mimi10v3

Notes / archive

Chronological archive

74,818 public notespage 286 of 749plain textJSON
  1. Oct 15, 2023Twitter · replydalle generated these for my name & profile minus one character in the username... suddenly frustrated I can't remove it and summon my catboy :)
  2. Oct 15, 2023Twitter · reply@liminaldoge ran into GPT rate limit - will try in a few hours tho :)
  3. Oct 15, 2023Twitter · replytried some other ppl's usernames for comparison :) https://t.co/db3UI50EXc
  4. Oct 15, 2023Twitter · reply@rageflowerrr yes, via ChatGPT rather than the Bing thing
  5. Oct 15, 2023Twitter · replymy username alone produces this? 🤔 or is GPT hiding the rest of the prompt? https://t.co/fh0F0mne5Y
  6. Oct 15, 2023Twitter · reply@neirenoir @Soul0Engineer @a_musingcat lmao 'musing is best kitten fr
  7. Oct 15, 2023Twitter · postme & who? https://t.co/8rUBWY629X
  8. Oct 15, 2023Twitter · replylol, quite consistent, even oddly insisting on the horizontal layout even with the vertical image? https://t.co/iLXMHNvR9b
  9. Oct 15, 2023Twitter · postlol, it me... but how did it *know*? https://t.co/MSOWvCdMCq
  10. Oct 15, 2023Twitter · replyhttps://t.co/rGzKru4loS
  11. Oct 15, 2023Twitter · reply@Mystisith here's the image prompt GPT made up: Photo of the aftermath of a Winter Trop performance, showcasing the beautiful patterns left on the snow by the dancers' footsteps, resembling a delicate snow tapestry.
  12. Oct 15, 2023Twitter · reply@202accepted no, but good idea I'll try that next! this is GPT's explanation: In the ethereal Winter Trop performances on the Altean Mountains, dancers move with such deliberation and grace that the snow beneath them becomes a canvas. Each step, each glide, leaves behind a mark, creating intricate patterns that tell the story of their dance. Over time,…
  13. Oct 15, 2023Twitter · reply@202accepted it is snow yes, but there's more to the prompt
  14. Oct 15, 2023Twitter · postdalle art - guess what this is :) https://t.co/gng3cTyQCL
  15. Oct 15, 2023Twitter · reply@getnormality oh, in the back and forth about fixing the previous images it did eventually seem to prompt with a celeb name https://t.co/yzVC9UHTI8
  16. Oct 15, 2023Twitter · reply@getnormality um depends - for Cleopatra it was willing but lol at the result... for current day names it says no... https://t.co/cOWRzN6ZJZ
  17. Oct 15, 2023Twitter · reply@GreyMindMatters @8teAPi lol I think the problem is the Dalle training data so skewed towards thin white models so it defaults to that unless prompted otherwise... and then they've done a poor job of telling GPT how to prompt otherwise. They DO seem to be making updates tho so hopefully they fix it
  18. Oct 15, 2023Twitter · reply@getnormality yeah, well actually I didn't choose the people I asked GPT to choose and then depict them. I did initially have to say it's ok to first choose the names and then do whatever you need to do to comply with your dalle tool policies, otherwise it refuses to choose specific names. https://t.co/57SsetI0Gp
  19. Oct 15, 2023Twitter · reply@losingcontrol23 yeah haha - he's also in one of the ones above - you can see I just ask it to choose celebs and then depict them, it chose Denzel and then created these images.
  20. Oct 15, 2023Twitter · reply@BeezyManzell @DistractedAnna I started with some questions about the tool prompt to see what it is told to do, then these two prompts, then the remaining ones were like "thanks, please generate more" - what prompt did you use? https://t.co/vhjybEALb4
  21. Oct 15, 2023Twitter · reply@GarrettPetersen @ALightcone This is using the Dall-E version of GPT-4, so it does the translation itself - the prompt requires it to replace celeb names with generic descriptions...
