Wikipedia:Articles for deletion/Huang's law

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 no consensus. Sandstein 13:59, 4 October 2020 (UTC)[reply]

Huang's law

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

(Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL)

Not seeing enough depth of sourcing or coverage to justify an article. Plus the basic error in the 2nd paragraph. According to the first para, Moore's law would predict a 32-fold increase, not 10-fold. Edwardx (talk) 18:42, 23 September 2020 (UTC)[reply]

You misunderstood. The Moore's law says "doubles every two years", not "doubles every year". The topic of this article is very important and even if the article was removed, the concept should be preserved either on the Moore's law page, or on some page related to Huang or Nvidia. The GPU design industry is much more secretive and mysterious and doesn't get as much attention and coverage as the topics related to CPUs, so even this tiny bit of knowledge is valuable. But this law also explains why so little is known about GPUs - the progress is just happening too fast to even track it. 37.225.41.14 (talk) 20:11, 23 September 2020 (UTC)[reply]
Keep - These simple errors can be fixed with edits, deletion is not a solution. WhoAteMyButter (📬│✏️) 21:26, 23 September 2020 (UTC)[reply]
Delete Yet another neologism written about far too soon. One reference is a not-in-depth primary source which throws in "Huang's law" for the sake of a clickbait headline (and headlines are not to be trusted), following it up with an admission that nobody actually called the idea "Huang's law" yet. The other two sources are newsblogs from this past week. Nothing here establishes notability. The best source that further searching could find is another opinion column saying the idea doesn't make sense. Wikipedia is not for trying to make fetch happen. XOR'easter (talk) 22:13, 23 September 2020 (UTC)[reply]
Addendum The article has been expanded since I !voted, but I don't think WP:HEY applies. Nothing in the expansion indicates to me that the claim has been around long enough to have been evaluated seriously (beyond the level of industry hype and omnipresent tech churnalism). XOR'easter (talk) 19:20, 24 September 2020 (UTC)[reply]
You've completely missed in your arguments the fact that ExtremeTech is one of the most important outlets that discusses the hardware architectures of GPUs. On Wikipedia, the term Computational power links to the article about Moore's law, which revolves around transistor counts, rather than computational power of hardware. The graphs in the Moore's law article that depict the progress of computational hardware are outdated and incomplete. To make an up to date graph for articles about Moore's law and Accelerating change, it would be necessary to include smartphone CPUs and GPUs, the tiniest of microcontrollers, modern GPGPU hardware from Nvidia and AMD, the Intel Xeon Phi, the AMD Threadripper, and various other accelerators like for example the ones used for AI acceleration. This graph would have to use a logarithmic scale, on which it would be then possible to draw a grid of slopes that would show the Moore's law growth rate, which would allow to verify Moore's predictions. I wasn't able to find a thorough examination of these trends or any relevant knowledge about the recent advances in those areas anywhere on Wikipedia. That's why i think that the idea discussed in this article is relevant - it's simply missing from Wikipedia. I think that your argument that this article is trying to make fetch happen is completely on point - naming a whole article after just two news pieces doesn't make much sense. The term "Huang's law" should only be mentioned as an anecdote, rather than be the sole reason for the articles existence at this point. The TOPS500 website includes data about adoption of Nvidia hardware in datacenters, and there is a dataset on a website related to progress of AI which includes the computational power of GPUs from the last two decades. I am not sure if there are other datasets like this available for other types of hardware like smartphone chipsets, but either way making a complete graph would likely take at least a day. The TOPS500 statistics show how important Nvidia has become in the supercomputer market, and these statistics alone aren't included in the articles about supercomputers or GPGPU. Joel Hruska hasn't objected the observation that the progress of GPU hardware hasn't been limited by Moore's law in the same way that it has affected the CPU market. Joel merely says that "Huang's law" never existed in the first place, that this term came out of nowhere, to object the recent articles that appeared in the less technologically fluent news outlets. Many of the Joel's observations would be very important in an article that would discuss this topic. Joel seems to be trying to say that "Huang's law" is not predictive of further progress of GPU hardware, and that it is subject to the trend of diminishing returns and as such, it has no power to overcome the limit of Moore's law. 37.225.41.14 (talk) 07:59, 25 September 2020 (UTC)[reply]

Oh my god, Danny Devito! I love your work! –Deacon Vorbis (carbon • videos) 15:15, 25 September 2020 (UTC)[reply]

