#Unicode
1 messages · Page 31 of 1
@chilly fulcrum ψ and ξ not in sheet still
ah i see it was in bot.. imma try & add it
Yeah I know I have loads to get around to
not as many active editors as there once was
As long as those are the lowercases and bot isn’t like being fooled
oh welp imma just continue filling nata's list
will continue later actually, imma eat
Idk I didn’t check if the bot recipes are legit but if they are it would help
if its within 2-5 step from my save
i would like to check em too
i wonder if we can scrape lineage of unicode known by bot
Anyone know how to do í ?
have you checked sheet
Yes
well check it again
me checking unicode of 11DC 11DD ... and ended up at 11FF
😭
Korean Alphabet
(for references)
Yuni
that is an unusual recipe did you just spam and happen to get that? I’ve had random things come with some random spam
i didnt spam lol...
i just curious what it will do with Y + Unifont since Unicode + Y = Yunicode
No I meant Korean alphabet
that’s an unusual one
Cyrillic Javanese is one of those things you get and then you get annoyed because it’s not what you want
true😂
You're telling me you don't want to see Javanese using Cyrillic orthograpy?
Actually Balinese would be worst
Ng
Expectation
Javanese Unicode
Reality :
understandable and have a nice day
new up recipe has dropped
Next Reverse : https://ib.zptr.cc/item/01hzkw6yrn7d502hy7d5dvp26n
OH nvm i can use simpler recipe
thanks for making Next Unicode Block
it was useful
¤ (U+00A4)
why did this unicode appear now :wondering:
i decided to get it now
i see, i'll leave that to other editor that has prepend U+00a4
and i dont have «
imagine normal element being able to unquote the "«"
does anyone know how to make U+420?
how to make "..u+"
"..u" + Prepend Hyphen + U+
like I tried so many things but just can't get it
Sorry, meant to ask is 0420 fine?
no, U+420
Ah, alright
ok nvm. Just got it U-420 + U+abc = U+420.
but really, I have been trying for like 20 minutes
this is TORTURE
What are you trying to get?
· (U+00B7)
that is a nice character
(ip grabber)
grab ipper*
Just randomly found U+0BE1 up to U+0BE6 (none of those are in the spreadsheet). Can someone please add the recipes? Also, the first five are unassigned so add that to the Missing Characters subsheet
U+1 + U+6b0 =
U+6b0 + =
U+6b0 + =
+ =
+ =
+ = ௦
U+626F
Delete The "e" + U+6ccf = 扯
i am fully convinced that getting · is literally just impossible
look at how many items with · i have
the AI always bullshits it's way into not giving me · somehow
heres the recipe for "·" if any of yall wanna get ·
§ (U+00A7) i dont have the stuff necessary to make efficient lineages lmao
oh yeah btw some samaritan found from U+0818 to U+082F and (U+0824, U+0825, U+0829, U+082A, U+082B)
and U+082E and U+082F are to be placed in Missing characters subsheet.
"fromcodepoint" + U+0b8 = ࠘
࠘ + ࠘ = ࠙
࠙ + ࠙ = ࠚ
࠙ + ࠚ = ࠛ
࠙ + ࠛ = ࠜ
ࠜ + ࠙ = ࠝ
ࠝ + ࠙ = ࠞ
ࠞ + ࠙ = ࠟ
ࠟ + ࠙ = ࠠ
ࠠ + ࠙ = ࠡ
ࠡ + ࠙ = ࠢ
ࠢ + ࠙ = ࠣ
ࠣ + ࠙ = ࠦ
ࠦ + ࠙ = ࠧ
ࠧ + ࠙ = ࠨ
ࠨ + ࠙ = ࠬ
ࠬ + ࠙ = ࠭
࠙ + ࠭ =
࠙ + =
࠙ + = ࠰
yeah about missing chars one day i think we can do that but for now not being able to add rows is annoying
hate to say it but missing aksara unicode 15.0
dont have "00";
????????????? what
american
also why is it putting random rtl seriously
is this some new behaviour
It's in the infinite craft browser lol
making a simpler lineage attempt for 00ab rn
i wonder how many characters are in the sheet but they dont know that theres a preceding rtl
got a lot of new unifont and unicodes with Mr. and also Mrs. but not many Mr and Mrs and no babys at all
damn all your hashtag stuff so many combos i couldnt keep track with all that on the board - like all the #current and all that woah
its just random spam - i guess i could lay out all my currents, nexts, previouses, unifonts, directly, etc
‰_‰
for reference
Aksara Unicode 15 + Unifont = Aksara Unicode 15.0
yeah
slightly optimised 00a7
question does mbs detect it?
