Wikipedia:Articles for deletion/15.ai

Add links
From Wikipedia, the free encyclopedia
The following discussion is an archived debate of the proposed deletion of the article below. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.

The result was keep. Although not unanimous, consensus is that there are sufficient independent sources of sufficient significant coverage that WP:GNG is met. 78.26 (spin me / revolutions) 03:35, 30 January 2023 (UTC)[reply]

15.ai[edit]

[Hide this box] New to Articles for deletion (AfD)? Read these primers!

15.ai (edit | talk | history | protect | delete | links | watch | logs | views) – (View log | edits since nomination)
(Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL)

This is an anonymous creator's unfinished project at a dead website, where they collected all of the My Little Pony videos and ran them through open source voice cloning software. Their grandiose view of their own project doesn't mean that it's notable. This article is 49% original research, and 49% content that needs to be moved to the Audio_deepfake article, since it's so generalized. I'm very impressed at boundless energy this particular group of editors has towards curating this article, considering the topic. It ultimately doesn't belong on Wikipedia though. Habanero-tan (talk) 00:48, 16 January 2023 (UTC)[reply]

  • Delete Other than the IGN article and the Arxiv pre-prints, the rest aren't reliable as sources. I don't find any RS discussions we could use. Oaktree b (talk) 01:25, 16 January 2023 (UTC)[reply]
  • Keep Extremely important as it's one example of the first TTS voice generators and neural networks. I do not see why this is considered for deletion and it's currently under maintenance. There is also plenty of high quality citations. (talk) 05:16, 16 January 2023 (UTC) — Preceding unsigned comment added by Bloodclotboy12 (talk • contribs) [reply]
    "one example of the first TTS voice generators and neural networks"
    That is an extremely false claim, and I am sorry if this Patreon project mislead you into donating based on that. The 15.ai project was begun in 2020 as a labeled training set of My Little Pony clips. The project uses publically available TTS & neural network software. For the actual history of neural network TTS, which 15.ai has nothing to do with, see:
    - https://en.wikipedia.org/wiki/WaveNet#History
    - https://en.wikipedia.org/wiki/Speech_synthesis#History
    - https://en.wikipedia.org/wiki/Neural_network#History
    Habanero-tan (talk) 21:47, 21 January 2023 (UTC)[reply]
    At this point I have to assume you are doing this out of malice. You are, once again, regurgitating the same falsehood that others have called you out for: The project uses publically available TTS & neural network software, when this is clearly not the case, and are now actively trying to (and I don't use this term lightly) lie your way through this. Could an admin please do something about this? This is extremely concerning. Tacotron2 (talk) 23:28, 21 January 2023 (UTC)[reply]
  • Note: This discussion has been included in the deletion sorting lists for the following topics: Computing and Websites. –LaundryPizza03 (dc̄) 05:28, 16 January 2023 (UTC)[reply]
  • Delete, how the hell did this become a GA? LilianaUwU (talk / contribs) 13:27, 17 January 2023 (UTC)[reply]
    Pardon me, however isn't this more of a "Comment", rather than a reason for deletion?
    Could you please elaborate as to why you believe this article should be deleted. 2A00:23C4:6696:4801:8CDB:BA79:1349:F217 (talk) 15:07, 23 January 2023 (UTC)[reply]