Note: This discussion has been included in the list of Computing-related deletion discussions. XOR'easter (talk) 22:14, 23 September 2020 (UTC)[reply]

Note: This discussion has been included in the Article Rescue Squadron's list of content for rescue consideration. Lightburst (talk) 00:55, 24 September 2020 (UTC)[reply]

Keep I do not understand the nomination. It is easy to find RS. Wall Street Journal here, HotHardware here, Fox 24 News here, Yahoo News here and a naysayer Extreme Tech here. As a gentle reminder WP:AFDNOTCLEANUP and WP:BEFORE Lightburst (talk) 00:55, 24 September 2020 (UTC)[reply]
The Fr24News item is just a mirror of the WSJ item, which is the newsblog post where the writer says, "I call it Huang's law". Proposing a neologism is not evidence that the neologism has been adopted. The HotHardware item is churnalism, repeating the claims from the WSJ newsblog post without offering independent analysis. XOR'easter (talk) 05:29, 24 September 2020 (UTC)[reply]
Delete and merge into Jensen Huang for now with no prejudice against re-creation should it build sustained notability later on. I agree with the argument that it's neologism. Wikipedia isn't really a place to create new article as if creating news stories. Even a re-direct isn't appropriate, because we don't know if calling his observations "Huang's law" is more than just a brief press fad. Graywalls (talk) 02:15, 24 September 2020 (UTC)[reply]
Keep
1. Perry, Tekla S. (May 2018). "Move Over, Moore's Law: Make Way for Huang's Law". IEEE Spectrum. IEEE. Retrieved September 24, 2020. Graphics processors are on a supercharged development path that eclipses Moore's Law. ... GPUs are also advancing more quickly than CPUs because they rely upon a parallel architecture, Jesse Clayton, an Nvidia senior manager, pointed out in another session."
2. Mims, Christopher (September 19, 2020). "Huang's Law Is the New Moore's Law, and Explains Why Nvidia Wants Arm". Wall Street Journal – via www.wsj.com.(subscription required) reprinted in "Huang's Law is New Moore's Law and explains why Nvidia wants an arm". Fox 24 News. September 29, 2020.
3. Hayes, Caroline (October 11, 2018). "Jensen Huang: Moore's law is dead – long live AI". Electronics Weekly. Metropolis International. Retrieved September 24, 2020. ... there are two dynamics controlling the computing industry today – the end of Moore's law and software that can write itself, artificial intelligence, or AI. ... We can study where bottlenecks are. New software systems make the application go faster, not just the chip.
4. Hruska, Joel (September 22, 2020). "There's No Such Thing as 'Huang's Law,' Despite Nvidia's AI Lead". Extreme Tech.

These settle the issue. We should include this pro and contrary views, but should not purge the article from Wikipedia. WP:AFDNOTCLEANUP and WP:BEFORE. Exceeds WP:GNG. Indeed, not now the article it was when it was nominated for deletion, so WP:HEY applies. 7&6=thirteen (☎) 11:27, 24 September 2020 (UTC)[reply]

The Electronics Weekly story from 2018 does not use the term "Huang's law". XOR'easter (talk) 19:22, 24 September 2020 (UTC)[reply]

Keep - WSJ and Spectrum sources adequatly demonstrate that this has gone beyond nelogism. ~Kvng (talk) 14:13, 24 September 2020 (UTC)[reply]
- How does a source proposing a neologism demonstrate that it has gone beyond neologism? I am legitimately confused by how people are taking the WSJ item. It's literally a guy writing I call it Huang's Law. XOR'easter (talk) 19:20, 24 September 2020 (UTC)[reply]
Keep The timing of this AfD coincides with the publication of "There’s No Such Thing as ‘Huang’s Law,’" in ExtremeTech in which the journalist Joel Hruska says it is "too soon" to determine the existence of the law. However.. this is one journalist whose job it is to create stories, often through controversy as a sort of devil's advocate. But it passes Notability, and the concept is evolving - it has a critic, but so did Moore's Law. The existence of a critic is not reason to delete. Even if it is discredited, it is now part of the history of technology and would need a place somewhere on Wikipedia. -- GreenC 15:22, 24 September 2020 (UTC)[reply]
Comment I think it is sufficiently covered to comply with Neologism, Wikipedia not a dictionary. 7&6=thirteen (☎) 17:16, 24 September 2020 (UTC)[reply]
The existence of a critic is not a reason to delete, but the existence of only one critic is a reason to think that the whole idea is too recent to have been evaluated critically. XOR'easter (talk) 19:20, 24 September 2020 (UTC)[reply]
Obvious Keep -Really?? Detailed coverage in IEEE Spectrum is not good enough to establish notability? Trying to reconnect (talk) 22:11, 24 September 2020 (UTC)[reply]
I wouldn't call the IEEE Spectrum story "detailed". It's a few paragraphs, summarizing a keynote speech without soliciting opinions from other experts or generally doing anything that serious science journalism would. It's not even reporting on a fully-formed idea (Huang was throwing a variety of numbers out there; it seems he's still working out the exact multiple he's talking about). XOR'easter (talk) 22:59, 24 September 2020 (UTC)[reply]
I disagree. The entire article was about this. That it is not as detailed as you would have liked is irrelevant. IEEE is super reliable and reputable, it devoted an entire article to this, it knocks WP:GNG out of the park. Trying to reconnect (talk) 23:21, 25 September 2020 (UTC)[reply]

