] своб. [свободное]
Ds: ОРЗ [или ОРВИ]" and with the translation: "Arsenyev
Condition satisfactory.
Temp normal, skin coverings [skin] are clean, lymph nodes not enlarged.
In the throat [pharynx], moderate hyperemia [redness].
In the lungs, breathing is rigid [hard], no rales [crackles/wheezing].
Heart tones are rhythmic.
Abdomen is soft, painless.
Urination is free [unhindered].
Diagnosis: ARD (Acute Respiratory Disease)."
My first language is Russian. I can't fully understand this dreaded "doctor's cursive", but I can see that some parts of Gemini's text is probably wrong.
It's most likely "но кашель сохр-ся лающий" ("but barking cough is still present"), not "кожные покровы чистые" ("the skin is clean"). Diagnose is probably wrong too. Judging by symptoms it should be "ОРЗ", but I have no idea what's actually written there.
No, transcription has nothing to do with written text, it guessed few words here and there but not even general topic. That's doctors note about patient visit, beginning with "Прием: состояние удовл., t*, но кашель / patient visit: condition is OK, t(temperature normal?) but coughing". But unreadable doctors handwriting is a meme...
This is a historical church document from 19th century and Gemini got it right with common words but completely hallucinated the names of village and people.
Right, it can do modern writing but anything older than a century ( church records and census)and it produces garbage. Yandex Archives figured that out and have CER in a single digit but they have the resources to collect immense data for training.
I'm slowly building a dataset for finetuning TROCR model and the best it can do is CER 18% ... which is sort of readable.
I'm using TrOCR because it's a smaller model that I can fine tune on a consumer card, but the age of the model and resources certainly make it a challenge. The official notebook for fine tuning hasn't been updated in years and has several errors due to the march of progress in the primary packages.
I think I based my notebook on the official example but yes at some point new versions of the libraries completely broke it. I had to pin the versions for it to work again.
It's funny that game makers make a fuss about anti-cheat not working on Linux but then publish Switch versions of their games. That platform has almost zero security and is commonly emulated with cheats even in multiplayer these days.
If people cheat in the switch, they can blame Nintendo. If people cheat in PC, they can blame the anticheat. Without anticheat, they have to take the blame.
This. Even kernel level anti-c-spyware can't stop a cheap vision model hokked to a mouse, see youtube for examples from simple auto input up to full on elctromuscular stimulation.
Yes the channel “Basically homeless” has a few variations on this. Using electrodes to move your muscles to more practical a bot that moves your mouse pad for you to give you perfect aim. No anti cheat can detect that because there is nothing to detect.
Although who knows, they might be outright lying about that just to scare cheaters, but I tend to default to assuming what they're saying is more or less true.
They can’t detect me splitting my hdmi output, feeding one of them to a separate machine with a vision model to detect what needs to be detected and the same machine moving and clicking the mouse. People are already doing this.
How does it know what isn't visible? Can it handle glass? Frosted glass? Smoke? What if I can't see the player but I can see their shadow? What if I can't see them because they're behind me but I can hear their footsteps? What if I have 50ms ping and the player is invisible after turning a corner because the server hasn't realized I can see them yet?
To answer all those questions you either have to render the entire game on the server for every player (not possible) or make the checks conservative enough that cheaters still get a significant advantage.
I never understoof why google gave up so early on cloud gaming. Clearly it is the future, the infrastructure will need to develop but your userbase can grow by the day.
I live a bit remote on an island group, and even though I have a 500Mbit Fiber, my latency to the next GeforceNOW datacenter is 60-70ms (which is my latency to most continental datacenters, so not NVidias fault). That makes it unplayable for i.e. Battlefield 6 (I tried, believe me), but I have been playing Fortnite (which is less aim sensitive) for 100+ hours with that.
And under such system, how do you stop people from abusing latency-compensation to make their character appear out of thin air on the opponent’s perspective by fake-juking a corner to trick the netcode into not sending the initial trajectory of your peeks?
The Switch is a closed proprietary platform, so Nintendo can give some guarantees, and if the user does something at the Switch level, the responsibility of legal action will be on Nintendo, saving up headaches to the publisher.
Or how using what amounts to a state-owned search engine - when that state is a corrupt, censorship-heavy authoritarian state dedicated to the conquer of their sovereign neighbor states - is helping toward "the best results". Outside of the "funding invaders" issue, of course it's results will be skewed.
Especially if he simultaneously claims it's hardly used and only represents 2% of costs. If it's so infrequently used, why so resistant to offering a toggle?
reply