  • Delete, although context from this article could probably make more general adjacent articles (such as Deepfake) into a GA article. Iazyges Consermonor Opus meum 14:22, 18 January 2023 (UTC)[reply]
  • Strong Keep. As one of the primary editors of the article in the past (now mostly inactive due to real-life things), the article from several months ago was most certainly up to GA standards. However, this article has constantly been plagued by anonymous editors and new users making unsubstantiated edits ever since the article was first created. I have since reverted the article back to what it was back in November. As for notability, the subjects absolutely meets the requisite standards, and as pointed above, was extremely crucial in the development of TTS voice generation. I apologize for being inactive as an editor for so long, but I did not expect the level of vandalism to reach this high, causing concern for the legitimacy of this article. —HackerKnownAs (talk) 19:36, 18 January 2023 (UTC)[reply]
    • I have since requested a higher level of protection for the article. The last request was accepted and lasted for a month, but it appears that this was not nearly strong enough. —HackerKnownAs (talk) 19:36, 18 January 2023 (UTC)[reply]
  • Keep The guy who runs the site goes on hiatus like every other month for improvements, this isn't anything new. Page was still up the last time it went under maintenance. — Preceding unsigned comment added by AveTrueToCaesar (talk • contribs) 20:11, 18 January 2023 (UTC)[reply]
  • Keep, articles like NovelAI and character.ai, which have far fewer reliable sources, are allowed to stay but somehow THIS gets nominated for deletion? Seriously? 63.139.68.87 (talk) 22:59, 18 January 2023 (UTC)[reply]
    The OP's claims are so replete with falsehoods and misinformation that it suggests to me that this AfD nomination was not made in good faith. The first two sentences are completely wrong—the project is not "unfinished" but rather constantly improving, the website is not "dead," the creator did not collect "My Little Pony videos," the voice cloning software is not "open source"—that I suspect that this was done fully on purpose. —HackerKnownAs (talk) 23:58, 18 January 2023 (UTC)[reply]
  • Keep It appears that the article has become somewhat bloated with both marginal content and marginal sources, and could do with a good spring cleaning. However, basic notability is not in question - these four mainstream game magazine articles [1][2][3][4] already form a sufficient backbone for that. --Elmidae (talk · contribs) 14:37, 19 January 2023 (UTC)[reply]
  • Keep Speedy keep, subject meets WP:GNG per Elmidae above. Tacotron2 (talk) 21:22, 22 January 2023 (UTC)[reply]
  • Comment. I'd like to point out: 1. almost everything stated in the OP is incorrect and the website isn't dead, as it's currently under maintenance (only for three months, mind you - hardly considered "dead"); and 2. WP:DEGRADE. This project's possibly the single most widely recognizable modern TTS project on the Internet (yes, even more recognizable than my namesake), and taking into consideration the inordinately numerous IP edit problem this article has always had, it's rash and unreasonable to judge the notability and quality of an article from these overzealous editors. I was most active in contributing to this article back in June of last year, so while I'm disappointed and rather annoyed by the sheer number of gung-ho editors who have no standards for quality control making all of these unsolicited edits, let me assure you that the article definitely did (and still does) exceed the notability and quality standards of Wikipedia. There are numerous secondary sources from reliable outlets like Game Informer, Polygon, IGN, and Kotaku, which are all listed under WP:RS.
