|
The frustrating thing about transcripts is that I don't think it'll be possible with these tools to get the network to understand things like "be consistent about who's speaking" (e.g. no "KIRK: Spock here."). Have started generating "would you rather" lists and it seems a little more in word-rnn's wheelhouse: Would you rather a wizard ? or a massive wizard ? Would you rather adult yourself or fear a job you do with no high Would you rather an ninja ring or a lincoln Would you rather apartment in coffee . or live in space . Would you rather basically unlimited important yelling . or goat can do Would you rather be a nice eagle or be a zombie Would you rather be able to summon anyone with the same snail or be able to eat the power to change the sex Would you rather be boy from spongebob or be killed in a generation Would you rather be in the future or have one as your brother Would you rather butter pie ? or tattoo ? Would you rather do a chocolate nachos or eat right beef Would you rather fire or australia Would you rather galactus or fire rhino Would you rather have no madrid or never pancakes Would you rather have spongebob or have cyborg michael jackson Would you rather ipod burrito or sound ghosts Would you rather live or live in a helmet Would you rather minecraft or dubstep Would you rather mitt romney or mitt obama Would you rather punch your anus or become your mom
|
# ? Jun 9, 2016 14:21 |
|
|
# ? May 15, 2024 04:00 |
|
lolling irl at some of those questions, especially the last one
|
# ? Jun 9, 2016 14:38 |
|
lmao that last one
|
# ? Jun 9, 2016 14:45 |
|
Captain Foo posted:http://arstechnica.com/the-multiverse/2016/06/an-ai-wrote-this-movie-and-its-strangely-moving/ this is pretty funny but also really exposes the limitations of generating scripts like this
|
# ? Jun 9, 2016 17:26 |
|
Snapchat A Titty posted:this is pretty funny but also really exposes the limitations of generating scripts like this yeah, to produce coherent screenplays requires more context than today's rnns can get from just looking at other scripts - it's not like Go where all the data you need to succeed is there in the game board :/
|
# ? Jun 9, 2016 17:31 |
|
twit-posting every 3 hours:
|
# ? Jun 23, 2016 16:15 |
|
Crusader posted:twit-posting every 3 hours: groovy. can you make it so that it makes each tweet a poll? that would be sw8
|
# ? Jun 24, 2016 02:35 |
|
https://twitter.com/WYR_bot/status/746085596964265984 lmbo
|
# ? Jun 24, 2016 08:11 |
|
lolll
|
# ? Jun 24, 2016 12:58 |
|
cheese-cube posted:groovy. can you make it so that it makes each tweet a poll? that would be sw8 yes please
|
# ? Jun 26, 2016 06:48 |
|
i guess polls still aren't supported in twitter's api? https://twittercommunity.com/t/poll-support-for-rest-api/54775/34 quote:Feb 24 if anyone has a way to do it (i guess there's always browser automation) let me know though
|
# ? Jun 26, 2016 22:46 |
|
Crusader posted:In fact, I have no orders to do you.
|
# ? Jun 27, 2016 08:42 |
|
|
# ? Jun 28, 2016 03:40 |
|
markov polov game is still very strong https://twitter.com/markov_polov/status/747813919742275587
|
# ? Jun 28, 2016 16:37 |
|
i love that bitcoin news just automatically likes any post with "bitcoin" in it, including https://twitter.com/markov_polov/status/751420605400952832
|
# ? Jul 8, 2016 15:25 |
|
Trig Discipline posted:i love that bitcoin news just automatically likes any post with "bitcoin" in it, including lmao
|
# ? Jul 8, 2016 15:32 |
|
heh https://twitter.com/NewFintech/status/753575212864638976
|
# ? Jul 14, 2016 14:56 |
|
lol MP got a bunch of love for https://twitter.com/markov_polov/status/753523252362739712
|
# ? Jul 14, 2016 16:47 |
|
NoneMoreNegative posted:gay as gently caress in every device, every country, every amount. please don't tweet the `objective` from my resume
|
# ? Jul 14, 2016 17:00 |
|
lol the tagline for iPhone7 has been leaked
|
# ? Jul 14, 2016 17:05 |
|
oh god dammit, polov watched Swingers https://twitter.com/markov_polov/status/753702082046066688
|
# ? Jul 15, 2016 00:36 |
|
wyr bot is killing it, good poo poo crusader. if twitter ever make posting polls available in the API then it would be legendary
|
# ? Jul 15, 2016 02:11 |
|
cheese-cube posted:wyr bot is killing it, good poo poo crusader. if twitter ever make posting polls available in the API then it would be legendary thanks! i'll be re-training it soon; the last run still only had a < 2MB corpus, and I think I can cobble together additional material for it to learn more interesting comparison components from.
