Da zepp cuntduckt zum dizturbin rezearch


#21

tru

:sunglasses:


#22

And both those hos are men in drag

Da tranny/crossdressing tech got bettah but so did our methods of clocking dem.


#23

Daim is the last one a painting? What about 150 years ago?


#24

:wink:


#25

HAHAHAHAHA REZPEC!!!

tru da MART cummah wiz a very reazonable refutation :sunglasses:

howevah, da zhort diztance ZPRINT i dun think iz cumparable to 88 playin fo 1 zimple reazon:

diz a zheeyat mofoz haff been doin fo literally millionz of yrz fo zurvival even b4 dey iz modern mofoz, tiz az natural n ezzential az eatin n zheeyatin n da zpeed iz lykly already evolutionarily optimized :monkey: :monkey:

to be cumparable to da 88, tiz haff to be an ‘unnatural’ zheeyat u dun uze in daily lyf, an invented activity bazically where dere vil be room fo tech development. n TRU i agree wiz da ‘we iz ztill in da naive age of tech’ point :sunglasses:

n randomly even wiz runnin, if u make it a lezz natural zheeyat by zimply increazin da diztance, u get very zimilah rezultz. da marathon tym doublin cummah very cloze to da birth of da PIMP. da long diztance WALK around brotha tym:

marathon 1908 175 121 1.003329815 208.5102718 1810.489728 1810
50km walk 1921 280 212 1.002842843 244.1682638 1774.831736 1775
20km walk 1911 98 76 1.002356793 294.4525703 1724.54743 1725

mo data cummah. da lazt column iz da projected doublin-yr tru :sunglasses:

110 hurdles 1908 15 12.8 1.001429895 485.1002327 1533.899767 1534
400m hurdles 1908 55 46.78 1.001459419 475.2938063 1543.706194 1544
100m sprint 1912 10.6 9.58 1.000946021 733.0438879 1285.956112 1286
200m sprint 1951 20.6 19.19 1.001043217 664.7790469 1354.220953 1354
400m run 1900 47.8 43.03 1.00088382 784.6096358 1234.390364 1234
800m run 1912 111.9 100.91 1.000966603 717.4424308 1301.557569 1302
1500m run 1912 236 206 1.001271421 545.5216891 1473.478311 1473

alzo zum diztance zheeyatz. note da doublin (or halvin in diz caze) fo all da THROWIN ZHEEYATZ (ie optimization of da mvmtz of da front limbz to generate zpeed n powah :raised_hand_with_fingers_splayed: :fire: :musical_keyboard: ) cummah durin PIMPz lyf tym :sunglasses:

high jump 1912 2 2.45 1.001898443 365.4599383 1653.540062 1654
long jump 1901 7.61 8.95 1.00137544 504.2924074 1514.707593 1515
triple jump 1911 15.52 18.29 1.001521758 455.8375361 1563.162464 1563
#DIV/0! #DIV/0! #DIV/0! #DIV/0!
shotput 1909 15.54 23.12 1.003618173 191.9201967 1827.079803 1827
discus 1912 47.58 74.08 1.004146264 167.52023 1851.47977 1851
hammer 1913 57.77 86.74 1.003841752 180.7710898 1838.22891 1838
javelin 1912 62.32 98.48 1.004285522 162.087928 1856.912072 1857

#26

Hahaha dis “Tocatta by Scarletti”


#27

tru juz google

19th century figurative painting
19th century realism painting

n haff an open mind :sunglasses:


#28

fuckkkkkkkkkkk da zepp haff randomly cunzidahed diz zubject b4

cuntceptually

writin a zoftware algorithm dat control two groupz of 5 pozzible ztrike pointz wiz zum maximal-diztance reztrictionz in between iz a zumwut ztraightforward zheeyat

zo da zepp thought zhorly tiz zhud be eazy to build a robo-penizt wiz handz dat look n move lyk human handz, upload diz algorithm, n it vil be able to quickly analyze any 88 zong inputed az a midi file to work out viable fingahrinz n perform an inztant ZITERAPE :lib:

but when i looked mo in-depth into diz, it appearz da ROBODICKZ technology iz nowhere near advanced enuff to handle da precize micro mvmtz of 88 playin while maintainin zpeed n rhythm. zo not only iz a ziterape-bot out of da queztion, even a pre-programmed rape of a moderately hard zong iz a non-trivial zheeyat if u iz reztricted to 10 ztrike pointz wiz inta-digit diztance rezitrictionz (da DONGAH mention a robot penizt wiz 50 ztationary fingahz attached to da keyz which iz indeed cheatin n wud only require quazi 1960z technology)

zo da well trained mvmtz of a HUMAN HAND on da :musical_keyboard: iz hilariouzly

an inzanely cumplicated zheeyat fo machinez to emulate

da 10-digit ROBOpenizt iz lyk da holygrail of ROBODICKZ technology

equivalent to GO programmin fo machine learnin AI zheeyatz pozz :sunglasses:

zo NOZTRADAzepp predickteth

da TRU END OF HUMANITY vil cummah when

a ROBO-PENIZT wiz 10 fingahz

defeat da reignin FAZTEZT HUMAN 88 MOFO in a DUEL :robot: :fire: :musical_keyboard: :sunglasses:


da predickted tym fo diz to happen iz 2027

incidentally, 2027 iz alzo da tym when da zepp plan to reach









PEAK TECH :zepp: :icon_stopw_sdc_473: :sunglasses:


#29

This got me thinking. :lib: I’m not a computer scientist but it wouldn’t surprise me if CNNs are on top of the control aspect today, at least when it comes to imitation. You have a clear objective function to minimize and a simulator in piano + robot hand, so it should “just” be to let the network fool around with the piano and learn which movements brings it closer to sound like the human recording. At least if you can pre-program the network so the initial weights correspond to a midi of the piece I think there’s a fair chance it would start to converge (if pozz after 5 years or so of uninterrupted practicing with current systems).

Even so it’s true that you still have the hydraulics and mechanical aspects left to worry about though. But if that can be solved I think you could even take it beyond imitation with a generative system and Spotify or something as training data, but to create meaningful performances on its own - of any piece - and hence really compete with a human pianist I think you’d need a complete AGI system with strong levels of abstraction, reasoning, understanding and real world experience.

Maybe by 2050… But it’s cool these things have become valid science problems all of a sudden, and not just hopelessly unrealistic science fiction.


#30

ahahaha tru u can uze da NN if u want a prog to mimic a GOLDENAGE INTERP includin rape, random fuckupz, wtf rubato etc. da MART haff dizcuzzed diz on facerape n pozz already developed a prototype awaitin NAZDAQ liztin :sunglasses:

da zepp n da MYZTERY PENIZT can both provide unlimited datazetz fo trannin :sunglasses:

rite now da zoftware zheeyatz r no problemz

ROBODICKZ iz da tru bottleneck :robot: :eggplant:


#31

Still, I don’t really need a robot hand - I’d be totally fine with a system you just plug on to the keyboard, with one robot finger per key. Getting that to play the Auber Tarantella like Cziffra would be super cool, and just having it do a random generic Spotify trained performance of the Appassionata would be great to hear too. And then if AGI becomes a thing before a perfectly emulated hand does, just let it learn to produce piano sound virtually.

I think all of this will be possible a long time before it becomes reality though. There’s no money in it. :rage:


#32

hhahahah how iz diz different from da dizkklavier zheeyat

tru alzo da :robot: zenzory feeback iz prehiztoric compared to da human proprioception zyztem zo da :robot: wud have to rely on zound feedback only, a much mo hardah zheeyat


#33

I predick da nextgen captchah will be

“sightread diz bar” :sunglasses:


#34

It would be a mechanical system which can play the piano like Cziffra, if you’ve trained it to emulate a particular Ziff recording.

Ideally anyway. :lib: I guess it’d be like a modern AI powered successor to the piano roll.