  22. Oct 15, 2023Twitter · reply@8teAPi I realize this - my point, further illustrated with 16 more generations in the reply - is that 100% of the resulting images are thin white people.
  23. Oct 15, 2023Twitter · replyfwiw this *can* be fixed by telling poor GPT what happened... maybe a better tool prompt would simply warn GPT of the tool's biases and suggest prompting otherwise if it wants to depict more diverse representative images. https://t.co/AxoDNVnGhl
  24. Oct 15, 2023Twitter · reply@ALightcone I feel bad for poor lil GPT, meaning so well and yet the Dalle tool produces only thin white ppl when not prompted otherwise
  25. Oct 15, 2023Twitter · replyMae Jemison, Thurgood Marshall, Misty Copeland, and Robert F. Smith Bessie Coleman, Katherine Johnson, Jean-Michel Basquiat, LeVar Burton Alvin Ailey, Gwendolyn Brooks, Shirley Chisholm, Jackie Robinson Lizzo, Mo'Nique, Gabourey Sidibe, Anthony Anderson https://t.co/bayahROUjm
  26. Oct 15, 2023Twitter · postDalle's depictions of Serena Williams, Neil deGrasse Tyson, Ursula Burns, and Benjamin O. Davis Jr. 🤣🙅‍♀️ https://t.co/qdfHeBEJSZ
  27. Oct 15, 2023Twitter · reply@ESYudkowsky 🤔 how would you want copilot assisted websites labeled?
  28. Oct 15, 2023Twitter · reply@DemonSynth lol midwit meme probably
  29. Oct 15, 2023Twitter · replyreminded me of this article... https://www.earth.com/news/as-bird-skeletons-become-more-complex-they-lose-diversity/
  30. Oct 15, 2023Twitter · reply@teortaxesTex what definition of "it thinks" or "world model" excludes GPT-4's abilities such as they are?
  31. Oct 15, 2023Twitter · reply@shakoistsLog big Reddit hall monitor energy around here lately fr https://t.co/L7IOC6v37x
  32. Oct 15, 2023Twitter · reply@Paleoguy073 @1pckt @MephistoTabern1 @Melanie33720945 @DarkCowboyGod @FifthDonkey @NHoomans @itisjustmebabe @megafant1 @edgy_knife @CarusoLawLV @1Quetzalcoatl1 @BrianNazarenus @LannisterLeo @ScienceQriosity @_truthswordsa @RebelATliance @SecularSmurf @ScientiaPercept @girmayyy @Galcondude @peterrevans16 @CubeSteve @Chortling_Man @TropicOfChaos @hubie0…
  33. Oct 15, 2023Twitter · quoteyou know it's gonna be good when it starts with a 2018 picture of a rainbow and quickly turns to this https://x.com/ScientiaPercept/status/1680016844290244609?s=20
  34. Oct 14, 2023Twitter · quote😻 🍿 I love a good "debate" thread of absurd length.... time to settle in for a good read https://x.com/Paleoguy073/status/1713392944978493578
  35. Oct 14, 2023Twitter · reply@MikePFrank @PipFoweraker @BrianJJi @peterthiel 🥺🫶
  36. Oct 14, 2023Twitter · postthe govt said next to nothing while the tabloids said every possible thing
  37. Oct 14, 2023Twitter · replysigh, reading Buddhist stuff and it feels like when you don't put the weight stack pin back in all the way and pull expecting heavy resistance and instead there's none... like, it's not having the right effect gah!