>This is an anonymous creator's unfinished project at a dead website, where they collected all of the My Little Pony videos and ran them through open source voice cloning software. Their grandiose view of their own project doesn't mean that it's notable.
That's a very gross misrepresentation of just how significant the project was in the history of TTS research and I would advise editors to not take what is written in the OP at face value. I'm disappointed not only because such a blatant misrepresentation is being used to justify the deletion of this article, but also because some editors seem to have bought into OP's misleading and objectively false statements without checking them. Tacotron2 (talk) 17:50, 19 January 2023 (UTC)[reply]
I wasn't aware that WP:SK was an option when I made the above post, so I've upgraded my recommendation to that now (for the reasons listed in #1 and #3 of WP:SK). Sorry, I'm new to this. Tacotron2 (talk) 18:22, 19 January 2023 (UTC)[reply]
"how significant the project was in the history of TTS research"
I see the My Little Pony fandom is now glomming onto this AFD, after the initial string of Deletes. First of all, welcome. Regarding your core claim - since 15.ai is ultimately an Audio deepfake implementation (or as you're calling it, TTS, even though that's a much older technology) - why do 0 of the 52 references at Audio deepfake make any mention at all of 15.ai? Seems odd. I hope non-biased editors can judge for themselves which side of this discussion is doing the lying. But, most likely the only new comments here from here out will be from the MLP Fandom, as the string of Deletes came from folks visiting via the daily AFD log. I have nothing against My Little Pony or this deepfake dubbing project, I only take issue with articles full of false claims, original research, and unrelated citations that belong in a different article. Habanero-tan (talk) 22:10, 21 January 2023 (UTC)[reply]
What makes you think that I or any of the other commenters are from the My Little Pony fandom? (I'm not.) If you check my previous edits, it should've been obvious that I'm an actual deep learning engineer, as the vast majority of my edits are of deep learning articles (not to mention my username) and the only tangentially MLP-related article in my edit history is this very one (and I only contributed to the technical side of the article, not the fandom side of the article). It's extremely disingenuous of you to try and blame this backlash on brigading when 1. there is no evidence of such brigading happening at all (and the burden of proof is on you), and 2. your initial nomination was full of falsehoods that you still haven't acknowledged and are trying to brush under the carpet.
For example, as pointed out above: Why would you claim that the project is dead? Why would you claim that the project was run through open source cloning software? Why would you claim that the creator collected all of the My Little Pony videos, and why would you insinuate that the MLP aspect of the project is most paramount compared to the actual technology behind the project?
The reason I will continue supporting a speedy keep is due to WP:EARLY: The fifth bullet point states, Nominations which are so erroneous that they indicate that the nominator has not even read the article in question., which to me is obviously the case from the falsehoods mentioned above. The fact that you seem to be ignoring editors calling you out and instead are blaming these grievances on an entire fandom is truly telling. Tacotron2 (talk) 23:03, 21 January 2023 (UTC)[reply]
