harlekin.me

Do you know whats so ama­zing about Ukrai­nes coun­ter offen­si­ve in Pischa­ne? Its now a day after the coun­ter attack that took the vil­la­ge in a day and its cur­r­ent­ly 05:30pm in MEZ Euro­pe, and for today the­res not even any figh­t­ing repor­ted in Pischa­ne. The vil­la­ge is basi­cal­ly hol­ding itself!

Never mind no artil­le­ry and no bombs fired yesterday. 

No infan­try enga­ge­ments today at all!

Wow!

Thats ama­zing!

edit: Es gibt aber auch wie­der gute Nachrichten:

src: click

Ja, Alex Amon kann kei­ne deut­sche Gram­ma­tik, war aber mal mein Chef­re­dak­teur. Und ja, der Stan­dard beschäf­tigt kei­ne Edi­to­ren mehr.

Qua­li­täts­jour­na­lis­mus nennt man das.

VW, make a 10k USD electric car. But make sure it has Tes­las FSD fea­tures in it for free.

China’s BYD Adding ‘High-Level’ Self-Driving to Its Bud­get $10K Electric Car

Becau­se at the high end, this is the chi­ne­se auto­ma­ted dri­ving expe­ri­ence for 35k+ USD cars (which then use three lidar sys­tems) today:

Con­text:

Hua­wei [as seen in the video] is the num­ber one sup­plier of auto­no­mous dri­ving tech in China.

Hua­wei sent 24 employees out to test dri­ve Tes­las in Cali­for­nia during the free tri­al peri­od Tes­la gave for FSD. Their respon­se invol­ved lots of laughter.

Musk’s mee­ting with Li Qiang did not get any appro­val for Tesla’s FSD roll-out in Chi­na. The­re wasn’t even any such news in China.

Even Chi­ne­se EV makers need to go through arduous steps in each city in Chi­na to get their Level 2 driver-assist sys­tems appro­ved (none have Level 3 appro­ved as of now).

Ogan belie­ves the most advan­ced auto­no­mous dri­ving sys­tem is Pony Ai, which has huge ope­ra­ti­ons in Chi­na but goes through extre­me­ly high hurd­les to exchan­ge data bet­ween its tests in Chi­na and the US (Tes­la doesn’t even have sizab­le data cen­ters in Chi­na set up to do the same).

src: click

This is so crazy.

The ger­man auto mar­ket real­ly slept through everything…

So now for them its - licen­se Micro­soft Auto for the dash­board, buy - Chi­ne­se FSD packa­ge from Hua­wai. Make a Hua­wai Micro­soft auto inte­gra­ti­on, that works on an Android Pho­ne and an iPho­ne. Put that into a 10K car. Stay cost competitive.

Its insa­ne.

So of cour­se thats never going to hap­pen - and they are tar­ge­ting the “real dri­ving fee­ling” niche in Chi­na, until it shrinks and shrinks and shrinks and dies.

Bra­vo.

Well, also lob­by the EU a bit more, why dont you?

The real­ly insa­ne bit is - that this has been out for two years -- and no one in the west even has a clue. Becau­se cul­tu­ral­ly - this couldnt have been hap­pe­ning, right? Someo­ne would have told us… On red­dit? Face­book? The news?

From the you­tube comments section:

@MrJohnson088
vor 6 Tagen
I just want to let my Wes­tern friends know how serious our infor­ma­ti­on bar­ri­ers are through this video. This kind of advan­ced stuff can’t be made over­night. It takes a long time of rese­arch and deve­lo­p­ment, prac­ti­ce and tes­ting. It’s not some­thing that sud­den­ly falls from the sky one day. Some­ti­mes when we are tal­king about some pro­blems, this is why the worlds we know [for] each other are com­ple­te­ly different.

Fuck.

edit: Oh sor­ry, Mer­ce­des wan­ted to pitch in…

The pri­ce of DRIVE PILOT remains unch­an­ged (star­ting at 5,950 euros incl. VAT).

src: click For SAE Level 3 (in their dreams…):

So the­re was a ukrai­ni­an coun­ter ass­ault towards the vil­la­ge of Dach­ens­ke, that appar­ent­ly was suc­cess­ful - and ama­zing in its own right, becau­se Ukrai­ne mana­ged to form coun­ter offen­si­ve for­ces in the Donbas.

Thats pret­ty much wide­ly repor­ted, and likely to have hap­pen­ed. But then the­re is this ent­i­re ama­zing ukrai­ni­an coun­ter offen­si­ve cur­r­ent­ly in Pischa­ne - that no one wants to talk about.……

Apparant­ly the Ukrai­ne mana­ged to reta­ke the ent­i­re­ty of Pischa­ne wit­hin a day (third rif­le sym­bol from the left:

Which is so very ama­zing, becau­se it hap­pen­ed right befo­re the Munich Secu­ri­ty Con­fe­rence later this week, and should be the most for­ti­fied part of the rus­si­an front oppo­si­te Pokrovsk.