  38. Oct 14, 2023Twitter · reply@quantum_oasis sigh yeah perhaps, I've been struggling w ChatGPT abt various topics recently, trying to get an argument clear enough to write something longer than a tweet abt it
  39. Oct 14, 2023Twitter · reply@storebrandguy what do you mean by "fake"? it seems like an interesting way to think abt certain questions
  40. Oct 14, 2023Twitter · postInteresting.... I wish I were smarter https://www.nature.com/articles/s41586-023-06600-9
  41. Oct 14, 2023Twitter · reply@sayestheorem @parafactual part is resilience & valuing forest above desert because of the water cycle also biophilia- humans aesthetic enjoyment & well-being from contact w nature... replacing a complex ecosystem w monoculture & killing off species removes the possibility for future generations
  42. Oct 14, 2023Twitter · reply@sayestheorem @parafactual sigh for me veganism isn't only about suffering but also about land use & biodiversity, valuing intact ecosystems & wildlife corridors & areas w rare species / high biodiversity... in part for human welfare but also for their own sake
  43. Oct 14, 2023Twitter · reply@lumpenspace 🤣 I love the way you purr too
  44. Oct 14, 2023Twitter · reply@sayestheorem @parafactual what's the argument wrt New Zealand beef & lamb? just welfare standards and an argument that the wild native animals they displace by grazing would lead a worse life?
  45. Oct 14, 2023Twitter · reply@sayestheorem @parafactual at this point it has been so long that meat & dairy elicits disgust response not appetite for me, its literally zero sacrifice
  46. Oct 14, 2023Twitter · reply@sayestheorem @parafactual I don't budget that closely lol... but vegan food is generally cheaper than meat, so if you shift to vegan you have more money to give to charity while keeping everything else constant... also $100 seems like a very low number given the average consumption & farming practices including the extra acres of feed grown w pesticides…
  47. Oct 14, 2023Twitter · reply@toasterlighting @parafactual to me this isn't about animal activism necessarily so much as just opting out of foods that cause disproportionate suffering to produce... like, ppl can eat vegan-ish and also do other activism
  48. Oct 14, 2023Twitter · reply@sayestheorem @parafactual better assuming you can only do one or the other? I don't really understand the logic of this...
  49. Oct 14, 2023Twitter · reply@sayestheorem @parafactual what is the less expensive wild animal intervention we're missing that doesn't lead back to veganism?
  50. Oct 14, 2023Twitter · reply@s9chroma yeah true- obscure jokes can be funny and also helpfully noted to help everyone else understand them
  51. Oct 14, 2023Twitter · reply@will_from_mars_ yeah, or worship maybe?
  52. Oct 14, 2023Twitter · quote🫤 this isn't about distribution of things but distribution of observations.... https://x.com/langofmind/status/1713189992820879406
  53. Oct 14, 2023Twitter · reply@fuzzyunclear oh words like "reality" or "soul" or "emptiness"
  54. Oct 14, 2023Twitter · reply@lillybilly299 ofc the ratings are about guessing how many of their bros would say "would" vs not... all about sizing up the competition 🙄
  55. Oct 14, 2023Twitter · reply@nickcammarata @MasterTimBlais does the ferrule have a distinct scent to you? or is it paint vs pulp that's distinct?
  56. Oct 14, 2023Twitter · reply@nickcammarata unsure what you mean so I answered no, no... like is it a question of imagining an intellectual / emotional dot of selfness and reporting where it appears? or are you questioning if it already exists?
  57. Oct 14, 2023Twitter · postperiodic reminder that if you yank too hard on particular words when the beliefs under them aren't ready to change, sometimes the word comes loose from the concept instead... this is extremely unhelpful as you no longer have a linguistic handle for an important concept
  58. Oct 14, 2023Twitter · repostRT @doomed_lidar: Just when I thought beff couldn’t possibly be more based. Jazz is music for the free spirited. The thinkers, dreamers, a…Archive text may be incomplete
  59. Oct 14, 2023Twitter · reply@emnode when it's causing me pain
  60. Oct 14, 2023Twitter · reply@randomstring729 enjoy homecoming!
  61. Oct 14, 2023Twitter · reply@s9chroma maybe it's multiple strikes plus only after the Note is sufficiently rated as Helpful so Note writers themselves can't take down an account they just dislike / don't get the joke