Establishing notability for a topic can be realized by publishing a book. In the easiest case, it would be an ebook sold in the Amazon kindle store. In the more advanced attempt, the book is available in a printed version and is used for teaching purposes in a university course. The reason why an encyclopedia is dependent from book publication is because of an information hierarchy. Primary sources are aggregated into text books, and text books are aggregated into encyclopedic databases. Bypassing the information layers is producing a larger complexity. For example, if the referenced article in the “Wall street journal” comes to the conclusion that “Huang's law” is different from previous articles, then the Wikipedia article has to be adapted quickly. This will result into a rolling release encyclopedic article which is maintained by activists on a daily basis.--ManuelRodriguez (talk) 08:51, 27 September 2020 (UTC)[reply]
Comment There is an ongoing discussion about inclusion of material, sourcing and reliability of sources here. I note that so interested editors can comment (if they choose) in the appropriate spot. I don't think the AFD should be cluttered with discussion about that, since it is immaterial to this effort. Cheers. 7&6=thirteen (☎) 13:16, 25 September 2020 (UTC)[reply]
Delete. The main sources bolstering this are not reliable. One of which starts out by describing Moore's law as a "prediction" rather than an empirical observation (and later a bit of a self-fulfilling prophecy). This per se renders the source unreliable. Another conflates the idea of transistor density (which is what Moore's law is actually about) with processor power (which is definitely isn't, and which is well-known). There's also a lot of overly credulous and hyped-up treatment of the concept rather than serious analysis of it. Maybe worst of all, that it's just not well-defined. "Huang's law claims that a synergy between hardware, software and artificial intelligence makes the new 'law' possible." The law claims something that makes the law possible? Huh? And there's still no clear statement of it. All this, at the very very least, adds up to a case of WP:TOOSOON. If this really turns into a better-defined thing that people really refer to on a large scale, then come back to this in 5 or 10 years. –Deacon Vorbis (carbon • videos) 15:09, 25 September 2020 (UTC)[reply]
We don't gauge WP:RS based on our own assessment of the accuracy of an individual article. WSJ and Spectrum are reliable sources. If we base an article on these sources and they turn out to be wrong, we'll fix the article but we won't delete it. WP:GNG and WP:42 trump WP:TOOSOON. ~Kvng (talk) 16:46, 25 September 2020 (UTC)[reply]
Yes, we can in fact. WSJ may be nominally reliable, but that doesn't mean every single piece it contains is. If we identify a major, glaring, fundamental error in an article there, then it does call its reliability into question. Determining reliability is editorial judgement that we do have a say in. As for as the Spectrum piece, as XOR also pointed out above, a deeper reading of it indicates that it's about his talk, and not about any actual law: "So Huang was throwing a variety of numbers out there; it seems he’s still working out the exact multiple he’s talking about." One specific benchmark measurement pointed out by one guy in one talk does not a law make. It's an ill-defined nebulous topic that cannot be written about in a cohesive way until and if it becomes more widely accepted and better specified. WP:GNG doesn't apply because there is no topic to apply it to. –Deacon Vorbis (carbon • videos) 17:23, 25 September 2020 (UTC)[reply]
Looks like a lot of sources about something that isn't a topic.. People play devil's advocate with the law to test the strength/predictive power. Oh wait, I said "prediction" because that is one thing laws can do for people similar to how the law of gravity provides predictive services for NASA, it's a common way people approach and use laws; but according to you because the source says "prediction" the entire source is invalid. C'mon your trying to discredit the sources, I get that, but is making mountains out of semantic and contextual molehills, it's not a reason to toss the source. -- GreenC 21:38, 25 September 2020 (UTC)[reply]
In a little more detail, look at the WSJ source:
I call it Huang’s Law ... It describes how the silicon chips that power artificial intelligence more than double in performance every two years. While the increase can be attributed to both hardware and software, its steady progress makes it a unique enabler of everything from autonomous cars, trucks and ships to the face, voice and object recognition in our personal gadgets.
So we're writing an article on a "law" that someone attempted to coin 6 days ago. A "law" which is so vague that the current article can't even properly describe it. So yeah, the source is not reliable insofar as it attempts to coin a new "law" of computing. XOR's fetch analogy above was both hilarious and eerily apt. If Gordon Moore had made his famous observation in a keynote speech 6 days ago, and then some people wrote about it and called it "Moore's law", it would also be inappropriate for an article. And it would have even have had the advantage of being specific: "transistor density doubles every 18-24 months" (or whatever the exact time frame is). It wasn't using some vague notion of "performance" that Mims is; it was talking about transistor density. We wouldn't have been able to write about Moore's law for quite a few years at the very least...until we knew that people were continuing to use the term in the same way, and that people had written about it and studied it. This isn't a topic. –Deacon Vorbis (carbon • videos) 00:39, 26 September 2020 (UTC)[reply]
This is a well defined and widely researched topic, normally referred to as Beyond Moore or More than Moore, and it can be directly linked from this phrase in the Moore's law article. It's about reversing the Wirth's law, and about how the the room can be found within the software and hardware design itself, instead than in materials science. It touches upon the topics of GPGPU, hardware acceleration, and software optimization. It's about learning to do better engineering, instead of throwing billions of dollars at another company to do the work for you. 37.225.41.14 (talk) 05:48, 26 September 2020 (UTC)[reply]