Again, I'm new to the AfD process, but the OP's accusation of brigading does not sit well with me. Could an admin clarify whether this is allowed or not? Tacotron2 (talk) 23:07, 21 January 2023 (UTC)[reply]
>or as you're calling it, TTS, even though that's a much older technology
As an ML engineer, I cannot even begin to explain how incorrect you are. Look up any "audio deepfake" implementation paper/repo, and I guarantee that you will see the term TTS being used in the title or description. I mean no offense, but claiming something like this shows just how blatantly ignorant you are of this topic. Here are three examples off the top of my head:
Tacotron2: https://arxiv.org/pdf/1712.05884.pdf
Glow-TTS: https://arxiv.org/pdf/2005.11129.pdf
Transformer-TTS: https://arxiv.org/pdf/1809.08895.pdf Tacotron2 (talk) 23:21, 21 January 2023 (UTC)[reply]
No. TTS stands for Text to Speech. It just means speech synthesis. It's not a specific implementation or technique. If you read the comment being directly replied to, it calls 15.ai "the first TTS". TTS is much older than that. Habanero-tan (talk) 02:37, 22 January 2023 (UTC)[reply]
What in the world are you talking about? You need to re-read the comment you directly replied to, because I never stated anything of the sort. I'm using the term "TTS" as it's commonly used in the field of ML-based TTS research. You were the one who claimed that 15.ai is not a TTS, when it by definition is. Tacotron2 (talk) 05:11, 22 January 2023 (UTC)[reply]
Please don't purposely strawman the wrong semantic sense into other folks comments, it only wastes time and energy. We both know what Bloodclotboy12 meant when they stated 15.ai is "one example of the first TTS". We both also know what the actual historical first TTS technologies were. Let's steelman and assume WP:GOODFAITH instead, and move back to the focus of improving Wikipedia. Habanero-tan (talk) 21:55, 22 January 2023 (UTC)[reply]
As much as I'd like to assume good faith on your end, it's exceedingly difficult for me to do so when you're still dodging the questions asked above. Could you explain how you came to the conclusion that 15.ai uses "open source voice cloning software"? Tacotron2 (talk) 22:50, 22 January 2023 (UTC)[reply]
information Administrator note Going to draw a line here, y'all. Saying the nominator is wrong is allowed. Arguing with passion that the nominator is wrong is allowed (although maybe not always the best idea). But personal attacks are not allowed, and further ones, by anyone, will be met with blocks from editing. -- Tamzin[cetacean needed] (she|they|xe) 23:49, 21 January 2023 (UTC)[reply]
The following discussion has been closed. Please do not modify it.
  • Strong Keep and ban Habanero-tan Every single claim this user has made about the project is blatantly false, and it's equally obvious that nominating it for deletion was an act of malice. Kyle horse (talk) 20:18, 20 January 2023 (UTC)[reply]
    This AFD was submitted out of a love for Wikipedia. I have zero malice towards the My Little Pony fandom. What I don't like is seeing articles made up entirely of WP:OR, WP:REL, & WP:RS violations. Full disclosure I have been working full time as a Deep Learning researcher since 2018, which includes TTS deepfakes, which is why I took such an astounded interest in this article's many unproven claims of significance. Habanero-tan (talk) 22:24, 21 January 2023 (UTC)[reply]
    Yeah you're very obviously just concern trolling. You have offered no proof for your claims, and you can't because you made them up. Get a life mate 2A02:A442:581E:1:7004:8562:4C6D:34B5 (talk) 23:01, 21 January 2023 (UTC)[reply]