|
# ? Jul 15, 2016 15:55 |
|
cheese-cube posted:wyr bot is killing it, good poo poo crusader. if twitter ever make posting polls available in the API then it would be legendary quotin because i agree
|
# ? Jul 15, 2016 22:41 |
|
yeah wyr bot is amazing i do love it when my offspring call out random strangers with complete nonsense
|
# ? Jul 16, 2016 06:42 |
|
Trig Discipline posted:yeah wyr bot is amazing Markov polov flat out told me killing Jews is good
|
# ? Jul 16, 2016 19:30 |
|
yeah it took about an hour for poliovirus to start being super racist after being released into the wild
|
# ? Jul 16, 2016 20:13 |
|
^^^ well that's an interesting autocorrect https://twitter.com/markov_polov/status/757413791441129472
|
# ? Jul 25, 2016 06:01 |
|
https://twitter.com/markov_polov/status/759058122757386240
|
# ? Jul 29, 2016 17:11 |
|
awwww just like his daddy
|
# ? Jul 29, 2016 22:36 |
|
hey trig i think some dingus is following markov polov because it occasionally spits our dumb tweets like this: https://twitter.com/markov_polov/status/760007020762103808 https://twitter.com/markov_polov/status/759734442810224640 https://twitter.com/markov_polov/status/759731911769980929 any way to filter out the dingus? e: those tweets aside polov continues to be a source of amazement.
|
# ? Aug 1, 2016 11:44 |
|
idk i actually think those are pretty fun because random people get confused and try to engage polov in conversation sometimes also
|
# ? Aug 1, 2016 22:16 |
|
has the account received any DMs. if so them
|
# ? Aug 2, 2016 01:15 |
|
nope just replies on the tweets
|
# ? Aug 2, 2016 01:29 |
|
Read through the first few pages and skipped ahead. I made something similar to this decades ago for IRC and wanted to share an idea with you. The goal of mine was to learn from input on IRC and try to retain some sort of context in its replies. Probably one big difference is the vocab structure, as it could find chains in either direction along the sentence. When reading a line of text, it would find the rarest known word in the input and mark that as the "context" word. It would then find two paths from the context word to the beginning and to the end of a sentence. Then, if possible, find a third path that connects the words on either side of the "context" word to each other (using something similar to a* to keep it short) and swap the context word out for that path which is often a single word. This makes for a lot of uncanny or even spooky results when people engage with it. eg: on this input "Some words with context in the middle." it does this internally : Taken < out < of <- context -> is > everything. and the reply: "Taken out of something else is everything." obv there are similar rules when the context word appears at the beginning or end. Might have to tweak it for the twitter character limit but considering you'd already have to write the short path finder for the context bridging words, that should be pretty easy.
|
# ? Aug 3, 2016 13:58 |
|
that sounds pretty cool, please report back once the bot is live
|
# ? Aug 3, 2016 16:52 |
|
cheese-cube posted:that sounds pretty cool, please report back once the bot is live
|
# ? Aug 3, 2016 17:46 |
|
the devil's tweet
|
# ? Aug 9, 2016 22:57 |
|
Crusader posted:the devil's tweet aaaaaaahhhhh
|
# ? Aug 9, 2016 23:04 |
|
|
# ? May 15, 2024 04:00 |
|
some clever fucker figured out how to tell which tweets were trump's and which were from a staffer. other than a bot that just tweets back "trump" or "staffer", i can't think of anything fun to do with this info, but it's drat interesting maybe i should tune trumpbot to only use authentic trump tweets?
|
# ? Aug 10, 2016 08:23 |