Its also so ama­zing, becau­se - the ISW for once is more con­ser­va­ti­ve with indi­ca­ting ukrai­ni­an ter­ri­to­ri­al gains in the area:

And doesnt omit, that rus­sia claims they still con­trol all of the area north to the area the Ukrai­ne claims it has captured.

Whats even more ama­zing is, that no one wants to talk about that stra­te­gic mar­vel of a mis­si­on -- as this is the ONLY news report on an offen­si­ve hap­pe­ning in Pischa­ne I could find so far:

In a cri­ti­cal deve­lo­p­ment on the eas­tern front, Ukrai­ni­an for­ces have laun­ched an offen­si­ve against Rus­si­an tro­ops in the vil­la­ge of Dach­ens­ke, loca­ted in the Donetsk region.

Accord­ing to reports, Ukrai­ni­an tro­ops, who are now in con­trol of the eas­tern part of the vil­la­ge, are pushing towards the cen­ter in an effort to rec­laim the ter­ri­to­ry. The ongo­ing con­flict in Donetsk has spar­ked inten­se batt­les, as Ukrai­ni­an for­ces enga­ge in fier­ce com­bat to regain stra­te­gic advan­ta­ges from Rus­si­an for­ces, reports Moochny, a figh­ter from the Ukrai­ni­an For­ces, on his Tele­gram channel.

“Our Cos­s­acks suc­cess­ful­ly atta­cked Rus­si­an posi­ti­ons from the north in the eas­tern sec­tion of Dach­ens­ke. Having secu­red this area, we’­re now enga­ging in the batt­le for the vil­la­ge cen­ter, with plans to steadi­ly move west­ward, aiming to regain full con­trol over the area,” Moochny reports.

Addi­tio­nal­ly, the Ukrai­ni­an for­ces con­ti­nue to mount attacks sou­the­ast of Kot­ly­ne and are enga­ged in com­bat around the Pischa­ne area.

src: click (uawire.org) -- from yes­ter­day 16pm

Whats also ama­zing is, that rus­sia appar­ent­ly has not moun­ted any air- or artil­le­ry­strikes on to Pischa­ne yes­ter­day or today!

Thats pret­ty ama­zing actually…

edit: The Black Bird Group is on it:

Ukrai­ni­ans are achie­ving limi­ted suc­cess. It’s likely an attempt to fur­ther grind down the Rus­si­an attacks which have slo­wed down during the last two mon­ths. The Ukrai­ni­ans cer­tain­ly also want to con­tain the thre­at of an encir­cle­ment around Pokrovsk. 2/

— Emil Kas­te­hel­mi (@emilkastehelmi) Febru­a­ry 10, 2025

edit: Deep­Sta­te has just updated its map:

If that was the direc­tion of the coun­ter attack - ehm “limi­ted” is the word that comes to mind, yes… Not so much the word counteroffensive.

Stan­dard: Blah, Blah, Blah, lüg, miss­ver­steh, nicht aus­kenn - falsch aus­sag, Bevöl­ke­rung ver­arsch, was bin ich für ein gro­ßer Exper­te, blah…

Und dann wie anno 1994, als Trot­tel wie sie noch nicht im Inter­net unter­wegs waren:

Ers­tes Kom­men­tar (zum Zeit­punkt des Aufrufs)

Kor­rekt.

Da rackerst du dich als Stan­dard Redak­teur ne Stun­de ab um kom­plet­ten bull­shit zu pro­du­zie­ren, und dann kommt ein Nut­zer daher, der mehr als eine Stun­de inves­tiert hat, der dich ein­fach kor­ri­giert, wo gibts denn sowas? Also nicht im Netz seit 1994!

Back­ground.

Ich lass mir seit drei Jah­ren Unter­ti­tel maschi­nell über­set­zen - und das vom Tool von “Akbar on Development”

https://www.syedgakbar.com/projects/dst

Ich habe mir das Equi­va­lent von 35 1950er comics maschi­nell über­set­zen las­sen, als ich die mal für drei Wochen gele­sen habe.

Dabei jedes mal die güns­tigs­ten Model­le ver­wen­det, um noch eine brauch­ba­re Qua­li­tät zu erhalten.

Hier ein Ver­gleich­front­end für eini­ge Model­le und Sprachen:

https://frontendtranslator.onrender.com/compare-translate

Bei TTS ist das aktu­ell viel­ver­spre­chends­te Modell kokoro-tts
https://huggingface.co/spaces/webml-community/kokoro-webgpu

Sie­he auch: https://claudio.uk/posts/audiblez-v4.html

edit: Die bes­te kom­er­zi­el­le Lösung zur Zeit ist: https://cartesia.ai

Das ehe­mals bes­te kos­ten­lo­se Modell, das noch kei­ne Into­na­ti­on from word con­text beherrscht Ivo­na Mar­le­ne für Deut­sche Into­na­ti­on und Ivo­na Amy für bri­tish English:
https://speech2go.online/ (Die Android Apps (apkpure.com) kos­ten­los nutzbar.)