Relisted to generate a more thorough discussion and clearer consensus.
Relisting comment: Relisting to get opinions on the state of the article after an editor's revert to the condition the page was in when it was designated a GA. I see there is room for improvement (and, at the least, protection) of this article but I'm very reluctant to delete an article that is a current GA. Just a note that those advocating Keep should cease attacking editors who see problems with this article. This is all part of the process of AFD when we consider how to handle articles that are seen as problematic. There is nothing nefarious going on, remember to have good faith towards other editors even when you disagree.

Just a reminder that a closer can close this whenever they deem there is a consensus. I just wanted to see if those advocating Delete had a change of mind with the revert of the article back to before vandalism occurred.
Please add new comments below this notice. Thanks, Liz Read! Talk! 02:17, 23 January 2023 (UTC)[reply]

  • I am leaning towards Delete. I 've read the article, quite interesting. But I failed to see RS dedicated on the topic of the article. May I ask, can anyone point to 3 solid RS (tweets, youtube, tiktok, about.com and most sources of the current version are not) that discuss 15.ai in depth? Thanks Cinadon36 08:36, 23 January 2023 (UTC)[reply]
    From the current version of the article:
    Game Informer: [5], PCGamer: [6], Kotaku: [7], Eurogamer: [8], Rock, Paper, Shotgun: [9]
    Game Informer is listed under WP:RSPSOURCES as a reliable source, and searching through the others in the noticeboard archives shows that these outlets have been used as reliable sources numerous times in the past, e.g. ([10], [11], [12], [13]). Tacotron2 (talk) 09:42, 23 January 2023 (UTC)[reply]
    Actually, I just came across WP:VG/RS, which lists Game Informer, PC Gamer, Kotaku, Eurogamer, and Rock Paper Shotgun all as solidly reliable sources. Tacotron2 (talk) 09:52, 23 January 2023 (UTC)[reply]
    Thanks for the reply. May I pose a couple of concerns more?
    • All of the articles cited are from January 21 (from 17th to 19th). Wikipedia:Notability requires that the attention by the world should be spread over a period of time.
    • Those articles are not an in depth presentation of 15.ai. They are pretty short and are sounding like "hey, look this weird app" Cinadon36 10:04, 23 January 2023 (UTC)[reply]
    • I believe you may have misread the dates -- the Eurogamer article was published on 2022 January 17, which was a year after the others. 15.ai most recently came under attention in early 2022 due to the Troy Baker NFT incident, which occurred over a year after 15.ai gained popularity in early 2021, so the subject isn't simply one of short-term interest.
    • As far as I'm aware, conciseness doesn't detract from these sources as having significant coverage (and in my humble opinion, the articles do a fine job of summarizing the important parts of 15.ai). The page for WP:Notability states that the sources should have significant coverage, be reliable, provide objective evidence of notability, and be independent of the subject, all of which seem to apply to the sources listed here. Tacotron2 (talk) 10:15, 23 January 2023 (UTC)[reply]
    • To give some more context, at the time of those initial articles in early 2021, no one had a clue where 15.ai had come from, as very little was actually known about the developer themselves. The website gave very little information about how the tool worked and there was no contact information, so it makes sense that those articles weren't able to talk much about the creator of the tool itself or how the TTS worked behind the scenes, but rather focused on the content created by the tool. Hope this clears things up, thank you. Tacotron2 (talk) 10:23, 23 January 2023 (UTC)[reply]
      Yes but. Article of Eurogamer was a year later, but the main issue was not 15.ai. Look, I understand this is borderline. They coverage is limited. For some, that s enough though. For me, it is a weak delete. Cinadon36 20:39, 23 January 2023 (UTC)[reply]
  • Keep per Elmidae and Tacotron2. Notability isn’t under question due to sources posted above that sufficiently meet WP:RS so I’m surprised to see discussion still going on about that. It’s one thing to scrutinize the quality of the article before or after the vandalism but to my knowledge that isn’t grounds for deletion. Full disclosure, I might be biased because I have heard of 15.ai before and only just found this AfD from the article itself. — Preceding unsigned comment added by Axiomofyourchoosing (talk • contribs) 11:17, 23 January 2023 (UTC)[reply]
  • Weak Keep. It's not 'extremely important'; but it seems well sourced enough for a non-contentious article. JeffUK 16:28, 23 January 2023 (UTC)[reply]
  • Keep, with the same points HackerKnownAs made above. 15.ai is constantly updated. While it is down currently, that isn't an uncommon state for the site to be in, and shouldn't be considered reason for deletion. Similar web projects are not available online, yet that does not discount their notability for an article. 15.ai is influential in fandom and AI culture, with coverage of itself and of creations made using it. —FrostyBeep 19:04, 23 January 2023 (UTC)[reply]
  • Comment (I already voted Strong Keep above, but would like to expound on my vote as I continue to clean up the article). There are at least four unchallenged sources from reliable organizations, plus a few others I have found on the article that contribute toward a solid GNG. The following is a summary of these sources:

Source assessment table: prepared by User:HackerKnownAs
Source Independent? Reliable? Significant coverage? Count source toward GNG?
[[14]] Yes Yes Game Informer is listed under WP:VG/RS Yes The source discusses the subject directly and in detail ✔ Yes
[[15]] Yes Yes PCGamer is listed under WP:VG/RS Yes The source discusses the subject directly and in detail ✔ Yes
[[16]] Yes Yes Kotaku is listed under WP:VG/RS Yes The source discusses the subject directly and in detail ✔ Yes
[[17]] Yes Yes Rock, Paper, Shotgun is listed under WP:VG/RS Yes The source discusses the subject directly and in detail ✔ Yes
[[18]] Yes Yes Eurogamer is listed under WP:VG/RS ~ The article mentions the subject briefly, but does not offer much detail ~ Partial
[[19]] Yes ~ While Yahoo! News is listed under WP:RSPSS, Yahoo! Finance is possibly less reliable; leaving as uncertain Yes The source discusses the subject directly and in detail ~ Partial
[[20]] Yes Yes Generally considered reliable; see [[21]] ~ The article mentions the subject briefly, but does not offer much detail ~ Partial
[[22]] Yes Yes Source is in Japanese; unclear whether the source (DenFamiNicoGamer) is reliable Appears to be reliable, as the website has its own editorial board, has been used unchallenged in other GA articles, and has been the source of exclusive interviews of celebrities and leaders in the gaming industry (see [23]) Yes The source discusses the subject directly and in detail ✔ Yes
[[24]] Yes Yes Source is in Japanese; unclear whether the source (Automaton Media) is reliable Appears to be reliable, as the website has its own editorial board, has been used unchallenged in other GA articles, and has been the source of exclusive interviews of celebrities and leaders in the gaming industry (see [25]) Yes The source discusses the subject directly and in detail ✔ Yes
[[26]] Yes ? Although LaPS4 has an article under the Spanish Wikipedia, unclear whether the source is reliable Yes The source discusses the subject directly and in detail ? Unknown
This table may not be a final or consensus view; it may summarize developing consensus, or reflect assessments of a single editor. Created using {{source assess table}}.
In addition, I have made a request that the article be placed under extended-confirmed protection, which has been accepted as of yesterday. Hopefully, this will keep some of the persistent disruptive editing at bay. —HackerKnownAs (talk) 20:12, 26 January 2023 (UTC)[reply]
Thank you, the table is quite impressive. But I am not sure that the sources discusses the subject in detail. For example, Game Informer's article is of 334 words (counted by https://wordcounter.net/). 'Rock paper shotgun is at 340 words. Kotaku is 216. PC Gamer at 362 words. So I am not very confided that is significant coverage. Cinadon36 20:46, 26 January 2023 (UTC)[reply]
Thank you for the reply; however, I must disagree with your assessment. According to WP:WHATSIGCOV and WP:100W, "sources with at least 100 words of coverage of a topic generally count" as part of significant coverage, a condition that all of the articles above satisfy. In addition, the definition of "significant coverage" as stated in WP:GNG has no mention of a minimum word count, and only stipulates that the sources must "address the topic directly and in detail, so that no original research is needed to extract the content." —HackerKnownAs (talk) 20:59, 26 January 2023 (UTC)[reply]
The 100 words bar is derived from an essay (Wikipedia:One hundred words), it is not a wp consensus. When reading those articles, nowhere did I felt I was reading a detailed account of 15.ai. Cinadon36 06:41, 27 January 2023 (UTC) Cinadon36 06:41, 27 January 2023 (UTC)[reply]
And similarly, that the 100 words guideline is derived from an essay does not mean that it is precluded from being a Wikipedia consensus; I don't see anywhere in the Wikipedia documentation that stipulates that sources with ~300 words are exempt from being meeting WP:SIGCOV, and I would consider that limit to be just as arbitrary as the 100 words limit in WP:100W. I regarded the articles as more than adequately meeting significant coverage of the subject, and I'm sure that many other editors would regard them as such as well. Your standards for what meets significant coverage might just be stricter than most, and that is completely fine—I simply disagree that an article must meet a minimum length in order for its subject to be considered noteworthy. —HackerKnownAs (talk) 11:59, 27 January 2023 (UTC)[reply]
Thank you for the polite discussion. Even though we disagree, I really liked the way you present your arguments and the table above. Cinadon36 14:12, 27 January 2023 (UTC)[reply]
It's no problem at all. Thank you for the civil discussion as well. —HackerKnownAs (talk) 16:29, 29 January 2023 (UTC)[reply]
Edited the table to include Japanese and Spanish sources, although it's still unclear to me whether these sources are reliable or not. Could someone who understands Japanese or Spanish chime in on the discussion? Thank you. —HackerKnownAs (talk) 21:01, 26 January 2023 (UTC)[reply]
  • Keep per table above, subject meets GNG. Also, shouldn’t an admin have closed this by now? It’s been well over a week since the AfD was created. Throwaway1112 (talk) 19:36, 27 January 2023 (UTC)[reply]
The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Wikipedia:Articles_for_deletion/15.ai&oldid=1136397339"