User talk:Rhg2

Hello, Rhg2, and Welcome to Wikipedia!

Thank you for your contributions to this free encyclopedia. If you decide that you need help, check out Getting Help below, ask at the help desk, or place {{Help me}} on your talk page and ask your question there. Please remember to sign your name on talk pages by clicking or by typing four tildes (~~~~); this will automatically produce your username and the date. Also, please do your best to always fill in the edit summary field. Below are some useful links to help you get started. Happy editing! KylieTastic (talk) 21:55, 12 June 2025 (UTC)[reply]

Thank you KylieTastic ... I've tried putting a question on my Talk page with the "Help me" Rhg2 (talk) 23:44, 12 June 2025 (UTC)[reply]
You did it correct, it can just take a while for someone to answer (depending on the question). Remember, everyone is just a volunteer editor like you. Also we live around the world so all edit at different times. Cheers KylieTastic (talk) 09:04, 13 June 2025 (UTC)[reply]

Your submission at Articles for creation: Untether AI (June 12)

[edit]
Your recent article submission to Articles for Creation has been reviewed. Unfortunately, it has not been accepted at this time. The reason left by CF-501 Falcon was:
This draft's references do not show that the subject qualifies for a Wikipedia article. In summary, the draft needs multiple published sources that are:
Make sure you add references that meet all four of these criteria before resubmitting. Learn about mistakes to avoid when addressing this issue. If no additional references exist, the subject is not suitable for Wikipedia.
Please check the submission for any additional comments left by the reviewer. You are encouraged to edit the submission to address the issues raised and resubmit after they have been resolved.
CF-501 Falcon (talk · contribs) 01:35, 12 June 2025 (UTC)[reply]
Teahouse logo
Hello, Rhg2! Having an article draft declined at Articles for Creation can be disappointing. If you are wondering why your article submission was declined, please post a question at the Articles for creation help desk. If you have any other questions about your editing experience, we'd love to help you at the Teahouse, a friendly space on Wikipedia where experienced editors lend a hand to help new editors like yourself! See you there! CF-501 Falcon (talk · contribs) 01:35, 12 June 2025 (UTC)[reply]

Your submission at Articles for creation: Untether AI (June 12)

[edit]
Your recent article submission to Articles for Creation has been reviewed. Unfortunately, it has not been accepted at this time. The reason left by CanonNi was:
This draft's references do not show that the subject qualifies for a Wikipedia article. In summary, the draft needs multiple published sources that are:
Make sure you add references that meet all four of these criteria before resubmitting. Learn about mistakes to avoid when addressing this issue. If no additional references exist, the subject is not suitable for Wikipedia.
Please check the submission for any additional comments left by the reviewer. You are encouraged to edit the submission to address the issues raised and resubmit after they have been resolved.
'''[[User:CanonNi]]''' (talk • contribs) 07:52, 12 June 2025 (UTC)[reply]

Your submission at Articles for creation: Untether AI (June 12)

[edit]
Your recent article submission to Articles for Creation has been reviewed. Unfortunately, it has not been accepted at this time. The reasons left by KylieTastic were:
This submission is not adequately supported by reliable sources. Reliable sources are required so that information can be verified. If you need help with referencing, please see Referencing for beginners and Citing sources.
This draft's references do not show that the subject qualifies for a Wikipedia article. In summary, the draft needs multiple published sources that are:
Make sure you add references that meet these criteria before resubmitting. Learn about mistakes to avoid when addressing this issue. If no additional references exist, the subject is not suitable for Wikipedia.
Please check the submission for any additional comments left by the reviewer. You are encouraged to edit the submission to address the issues raised and resubmit after they have been resolved.
KylieTastic (talk) 21:55, 12 June 2025 (UTC)[reply]

thread on citations help

[edit]

I thought I'd try to start a Wikipedia page for the (what I think is interesting) story of Untether AI (Draft:Untether AI. I'm very new to Wikipedia and just getting used to how things work and had missed how careful I needed to be with citations at first. So after my first submission came back because of improper citations, I tried adding better citations (like an IEEE symposium). When that came back, I thought I'd be more careful about titles, publishers and names (i.e. I found I can use first2 and last2 to name a second author). So now I think I've got some pretty good independent sources, and I think I've got the format in pretty good shape. But still the citations aren't up to the mark. Would you have any suggestions as to some next steps?


Rhg2 (talk) 23:40, 12 June 2025 (UTC)[reply]