Hier das tat­säch­li­che Fazit:

Es liegt am Lan­guage pair.

Eini­ge Model­le sind deut­lich bes­ser als andere.

Das Resul­tat war seit mehr als einem Jahr immer brauchbar.

Pri­ce per­for­mance Sie­ger war ChatGPT 4o. Rea­so­ning Model­le habe ich dafür noch nicht ver­wen­det. Mul­ti line trans­la­ti­on wird oft­mals zum Pro­blem beim Untertitel-Formating, wes­halb die meis­ten im Pro­duk­ti­ons­um­feld wohl nur sin­gle line Kon­text nut­zen dürf­ten. Promp­t­ing to skip, if unsu­re hilft um Gib­be­rish wei­ter zu reduzieren.

Deepl wird aktu­ell bereits von aktu­el­len LLM Base models ausgestochen.

Model­le die ähn­lich wahr­schein­li­che Wort Über­set­zun­gen anzei­gen, und die bei Bedarf wech­seln las­sen - erspa­ren bereits in 99% manu­el­les Nach­über­set­zen, nach ein­ma­li­ger manu­el­ler Korrektur.

Idio­me blei­ben problematisch.

TTS die Into­na­ti­on aus dem Kon­text auto­ma­ti­siert ablei­tet ist bereits sehr gut.

TTS die es schafft dabei das Voice Pro­fil des Spre­chers zu clo­nen noch mal deut­lich besser.

Sie­he:

Was mich bereits zu fol­gen­dem Nut­zungs­pro­fil bewo­gen hat -- auch wenn mein Eng­lisch in Rea­ding, Use of Eng­lish and Lis­tening der höchs­ten Punk­te­be­wer­tung von CEFR Level C2 ent­spricht (bes­ser geht nicht) las­se ich mir doch bereits fürs casu­al watching einer doo­fen Serie, oder eines Films Unter­ti­tel auto­ma­ti­siert von eng­lish auf Deutsch über­set­zen, da ich Deutsch ent­spann­ter (weni­ger Akti­vie­rung) lese. Und das seit nem Jahr. For­mel­les Sie Feh­ler, fal­sche Pro­no­mi­na Wech­sel (erneut, Kon­text­li­mit “per Zei­le”) -- alles egal, ich les es lie­ber auf Deutsch.

Zwei drei gro­be Schnit­zer Pro Film, bei denen man kurz auf­lacht, die aber meist im Kon­text des Films sofort rich­tig iden­ti­fi­ziert wer­den kön­nen -- dafür kor­ri­gier ich nichts.

Beson­ders wich­tig wird es für mich bei Fil­men die ich ger­ne in einer Spra­che sehen wür­de, die ich ver­ste­he, bei denen aber nur ander­spra­chi­ge Unter­ti­tel vor­lie­gen. Auch das klappt bei­na­he immer. Über­set­zun­gen nach eng­lish sind dabei qua­li­ta­tiv bei­na­he immer bes­ser. (Hol­län­disch Deutsch ist die ein­zi­ge Aus­nah­me von der Regel, die mir bis­her unter­ge­kom­men ist.).

Mensch­lich erstell­te Unter­ti­tel sind in 80% der Fäl­le in Punk­to Text­qua­li­tät immer noch deut­lich besser.

Give Selen­skyj a fal­se dicho­to­my, and he will hang hims­elf with it.

“Mr. Selen­skyj, are you a pre­si­dent for vic­to­ry, or a pre­si­dent for peace?”

60 IQ Mana­ger pos­se to the res­cue! Oh no - they got stuck in a near­by room, cir­cle­je­r­king on a coo­kie. (Poli­ti­cal lea­ders­hip­TM team of a country.TM)

Oh, and of cour­se I’m such a gre­at pre­si­dent and mili­ta­ry lea­der, that I even mana­ged, that Cri­mea will come back to ukrai­ne. Kursk was essen­ti­al in this. Gre­at Kursk. Such for­ce - our peop­le. Wow.

But dont you worry, neit­her you - nor the ent­i­re ger­man spea­king news­me­dia was able to iden­ti­fy that this was hap­pe­ning - for at least one and a half years past the point whe­re it would have been important to report on that.

They sim­ply fed you lies -- which now they claim to not see any­mo­re… Oh no pro­blem - just pay us for our gre­at jour­na­listic work a bit more… LOves that Selen­skyj guy! Such bra­ve. Such power.

Cor­rect ans­wer would have been “I’m the pre­si­dent of Ukrai­ne” of course.