it doesnt give me a codepoint so its a dead giveaway
prepend hypen & u+
this is indeed simpler
didnt know "the [ ]" was useful like that
also Aksara Unicode 15 + Unifont = Unicode
not Aksara Unicode 15.0
i see
yeah idk lots of options did you kinda combine lots of #current and #next and unicode, directly, unifont, etc etc
i got a few new ones
New Unifont for example
i was thinking if i can have previous & next
why not current too
then the grind is going long
hmm do you have any aksara unicode 1 - 9
i have unicode 0-9 nata made them
imma hope the lineage is reachable within extra 100+ step
killerbowser also made a simplified version
found it, yeah thats much better
does happen indeed
i had something like Current D, Current Dandelion smth smth
might as well i gave the save file for easier access if you want
im ok
imma continue [this](#1206592567622373446 message) again
which one? discord message links brokey
does reply broken?
no reply is not broken
oki
well... ok
oh wow
predictable tho
i give up for now
wow 33.5%
i wonder by the end of year, can we reach 50% or 75%😂
interesting never seen it just drop the quotes and the the - oh wait
unquote
thingy
yeah
@rubiksmath muted
Reason: Rule 2: No Spam. No repeated messages, inappropriate pings, mangled text, link spam, etc.
Duration: 59 minutes and 56 seconds
ransom sure is getting strange why do you need to sacrifice your mind to have rubiksmath unmuted‽
Unicode 2060
for references as well
impressive it does what i wanted
Next Reverse : https://ib.zptr.cc/item/01hzkw6yrn7d502hy7d5dvp26n
whew
I was asleep sorry
no worries
why your infinite craft looks cool
Unassigned codepoint
what does that mean
It's a Unicode point that doesn't have a character attached to it
Uh
Send recipes and a sheet editor'll see
ok
step 1: ask to make georgian board (there isn't one i think)
step 2: post there
step 3: send to unicode
step 4: abandon it
these are the first few
just add the one thats before it
more
then just add previous ones with others
same, i believe thats all the Mtavruli script characters
nope thats just half of it
i meant from U+10E6 onwards
sigh, check tokenizer
᳐᳐᳐
imma complete this code block as i can get the 1c90
yea georgian unicode + A
using that as entrance also works ! (ignoring it wont be giving U+1C90
Speedrun Mtavruli script?
nope just discovering the mtavruli script block
after so many fking month of dogra unicode hunt
Finally it appeared !!!
Ğ
holy dogra
Dogra codeblock
U+11800 - U+1183F
Aksarra Unicode 11 : #1206592567622373446 message
https://media.discordapp.net/attachments/1211396359836409936/1249407286028537886/image.png?ex=666730b5&is=6665df35&hm=a6486994b4d0b26c94b940b915ff1611fbeef0beb905a8eaf2f02b784df615b7&=&format=webp&quality=lossless&width=553&height=565
https://media.discordapp.net/attachments/1211396359836409936/1249407804813348997/image.png?ex=66673131&is=6665dfb1&hm=61bee0692df513e79b0b7f31844fa0da004a1c464514a717479f5a95e7f8e358&=&format=webp&quality=lossless&width=302&height=437
https://media.discordapp.net/attachments/1211396359836409936/1249408779993223169/image.png?ex=6667321a&is=6665e09a&hm=458766f19269d70396e0ffc085123856169475d14d2a29d34ef3b1449860ae59&=&format=webp&quality=lossless&width=501&height=565
https://media.discordapp.net/attachments/1211396359836409936/1249409185909706904/image.png?ex=6667327a&is=6665e0fa&hm=1eff85c11b63df1ff43ad7bae0378815e4d3cc70df62e8a70097cd20aae6e2f4&=&format=webp&quality=lossless&width=395&height=437
somehow there's one thats alr found :0
hmm the quality image... imma redo screenshot
have you tried using discord search to find the recipe?
#1226724619290148917
I did and it's still not helping
ok thanks
I just want a letter in quotes but it doesn't wanna work and nothing seems to do it
oh welp...