coin 6 days ago .. actually some years ago. The term with capital letters has been in usage since April and May 2018. The HotHardware piece claiming a 2020 origin is not surprising as etymology is often confusing, for journalists or Wikipedia and anyone, but that sort of error doesn't invalidate the entire HotHardware article as unreliable. -- GreenC 13:32, 26 September 2020 (UTC)[reply]
@GreenC: If you're talking about the Spectrum piece, all it says is: "Huang, who is CEO of Nvidia, didn’t call it Huang’s Law; I’m guessing he’ll leave that to others.". That's the exact opposite of using it, and the piece was very clear to point out that Huang was being pretty vague with his claims, which also supports the fact that there's no topic here. Let me stress this point, because it's vital: there is no topic here. The ExtremeTech piece, contrary to trying to use it to support notability, goes out of its way to also explain why this is not some well-defined thing, like Moore's law is, and to criticize Mims for not knowing what the hell he's talking about. We can't write about a topic if there is no well-defined topic. –Deacon Vorbis (carbon • videos) 14:13, 26 September 2020 (UTC)[reply]
Nearly every source in the article use the term and consider it a topic. Who coined the term originally is beside the point only that it's in currency. It also doesn't matter if the Law is accurate or not, only that it is in currency. Wikipedia could have articles on real laws, fake laws and controversial laws. The question is the degree of coverage for notability purposes. -- GreenC 14:21, 26 September 2020 (UTC)[reply]
But it can't have an article on a "law" that doesn't even have a definition, because it was an attempted coinage by someone a week ago who didn't really know what he was talking about. If this sticks, and there's actually some nailing down of precisely what this is about, and people actually write about it seriously, then we can. But that takes years to get a proper perspective on. And until then...there is no topic. –Deacon Vorbis (carbon • videos) 14:28, 26 September 2020 (UTC)[reply]
Keep I came here today looking forward to learn about this concept sooo it is relevant enough and deserves its place in Wikipedia — User:Ejrrjs says What? 20:50, 25 September 2020 (UTC)[reply]
Delete, and merge content into Jensen Huang, as another editor said. This is a neologism that seems to be borne out of Nvidia marketing/PR. I have not heard of this being widely or seriously referenced in academic publications. On the other hand, there has been some coverage of it in reputable sources, so its content merits a place in Wikipedia after all. If/when this gets widely referenced, we can split it out to its own page. This is a message from Mr. XYZ. 18:29, 2 October 2020 (UTC)[reply]
Mr. XYZ, not that it is critically relevant but do you have any evidence to back your claim that this comes from Nvidia marketing/PR? ~Kvng (talk) 15:22, 3 October 2020 (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.