The software is complaining about multiple |date= parameters in one (or possibly more) of your citation templates. The references generally are working, technically, but I'm not sure you have the kind of sources that can be used to establish notability for companies according to WP:NCOMP. You want to avoid press releases and things that are considered 'routine' coverage. The final press release talking about the shutdown should stay, but it doesn't help for notability. — jmcgnh(talk) (contribs) 01:06, 13 June 2025 (UTC)[reply]
Rhg2, the key to any article is first to show notability as the declines notices describe in a set of bullet points. All new articles on Wikipedia have to show the subject is notable (See WP:N) which in most cases requires significant coverage (WP:SIGCOV) in multiple independent (WP:INDY) reliable sources (WP:RS). You did a good job regards the citations, but although important for polished articles, format is secondary until notability is shown. So although you have 7 sources they do not show notability: [1] is IEEE so reliable publisher, but not independent as from Untether Ai; [2] is run of the mill funding reporting, also the way it is written shows most of the info is direct from Untether directly. [3] and [4] are fine but based on the undying MLPerf Benchmarks. [5] and [6] are PR, so not independent so ignored for notability, but in general Wikipedia cares what independent sources say about subjects not what subjects say about themselves as they are always going to be biased/selective. [7] is fine. So in general not enough to show notability, in my opinion as it stands. Look for if any other sources exists that are independent and have some depth to them. Hope that helps explain things a bit more. Cheers KylieTastic (talk) 09:23, 13 June 2025 (UTC)[reply]
Thank you for your replies @KylieTastic and @Jmcgnh - they are very helpful in understanding the concerns that led to the rejection of the article. As you outline, the 2 main sources that contribute to the notability of the events of Untether AI are the 2022 IEEE Hot Chips 34 Symposium and the "New MLPerf Inference v4.1 Benchmark Results" from mlcommons.org.
For the IEEE, (and the questions from KylieTastic around " is IEEE so reliable publisher, but not independent as from Untether Ai"?), I'm confident that many will find IEEE Symposiums to be reliable, credible and independent. The Institute of Electrical and Electronics Engineers itself has a Wikipedia page (https://en.wikipedia.org/wiki/Institute_of_Electrical_and_Electronics_Engineers) and has been the publisher of professional peer-reviewed journals and symposiums since 1963. I suggest the acceptance of at-memory computing within the IEEE as a significant AI/ML architecture provides credibility and, to some extent, notability (there is an extensive peer-review process to be accepted at the symposiums).
I think the credibility of the mlcommons.org and MLPerf benchmark is a subject of greater debate. It is not uncommon to struggle to benchmark new technologies (a similar debate existed around database transaction processing and https://www.tpc.org/ in the 1990s). So yes, as I'm writing this, I'm wondering
1. how to convey the credibility and notability of an MLPerf disclosure to people within and outside of the AI/ML industry and
2. if a Wikipedia page on MLPerf would be worth looking into as well (see https://en.wikipedia.org/w/index.php?title=MLPerf&redirect=no )
Towards the credibility and notability of an MLPerf disclosure, to the best of my knowledge, mlcommons.org and the MLPerf benchmark is the best the AI/ML industry has to compare performance. I'm not expert enough to comment on shortcomings or bias in the benchmark. So the best I can offer on the credibility of the MLPerf benchmark is the relatively widespread coverage of MLPerf by the large AI hardware industry players (i.e. https://www.nvidia.com/en-us/data-center/resources/mlperf-benchmarks/, https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets/documents/Cloud%20AI%20100%20MLPerf%202.0%20inference%20performance%202022-05-16.pdf, https://rocm.blogs.amd.com/artificial-intelligence/mlperf-inf-4-1/README.html and https://newsroom.intel.com/artificial-intelligence/what-is-mlperf-understanding-ais-top-benchmark ).
Further, while the IEEE may publish papers and discussions on many AI/ML architectures, few have made it all the way to commercial production and an MLPerf disclosure showing best-in-class. Of particular note, Untether AI is one of very few startup companies that have made an MLPerf disclosure (which are typically dominated by companies such as NVidia, AMD, Qualcomm or other much larger companies). I suggest the combination of the IEEE citation and the MLPerf citation is sufficient to support the notability of the article.
I hope you'll take the above into account when considering the notability of my article. Any further suggestions for improvement would also be greatly appreciated. Thank you again for your time and consideration. Rhg2 (talk) 21:30, 13 June 2025 (UTC)[reply]
Our job as reviewers is to judge on what we believe the consensus view is, not our personal views. So all I can say is that in my experience it does not matter that a very respectable publisher (IEEE) published the first source, only that it is written by Untether AI so it is primary and does not count for notability. The general view is that if such a source is notable then other independent reliable sources would have written about it, giving you secondary sources. As it is not a peer reviewed paper but just a PowerPoint presentation is is also of dubious value without supplementary notes.
I think a key thing to point out is that English Wikipedia has its notability guideline as a way to select what topics are acceptable. It works, but is not perfect by far. There are a lot of subjects that are worthy information but just don't get articles because no one writes about them. On the flip side people who are just famous for being famous etc get articles. As individuals we may not agree, but Wikipedia is run by consensus.
As to if MLPerf is a valid subject, I would suggest seeing if you can find 3+ independent reliable sources with significant coverage and then maybe ask Wikipedia:WikiProject Computing or the Wikipedia:Teahouse if they think it's enough. Asking for feedback from forums with multiple people who will see the question is always good here as it helps show the consensus. You could also ask for feedback on Draft:Untether AI, maybe someone at Wikipedia:WikiProject Computing will be able to suggest more sources. Regards KylieTastic (talk) 10:41, 14 June 2025 (UTC)[reply]
Thank you again for your reply, @KylieTastic. I looked into citations of the referencces I provided, and I did find 1 citation of the HotChips presentation (from https://ieeexplore.ieee.org/document/9895618/citations#citations ).
I also looked into a bit more of the history of at-memory compute and Untether AI and added references to an earlier Computational RAM paper that has 62 citations (https://ieeexplore.ieee.org/document/5727436/citations) as well as a Computational Memory patent that has 11 Patent citations and 15 Non-patent citations (see https://patents.google.com/patent/US11614947B2/en).
I'm hoping this will provide further evidence of the notability of my article.
Also thank you for your suggestion to post on Wikipedia talk:WikiProject Computing - Wikipedia - I have given that a try as well.
Regards Rhg2 (talk) 16:09, 15 June 2025 (UTC)[reply]
Retrieved from "https://en.wikipedia.org/w/index.php?title=User_talk:Rhg2&oldid=1295745790"