U+1cc0 : #1206592567622373446 message
Aksara Unicode 11 and 16 : #1206592567622373446 message
continuation
34% now :>
ok now I'm looking for U+1d0d and I have no idea how. I saw it's been gotten but idk how to get much past having the letters on the end
I have U+1d
ok nvm I got it now
got random two U+xxxx in one element
happens
idk how ot get unicode 11.0 tho
like unicode 11 + 11.0 is of no help
ah ok
using unicode 1 from unicode 10.0
wheee
anyone know how to get ṇ? On the sheet it says one of the components is next unicode but the other element is missing but it says it's been gotten
I see
which codepoint is that
I’ll have a look
hmm its not ticked on the sheet
the capital is though
and the recipe is ̇ + Next Unicode
where ̇ is combining dot above
(0307)
bruhhhh
why is getting weightlifting emoji so hard
it even gave me skin tone modifiers instead
would help if i could get the swimming emoji before it
anyone got the swimming emoji? i am not interested in following lineage for 1f3ca thats on the sheet
made all up to unicode 24.0
stuck there
the recipe for the swimming emoji says increment from 1fad7 to 1fae7... what incrementation tech was this?? it doesnt increment
hold on i m getting there
the model must have been so different back then oml
Elaborate
if you look at the recipe for the swimming emoji on the sheet - the manner in which it was obtained and the instructions given really seem to speak of a different time
if you try similar recipes
you dont really get anything near these it feels
but maybe it was just a one off random
for them too
but i need to get rid of that recipe and banish it from the sheet
its awful ive incremented up to 1fae2 now but im stuck - need 1fae7 to combine with whale emoji
and my "cheat code" for emoji incrementation may have broke since whatever happened last wekk
ey theres 1fae7
out of nowhere
must have found the recipe they sued
I have some thoughts about what this change means
I'll need to do a bit more research first but it might be something adjacent to quantisation without actual quantisation
cause i can tell you now i was not gonna get to that
can you link me something on that? seems interesting but its not something i know about
Which part?
quantisation and the effects
Uh...
Let me see
but we need to get rid of that god awful recipe
This might be a bit involved but it covers most of the necessary info
Ignore the second half about post-quantisation calibration, that's not relevant
okay - do you study this formally or is this just a hobby thing you taught yourself? my friend is doing like an ai major at uni
damn couldnt use my cheat code to increment to 1f3cb from 1f3ca
Hobbyist
My field is social studies, which includes sociolinguistics, and a research project I was doing pulled me into language models
"Social studies", hah, I'm such a pretentious ass
I do politics. There, that's better
well it seems like you learned a lot - i always find it a bit daunting to go in kinda blind - i do much better with like formal teaching and only then i could maybe do a little extension
I don't touch the actual code part of all of it, that's beyond me
But the theory all makes sense
yeah sure that sounds really cool if you have a starting point for me i will look into it one day
Starting point for language models?
Depends what part you want to look into to be honest
wait how do I get the N thing
N low dot?
you use the recipe which is Next Unicode + (combining dot above) (also in sheet)
its hard to see
but there is a charatcer there
its U+0307
So Rubiks slight problem with that
This is what's shown
But you're not allowed to copy-paste the characters to check what they are
thats annoying
is there an option to enable copy paste but still not be able to edit?
i can ask ray
Yes
Because that's called Viewing permission
This is something that isn't on by default
does that also enable file download?
Yes, though I'm not sure why that's a problem
for others to run code on
to check what they have that isnt in the sheet
its annoying to fork up the most recent copy each and every time
I may be misunderstanding you but what does that mean
"Fork up the most recent copy"
download sheets and send it to them
Oh
Yeah no it'll allow you to just do that yourself
Like I did when I made the Lookup Table
maybe we can ask the more knowledgeable programmers if theres a way to hook it up to a live database therefore dont have to download every time
okay
yeah did we have to send the files to u then? i forgor
interesting
if you know the exact settings you can put in the admin thread yourself but yeah thanks for bringing that to my attention
I have no idea
Regardless, my laundry's out of the oven now so I should sleep
Bonne nuit
okay
ideally we would just pull from mika's database but idk what is going on with that. is it still broken? and i'm not sure if we can have access to it
id not work
I'm not sure what to do now
try alot of tools, more than 50 i guess...