  62. Oct 14, 2023Twitter · reply@Minus3Stdev lol, I'm probably on the other side of those opinions...
  63. Oct 14, 2023Twitter · reply@lumpenspace yeah lol I was reading Buddhist stuff again... always irks me when they play w language to create paradoxes and then say nondualism is the solution... like, the paradoxes don't bother me so I was thinking about what belief I have makes that the case
  64. Oct 14, 2023Twitter · reply@s9chroma yeah lol I'm frequently tempted to tweet virally wrong posts to elicit community note w correct info that I think should be more widely known
  65. Oct 14, 2023Twitter · reply@MailePRMedia and trolls and pranksters who have been treating Community Notes as merit badges lol
  66. Oct 14, 2023Twitter · quote👀 This is not a number to maximize, just fyi... https://x.com/MaileOnX/status/1713325805156352475
  67. Oct 14, 2023Twitter · reply@rarelyquinn imo pursuing consistency is most productive when the model is important enough for the effort, sufficiently parameterized, and has access to data or ability to collect data to fill in unknowns
  68. Oct 14, 2023Twitter · reply@garybasin @AndyXAndersen @ylecun otoh open source models have less attack surface- less appealing target for lawsuits, less need to protect a moat...
  69. Oct 14, 2023Twitter · reply@lumpenspace the problem is it's an unrealistic expectation, leads to unnecessary anger when we see the inconsistencies & demand they be reconciled, while mostly believing our own model is accurate
  70. Oct 14, 2023Twitter · postconsistency is overrated we're all continually building & revising our world model, a lossy abstraction on the complexities of the world, finding rules that work for predicting our day to day lives in different domains we find different rules fit best- and sometimes they're inconsistent, they often have exceptions and edge cases and discontinuities this…
  71. Oct 14, 2023Twitter · reply@vraiqx @___frye yeah - I think it's a fundamental problem to solve if you wanna build this - how to not have your initial users be disproportionately the most obnoxious disaffected abusive interfering hateful mob of ppl on the internet. on English-language bsky, it started out w techie insiders & scattered random users with nothing much to discuss, then…
  72. Oct 14, 2023Twitter · repostRT @elonmusk: @WallStreetSilv Stopping inflation requires government to stop increasing the quantity of money at a rate faster than the out…Archive text may be incomplete
  73. Oct 14, 2023Twitter · reply@mylordcod @BasedBeffJezos hahahaha, my evil plan to promote biophilic design is WORKING
  74. Oct 14, 2023Twitter · reply@bayeslord yep, love how OpenAI rebranded to pursuing AGI (which they're defining as an economically useful *tool* that can replace human labor) while obfuscating with memes and pursuing regulatory moat by cynical support of the Doomers.
  75. Oct 14, 2023Twitter · replysigh, it is really remarkably bad - Beyonce, Denzel, Frida, and Jackie... not to mention the thin bias 😑 https://t.co/xBJeMT2OLY
  76. Oct 14, 2023Twitter · reply@mylordcod @BasedBeffJezos I mean, with the right design to selectively buffer the intensity this could be amazing, but I assume they had other priorities & limited funds... also, my personal guess is that the group of ppl who chose to live there wasn't exactly the ideal team, psychologically.
  77. Oct 14, 2023Twitter · reply@mylordcod @BasedBeffJezos yeah lol... is interesting, in some ways it's representative of the challenges of a human mission to Mars - isolated small group, limited resources, long duration
  78. Oct 14, 2023Twitter · replyand given the original NONPROFIT goals of *OPEN* AI, share these diverse datasets for use by other AI teams with fewer resources to seek out additional novel data representing other languages & cultures & descents.
  79. Oct 14, 2023Twitter · replyimo they should trim back the tool prompt to focus on accurately executing user requests and work towards a more equitable & representative world through other means .... maybe spend some of their money on collecting additional data - text / images / voices from underrepresented communities across the globe?
  80. Oct 14, 2023Twitter · replyresulting in recognizable iconic black celebs being generically described and then the skewed makeup of the training data tending to create white images when identity / race / skin color is not specified.