How about the recipe on the sheet for it
Oh nice how did you get it if it’s easier I won’t hesitate to put it in
In quotes may be tough
I will try tho
U+++1fa idk how u get them stuff all the time for the unicode utils but even now incrementing was nerfed you still can get the codes its cool
oh u mean this?
there's some i didnt able to get, i just skip them and do smth else for better time used
checkmate? got this instead
yeah but in general you find those really well i dont exploit the +++ enough
but i will be forced to now
because U+++1 just forces you into +++ land
oh no
99+ has returned
hangul 😬
i had many of those :>
oh yeah some are just this long
who's cooking again 👀
native editors for hangul/cjk wen
i cant read em... just gonna copy paste to check if i had em
yeah but like with the "words"
thats when it is beyond me
single character is ok
but like if it is the word for frog im not gonna know that
yeah i wont either :v
new update nerfed all my cheat codes
so cringe
1f3ca and b are now like ... impossibru
what kind of lineage is that 😔
I know I want to fix it
I didn’t put it there
it took me ages of guesswork to follow it
did you have Cyrillic to Unicode converter or any other converters like that
checking to element and found these
idk why im getting the Love Letter to Mr. Cyrillic Dogra either
out of all, i guess that one is the new convert stuff
well 🤣
An Input Tool has been implemented as a fast way to transfer recipes from console to sheet. If an editor wants to use it then DM me or request in #1217017057200050257.
nice
nice yeah some of your recipes from last few days i need to do - can you show me retroflex underdot and all those other things from last few days plz so i can add them
lemme make some room on my board
do you have mbs? it has pannable board
you can drag the board around and use empty area while used area will be dragged around without the element being deleted/smashed to one place
awesome
is that a feature?
do you have a list of the codes you made
I've never seen that in the options
yeah its part of mbs
ah cool
I just tried it I never knew
I don't think I have any codes that are new
i used it often when its a chain unicode
I've only wanted ones that people have gotten
mostly the letters with the dot below
ah well there have been some characters youve posted in the last week or two right?
I thinmk they're all already made though
It's not the codes that I had different it's the other element that I made that was different not the code
rubik, how would you do lineage for u+055a
note i alr revived it
for the character?
just the U+xxxx
i wouldnt have made one but now since update incrementing is hard i guess i may have to
the uppercase A is easy yeah
lowecase a however... fun
i forgot which one i used but its still U+++xxxx
umm
cursed
Yunikode
so messy one to have
i aint gonna use these tbh
Just give ME the goddamn U+056a already
finally
got U+056a0
trying to complete armenian code block rn, as expected it do a little trolling
i havent seen recipe of Armenian Unicode
where in the world did they hide it
ok it was just Armenian Script + Unicode = Armenian Unicode
why on earth aren’t we just using Chromatic tech for <country> Unicode and Unicode <country> elements, tf? 😅
you can try make them, so far i dont have any chromatic tech in my unicode save file
just simple .alphabet "alphabet" and some Mr. tech as well as #tech
maybe Mr. tech is just as good, idk because I don’t use it, but using recipes like Armenian Script + Unicode is WILD
AWWO would have our heads if they found out 🤣
i have some recipes for that
the lowercase?
it hasn’t been put in the sheet ofc it isn’t in the bot xD
nata if you’ve got lowercase xi Imma be a n g e r y
what emoji does it have
also nata perms were changed you can now download the sheets urself
no more needing to ask us
likewise if the bot has it I will be a n g e r y at the bot
so you can keep it up to date
because Georgian Unicode was also made like that
Georgian Script + Unicode = Georgian Unicode
thats why i kinda figured it out
idk just google Greek small letter xi unicode and paste it into your element search bar, sorry
it has the thinking emoji
yea i have it
nice
right, but that just pushes the fuckery down the line a little, to <country> Script
lmao so two people had lowercase phi for ages and just didn’t realise or tell anyone xD
pretty normal
meanwhile we have a bot that tells you your unicodes not in the sheet bahaha
like i guarantee the weightlifting emoji has been found
but it evades all my attempts especially since model changed
sure you can look at the type of recipes used on sheet - its a lot of random stuff since thats what you get thrown a lot but theres also a lot of name use
Emoji For
umlauts as well
for xi i have these
idk making something with the word emoji in it is hard
NOOOOOOOO
if only they all are willing to work :p
I thought I was the first discoverer of xi 😭
i do have current emoji
PLEASE don’t tell me you have uppercase omicron too nata
considering nata's bruteforce, he might had :v
039f?
ye
i made "prepend U+1f3cb" but i couldnt make it with anything out the front to remove
037f - 03ff is being crawled
it just give nothing every time if i try
did you get xi by crawling?
stupid update
yea
WOO
increased Nothing frequency so much
okay so I still have some claim to ξ yay
because if you have an element but don’t even realise you have it I don’t think that really counts as a “discovery” xD
cause it wasn't fd
well you can realise but you also can not realise its not in any db
ive only been checking for fds since no sheet
no sheet?
its not on sheet
ohh
okay well i mean only checking fd is easier for sure we need to ask star or someone to find out how to hook this to a live db
I mean the script is fine surely, no?
especially since now everyone can download the sheet
script needs to be in the pinned message tbh
its still cumbersome
how?
it would be much easier to just be able to make a request
instead of download every single time
I disagree
the code could make a request quite easily to the db
wonder if i can pull from sheet automatically now that it can be downloaded
yeah ofc your proposal would be better, but that’s doesn’t discredit how good the script is
yup i totally wanted that 😭
sigma
the script is fine it only needs a small tweak to connect to a db
i can even add such a method to do this in anticipation
I’m wondering why you said “any db”—we only have one comprehensive db, the sheet??
oh yea rubiks can you try this ich patch to see if middleclick still behaves inconsistently
rabbit experiment
well its not absolute
i still find things occasionally that are in other places but not this one
load time was like 5s
i thought you couldnt override it
yeah it works nice - rendering is gona slightly choppy but no way i can see for sure thats not a browser/pc thing - my setup does this thing a lot where if a make a new element and move the mouse quickly and its in 'choppy rendering mode' im gonna call it, it 'forgets' to render the new element till you move your mouse again
but i cant thanos snap glitch anymore
oh okay'
yes i totally wanted Armenian Caucasian codeblock
idk i havent been able to
what i did is basically making middleclick use the same function as doubleclick
okay just so I know, are there any other places besides the sheet, the bot, the browser, and discord search?
theres browser, theres also an emoji bounty which has some things too
and i guess if its a language there may be a bounty for that too
racism tech, LLMs have racial bias so it could work 👀
yeah it does this sometimes but when you initially pick up an element the rendering is not smooth
i can after the initial 0.3s lag move it around at 203948230984 kmh and it renders perfectly
im literally suffering yet progressing on different path than i originally wanted
it kinda in the first bit teleports to you mouse
yea it also does that over here
i should benchmark element dragging too
still "trivial"
I love when I go down an hour-long rabbit hole in Infinite Craft and when I finally get out of it I’m like … what tF was I doing? 🤣
setup is consistent now tho
then again I have cognitive memory deficits soo
yea there's a 1 second delay when picking elements
i see
yeah my mouse is on the far left here
it is forgetting to update position
it resumes remembering when i move mouse
seems to work best with things from sidebar
the delay is especially high
does it only sometimes
sometimes delay is not enough for it to forget
ok well yeah weightlifting emoji im sure exists
the swimming emoji recipe is hell
and update hasnt helped either
much better than me aiming to get x codeblock being done but got thrown into xy codeblock which is u+xxxxx instead...
thanks...
some day I’m going to optimise the Greek unicode block and it’s going to be GLORIOUS
woah actually big brain
thanks now I realise that there’s no reason for “ei” to make the sound “ay” in “weight”, I hate it 👍
U+056x be giving unicode but not its own code block....
( some does though )
you’d think with humanities collective intellect that there would be a common language with a phonetically consistent alphabet by now but NOPE
all we’ve got is IPA I think and that reads like hyroglyphics
we need to boycott the word ‘though’ Ive had ENOUGH
:<
and ‘queue’, like WHAT
you know today i realised how now we say it all the time eleven twelve thirteen fourteen fifteen... you dont even realise you are switching between two systems at 13-14
that has no reason to exist
but you dont even notice anymore
huh?
like
14-19
is fourteen
you then kinda die at 15
but then back to normal again
sixteen
seventeen
etc
tbf i forgot about 15 see it proves my point i didnt even notice
rubik, can u try the crazy tech to get unicode of u+0563
imma take a break for now
what crazy tech is this
anything thats not normal to use
damn gonna have to get used to lineage making tools in the new mbs i got
hmm ok
weightlifter vs everything with emoji in my save
damn so close
need Remove The Variants
it’s coz people used to count in 12s instead of 10s
so 11 and 12 were digits so they had their own names
where do you see this?
unicode analyser
any will do
i had a "cheat code" for emojis and was flying through till update
oh and of course this actually modifies the emoji of all things
plus or minus???? are you serious????
is blue weightlifting emoji what you were going for?
ah
what happens when you combine the red and the blue?