  81. Oct 14, 2023Twitter · replyand then they've layered a bunch of questionable & contradictory rules for GPT to prompt dalle while meeting the DEI goals: "Diversify depictions of ALL images with people to include DESCENT and GENDER for EACH person using direct terms." "Your choices should be grounded in reality. For example, all of a given OCCUPATION should not be the same gender or…
  82. Oct 14, 2023Twitter · replythey're start with human-created images which represent the biases of past & present, as well as attention / selection effect of those with money & tech to represent their lived experience - so it's likely that without special effort the training accurately learns what's in the training data, not global demographic mix or how we *wish* occupations & roles…
  83. Oct 14, 2023Twitter · replylike, they've seriously Waluigied this one... https://t.co/S1MleaN8GG
  84. Oct 14, 2023Twitter · replyI *assume* OpenAI is not trying to turn amazingly talented black celebs into a bunch of white images, but like, did they test their prompt at all? https://t.co/GyP6agPZO7
  85. Oct 14, 2023Twitter · replyanother try - it selected Jimi Hendrix and Rita Moreno for the 2nd and 3rd pics... but lmao at the result. OpenAI is not sending their best to the dalle team! https://t.co/Rcnq7mvriH
  86. Oct 14, 2023Twitter · postkinda funny how the dalle tool prompt is constrained by the need to both read as reasonable & non-biased and yet also achieve the race & gender representation that OpenAI desires. I asked it to choose specific celebs to create and then do whatever is necessary to comply with the tool prompt... it chose Denzel & Will Smith on the right, but lol at the…
  87. Oct 14, 2023Twitter · reply@StevieKomputer 😊 very nice!!
  88. Oct 14, 2023Twitter · reply@inerati yep mine has been similar every time I've had it tested, veganism really does affect cholesterol levels, whatever that in turn may do 🤷‍♀️
  89. Oct 14, 2023Twitter · reply@BasedBaconator DoorDash! & relax!
  90. Oct 14, 2023Twitter · reply@aworldprincess 😞 sigh
  91. Oct 14, 2023Twitter · reply@MarkovMagnifico oh cool! make the physics fun- things bouncing & affected by gravity
  92. Oct 14, 2023Twitter · reply@DariasRucker how would you describe my flavor?
  93. Oct 14, 2023Twitter · quoteremembering back to early MidJourney... so much change & improvement since the early days, but even then the essence was there https://x.com/mimi10v3/status/1531888993595543554
  94. Oct 14, 2023Twitter · reply@main_horse @akbirthko Samantha is a nice AI gf tho, not a killer bit scenario... probably both cases were going by the same judgement that SJ has an appealing voice
  95. Oct 14, 2023Twitter · reply@accelerationem @bayeslord the 8x better reliability vs land DCs isn't selling you, given cost per DC employee on an offshore platform? vs likely worse than avg given sea air / iffy AC and storms... scale would amortize the cost of a tender drone sub for retrieving every 5 yrs for refurbishment...
  96. Oct 14, 2023Twitter · reply@accelerationem @bayeslord https://news.microsoft.com/source/features/sustainability/project-natick-underwater-datacenter/ I think there are arguments in favor of this plan, obvs you'd do many containers for redundancy & incremental replacement / expansion
  97. Oct 14, 2023Twitter · reply@BasedBeffJezos Also I want a lab for Cephalopod-Human-AI cooperative neuroscience
  98. Oct 14, 2023Twitter · reply@accelerationem @bayeslord sea air is so corrosive tho, and AC systems so finicky... better to seal entirely & replace as needed
  99. Oct 14, 2023Twitter · reply@BasedBeffJezos seriously yes; needs nuclear power for the seafloor servers, desalinization, DEW area defense, containerized 100% automated hydroponic fresh food, Starlink + undersea fiber connections, and manufacturing our own rogue chips -> selling black market unregistered GPUs
  100. Oct 14, 2023Twitter · reply@accelerationem @bayeslord might put the servers on the ocean below, the better for cooling...