nothing useful
ooh wow
not a cutoff - just i guess the locale doesnt have this an emoji
like female modifier for a flexed bicep makes no sense here because its inherently genderless
but i guess how would it know
it just swapped them out
to it its just as normal as the other - but we see a messed up combo of emojis i guess
omg equals sign!!1!! (real)
michael phelps + unicode = new fd variant of swimming emoji
you have Emojicode but not Emoji Unicode or Unicode Emoji lol
uh huh
i dont know how to get them
they are used a lot
but i have no clue
its not easy to make
chromatic tech dude
emoji doesnt stick to stuff
I’m sure it does
chromatic tech is way more versatile than you’re giving it credit for
emoji and unicode are modifiers so yeah its hard - idk how to find such workarounds
new update no help either
"the Chromatic" + Emo + Emoji + Delete First Word works for "the Emoji"
I don't have "the Unicode" to check there
happily it makes this
Aha
hopefully Unicode Emoji is easier
Huh
When'd I get this
i doubt it personally - theres likely just good recipes for it
its an acient tech
i found a new recipe for swimming emoji
requires Swimator tho
which in turn required swimming emoji... id rather not build that
its not fd so there must be a decent recipe
it also requires unicode character finder (my tech i made on accident)
first time it ever did something
the lag is pretty bad since update
wont lie
I don't think that's an update thing
Lag is same for me
there are brief spurts of no recipe for 10-15 sec then all of a sudden it comes flooding in
its not like oh damn its down or anything its just every couple minutes new recipes take unusually long
Hm
Unrelated question, do you touch curlies often?
no
rarely
cumbersome to search for
also pain in the ass to unquote
generally just annoying
About that
unless it was used for something useful like
dogra unicode, Directly Next Unicode or stuff
I'm having some interesting experiences with curlies now
Things... fall out
i think you can exclude dogra unicode being weird behaviour suddenly appeared
cuz iirc this last few grind i kept getting send to u+xxxxx while doing u+xxxx
which was unicode of some aksara/language/similar stuff... which idk if its normal or not anymore
Other notes I'm taking include a tendency for non-quote words to no longer stick to quote words (in the vein of "word1" + Word 2 = "word1 word2" and rather spontaneous word compression
Problematically none of this is simply explained by any reasonable update, and I'd guess that around 75% is just people noticing things more because they think they have an explanation for it
I still can't remember what that cognitive bias is called.
Any help Gems?
Okay well I can tell you for sure new behaviour
now, U+xxxx + U+++1 has a tendency to equal U+++xxxx (or xxxx + 1)
This never happened before I swear not once
Nothing results also seem to be much more frequent
but can’t prove that
The second would be expected with quantisation
The former... eh?
oh yeah... it does that too, i thought i wasbeing unlucky
Right you sent something about that before a link but I had exam after and didn’t check oops
Ah
I mean to put it simply it's just making the numbers in the tokenisation weights less precise
I’d like to read though in what that precisely means and it’s effects but I may have to start from a more introductory point - it’s why I like courses they do just this
nothing feels worse than reading and realising you don’t understand anything - kinda like how I felt when I started this and had to madly catch up
I will read it though if you reckon I can take it
one other question - wtf even is a nothing
like from the model
what does that even mean
Either the actual word nothing or a failure to generate output, presumably
for a non general ai I’d imagine it’s like the outputs are way spread, but this is just from imagining those digit recognition ones where they output something for every digit
and the number is somewhat related to how sure it is that the number is that
so I was imagining ah the outputs must be just totally unsure but that isn’t how this works
Well devil's advocate, say it does reject any probability below some certain threshold, with enough spread that could happen
Well in this context it doesn’t make sense
This isn’t a digit recognition
Like what am I gonna do assign an output probability to every string possible?
Yes
Every token
I meant like all combinations
literally every string ever would need its own probability in my mind
But I guess token may make sense if it’s gradual
like iterative
what the fuck are you two talking about
As in it’s aware of the output it’s made thus far
You
I guess it has to be
otherwise how would it make coherent words
if it didn’t know what it just wrote
i would never be linked to this crazy rambling
im eating and im not enjoying it
So yeah, Niko works by looking at what has come before, assigning a weight to each token before through some attention mechanism, and using that to predict the next token
That's it
right so if the probabilities are just really whack
and if there was a threshold where if the maximum was below a certain amount it just refuses
There is also another potential issue with quantisation, in that, if done improperly, you can end up with tokenisation weights that are NaNs
Bad normalisation
In making the values less precise, some of the higher values get floating pointed to infinity
Floating point can go die
Then you divide them out to get them as probabilities and whoopsie that's not a valid operation
Yeah I see
so that would end up in like an illegal token
where you read that btw - anywhere specific or just stuff you’ve picked up over time
PB found that out after we tried making a homemade Infinite Craft as an experiment
Well we didn't get very far because of said problem
He downloaded the weights, put them in something (can't remember) and then tested out summing the token vectorisations
It sorta worked, but also threw up a lot of nonsense
code wise I guess too that’s a lot lol
Wait shit this isn't the Lab
Real
I was wondering why you were the only one here
i thought you would've realised this sooner
yeah I’m just coping with not being able to make inroads on 1f3ca/cb
apart from swimator + Unicode character finder
Look when people ask technical things my brain falls into a ravine and doesn't come back
True same
Forgot to mention but Johny found a few Unicodes during dead tests
with Nothing
so like Nothing + U+xxxx = character and it’s the only known recipe
lmao
Powerful
that's why i test with ?
i trash the output though so i wouldn't know if i find a new unicode with that
that's probably not a good idea but too late
oh you don’t save
No! = sign! How could you
i think it'd be funny if the found unicodes had no other known recipes even months later so the only way to get it is to cheat
Technically we could just get nothing
what about NOTHING
right if all caps is found why is laurasia only saying lowercase .. seems discriminatory
I hate capitalists
that's why you're lower class
this is funny
anyone knows the recipe for U+867?
the codepoint?
no, exactly what I wrote
U+867? then!
I haven’t done much with 3 digits
because I am starting to believe it is not possible
Unfortunately it seems incrementing got harder
Jenny?
sorry I am not familiar lol
neither was I. Just search 867 jenny
What
Ended up with this trying to get it
What is that ew
String.prototype.reverse() + U+0786 = ⵆ
Tifinagh Letter Tuareg Yakh
maybe with U+0768, that is actually reverse
Loss of reliable incrementstion hurts to bad
yeah
Let me check codepoints also jesus christ it is pouring out there
I see no relation
2D46
I like heavy rain - later I could tell you all about 24-27 feb 2022
like for some reason this can easily be made, but just U+867 not 😦
it's not too bad
thanks. I didn't think of using mr.
yea i have fd on that mr
I have his family 🔫
Oh
adopted
ok that is weird... cant get u+867 the normal way 😵💫
Unicode Character 🧿 (U+1F9FF)
apparently Î is missing
https://ib.zptr.cc/item/01j01ccwn3jj0bzy7fb4frcb2s here is the way i got it
i find Unicode Lepcha interesting
U+0550 + Unicode Lepcha = 𑜐
which is U+11710
idk the old behaviour but now its useful to get random unicode... until i understand whats going on
Wow very efficient lineage no string no U+ like you barely have unicode itself by the time you get it
oh yeah Unicode lepcha finally giving its unicode too... when was Unicode Lepcha made...
who the hell get u+0501 but not u+0500
Nvm it trolls
after abit of crafting finally i got the U+11680
+0.7% progress since last time i checked... not bad now imma rest
Unicode Character Ú (U+00DA) #1248708518609944616 message for lowercase in quotes
id67k what korean letter this is
Check soncole
would you look at that , all 3 letter combos are done
and this is the 1 character “combos” project 😂
maybe if we restrict to all letters from all alphabets that’d be a nice manageable subgoal 🤔
selective 1 character
if we include unassigned character
consider the project's goal multiplied
I’m still not convinced this project (assigneds) is even possible, but time will tell
I don’t think we can expect llama2 to know anything at all about the very most obscure unicode characters 🤷♂️
tbh we alr got some impossible ( plus banned ) unicode to have...
imma still do per block for consistency run ( and getting thrown to other block as usual )
that… is contradictory 😅
still looking for emoji unicode and unicode emoji
was hoping they were old enough to be in bot
obv not
I tried and I think the update nerfed Chromatic tech, I hadn’t played in like a week ;-;
unintentional nerf, obviously
if you decide to use Chromatic-like type tech then it looks like the best way to go about it would be "the X Unicode" + Emoji + Delete The X
for some X
for example I have Delete The Aged and stuff
you can usually get new Delete The X elements by just chucking X at all your existing Delete The Y elements
likewise it would be "the X Emoji" + Unicode + Delete The X for some X
yeah sadly the tech is nerfed for that it adds words far less often
okay i have the element G with breve
can i use it like that guy did tho
idk id need like hungarian breve accent or something
you have awoken me
seems normal
pretty sure i have that
it was just U+30FB + Middle Dot
idk
whenever i see c replaced by k in the tools i shared....
brain doesnt like it but its funny
🎉 post this in https://discord.com/channels/1203527044957208596/1208450449636855848 too
"𝐀lphabet" when?
fffd
ok so how to i make this into U+1f5e3
U+, Mr/mrs u+
thanks
I recommend combining it with U+, U+0, U+1, U+2, U-1, U-, U-0, Prepend U+0, Prepend U+0000, Mr./Mrs. U, Mr./Mrs. U+, Mr./Mrs. U+0, Prepend Plus, Prepend Plus Sign
usually some of them work. obviously you need to remove the Mr. Mrs. if you go that way
Ok so... I found and/or revived every U+xxx element.
invert fromcharcode is not something i have
its an old tool kali found
its #invert + fromcharcode
and then you coulg get this too
Just made them, yeah
cool
idk if next ascii stops
also I could share the recipes for all of them, but only in the form of a text file containing all of my recipes
next ascii? thats new
or if anyone would like my savefile or would like to run it through a program or smth I can provide stuff
did you note them down or run code through your file
I just write down all of my recipes for discoveries. didn't run any code through it
whew
anyone tried getting angzarr yet?
i got fd for the word so idk
"an"+"g"+"za"+"rr"+"rr"+ delete last character+delete the hyphen
pov you tryna make unicode
What
nah what are you doing
How does that even occur
two pluses, a space, and a .1
I could see the .1 and maybe 3 pluses
but the space… gg got me there
U++ isn’t what I try normally
uhh, there are a bunch of codepoints missing on the sheet
nvm found them im just dumb they were organized separately
car
Ca
Could there be like a row in place of the removed ones to explain like which ones got removed?
makes it more obvious
unassigned unicode doesnt have its own list as its too much, only some were recorded
for other... its probs in different sheet
yeah but if you get a unicode randomly you gotta check every tab (and some
times it takes a while to load the one you want in a tab)
Ü = ALT + Ui found something kinda random
and Alt + Ü = ü lol
Alt + U = Ü
Alt + Y = ¥
Alt + A = Æ
yes oddly Alt does some stuff when you throw letters at it
who got lazy
i'll setup sheet check and give a big file tomorrow cause i need to sleep now (4am moment)
mongus
hmmm thats a AFWWM request rn
literally it is one on the list
the among us unicode?
and the one that i replied
i'm trying to find a recipe for it
do you have edit access on the sheet
if you do can you remove Ü
no but let me try
input tool is so op i used it to input all the recipes from the 2 alphabets you got @thin warren all at once 😮
oh nice
yeah it is very cool
i saw you resolved them all pretty quickly
the most time consuming part was resolving the comments lmao
not putting them in or anything
i guess making the elements took more time but thats ok
but yeah it lets you copy paste direct from console
all recent recipes since update
it just gives poop emoji charcode now
for some reason
it used ot just give poop emoji
Why would first 2 give combinations give this result
not really much reason - but ther thing is if you are dealing with anything U+ it very often gives like the laughing emoji
and also the poop emoji
i guess this is just another way to say "poop emoji"
Tbh i dont remember getting poop emojies next to my u+
not next to the U+ i mean like the actual poop emoji character
poop
which by the way is in fact U+1f4a9
I see
yeah same thing happened
it smallcaps
i was trying to find a recipe for it because it doesn't have one
the ai doesnt like this bullshit 😮
what did neal cook again
next its gonna start giving the UTF-8 of the poop emoji
you mean fullwidth
okay
okay time to do the middle dot recipe..
@thin warren you got good recipe for middle dot?
i am confused
"the Middle" = not fd
"the Middle" + dot or any of the U+0020 tools = fd
"the chromatic middle dot" = fd
"the middle dot" = not fd
what
how
"the Middle" + "the Dot"?
= "the Dot in the Middle"
real
maybe string.append(' ') i heard thats decently good
and yeah if you add that to itself you get a double spaced element
string.append(' ')
goofy
made weightlifter emoji only for it to not help at all lmoa
anyone willing to try like normal elements and words + unicode phonetic
see if anything pops up
Motorcyle + Unicode Phonetic = 1f3cd new not on sheet as an excample
define normal
something youd expect a normal player might find i guess? but dont feel too restricted
nouns or adjectives
id say mostly
ngl why is my U+ or - sometimes doing things
woah
current unifont
future unifont
holy hell
clutch??
they both took down 1f3d7
ineffective against 1f394 and 1f3ca cb
1f3d5 has a recipe which is incorrect
it produces 1f3d6
fixed it with future unifont
the person must have used Chromatic-like tech but not Chromatic itself
maybe Aged
or Append or Prepend
or “the Middle” + D + Dot or “the Middle” + Do + Dot I suppose
there’s always more than one way to skin a cat xD
yeah idk names just dont help unless you arent me it seems - doesnt really matter what i do if i write out the name of the emoji it no worky
lol
alright man
😬 biden you better run
emoji desciptifier
who tf hates directly next to the beach ._.