Talk:Neats and scruffies

Untitled

Someone who subsequently decided they needed a wiki vacation said it isn't clear why this page is important. I find this a bit rich having just been going through the full theory of Dr. Who's origins on wikipedia (for no good reason --- o wait, I think the TARDIS was actually a featured article! now that's important) a few days ago, but anyway there may be a point. It's clear to a computer scientist that while we treat holy wars and everything else somewhat light-heartedly, that this page actually reflects a serious, long-term debate in the field. Is there a way to communicate this to the general reader? (besides linking to the holy war page?) --Jaibe 16:33, 12 March 2006 (UTC)

That's a good point, changing the name of the page might help I guess; there's no debate anyway scruffies RULE! :) But yeah anyone who says something like this isn't a worthwhile page should look in to the amount of Stargate pages ;) Ultima 22:41, 4 November 2006 (UTC)

I would go farther: it's not just a serious, long-term debate; it's central to the field of AI. I don't think one can really claim to understand AI (to the extent anyone does) without understanding the distinction and how various approaches fall on either side of it or combine elements of both. Slburson 23:06, 12 February 2007 (UTC)

Origin of term?

Does anyone know who came up with this terminology? The earliest I've seen is T.R.G. Green's chapter "The Nature of Programming" in "Psychology of Programming" by Hoc, Green, Samurcay & Gilmore, Academic Press 1990, p. 21: "Different programming cultures stress different values, on one hand neatness and well-defiinedness, and on the other hand openness and effectiveness. The neat-scruffy differences show themselves both in . . ."

good question, no idea... --Jaibe 21:21, 25 May 2007 (UTC)

I can't find any reference that is earlier than that from some quick searching. MattOates (Ulti) 13:18, 26 May 2007 (UTC)

FIXED. CharlesGillingham 04:46, 26 June 2007 (UTC)

Roger Schank first used those terms "scruffy" and "neat" at an AI conference in the 1970s. He proudly called himself a scruffy. 71.183.59.144 (talk) 02:17, 26 October 2011 (UTC)

The terminology is sourced to the late 1970s or early 1980s and originated by Schank according to this (argh, can't find the source now):

"In particular, certain personality traits go hand and hand with certain styles of research. Schank and Abelson hit upon one such phenomenon along these lines and dubbed it the neats vs. the scruffies. These terms moved into the mainstream AI community during the early 80s..."

Abelson presented the neats vs. the scruffies in a keynote address at the Annual Meeting of the Cognitive Science Society in 1981. Here are some selected excerpts from the accompanying paper in the proceedings:

“The study of the knowledge in a mental system tends toward both naturalism and phenomenology. The mind needs to represent what is out there in the real word, and it needs to manipulate it for particular purposes. But the world is messy, and purposes are manifold. Models of mind, therefore, can become garrulous and intractable as they become more and more realistic. If one’s emphasis is on science more than on cognition, however, the canons of hard science dictate a strategy of the isolation of idealized subsystems which can be modeled with elegant productive formalisms. Clarity and precision are highly prized, even at the expense of common sense realism. To caricature this tendency with a phrase from John Tukey (1969), the motto of the narrow hard scientist is, “Be exactly wrong, rather than approximately right”.
The one tendency points inside the mind, to see what might be there. The other points outside the mind, to some formal system which can be logically manipulated [Kintsch et al., 1981]. Neither camp grants the other a legitimate claim on cognitive science.... an unnamed but easily guessed colleague of mine (Schank?), who claims that the major clashes in human affairs are between the “neats” and the “scruffies”. The primary concern of the neat is that things should be orderly and predictable while the scruffy seeks the rough-and-tumble of life as it comes ... The fusion task is not easy. It is hard to neaten up a scruffy or scruffy up a neat. It is difficult to formalize aspects of human thought that which are variable, disorderly, and seemingly irrational, or to build tightly principled models of realistic language processing in messy natural domains.
What are the difficulties in starting our from the scruffy side and moving toward the neat? The obvious advantage is that one has the option of letting the problem areas itself, rather than the available methodology, guide us about what is important. The obstacle, of course, is that we may not know how to attack the important problems. More likely, we may think we know how to proceed, but other people may find our methods sloppy. We may have to face accusations of being ad hoc, and scientifically unprincipled, and other awful things."

Source is Chapter 5 of this book edited by Schank and published in 1994, titled "Beliefs, Reasoning, and Decision Making: Psycho-logic in Honor of Bob Abelson". Article needs clean-up, which I am doing now.--FeralOink (talk) 13:58, 2 August 2021 (UTC)

(1) We have three good secondary sources (McCorduck, Russel&Norvig, Crevier) that give Schank credit in the article, so no argument there.

(2) We have one paragraph about the "cultural" or "personality type" aspect of the distinction. I worry that the "personality type" part of the distinction is a distraction from the more substantive intellectual contrast. I noticed people have softened the "cultural" paragraph somewhat over the years. (It used to mention "hacking" and "scruffy AI" were born at the same time and place). So my vote would be to leave the personality stuff off unless we have good reason to add more. Let me know what you think.

(3) The long quote from Abelson is good, and I think we should use some of it; a paragraph in the origin section. Still thinking about what bits might be best. ---- CharlesGillingham (talk) 02:22, 9 September 2021 (UTC)

Elegant != Formal

A provably-correct solution is not necessarily elegant (see the Four Color Theorem for an example), nor is an elegant solution necessarily provably correct (for example, the human brain). I can't help but wonder if this is a false dichotomy. Metasquares 16:58, 9 November 2007 (UTC)

The human brain isn't elegant, but that's beside the point, of course. ---- CharlesGillingham (talk) 03:29, 29 June 2009 (UTC)

You're correct, Metasquares. I balked at the use of strictly defined terms in the article, e.g. being provably correct, converging, mathematical elegance, being complete vs being NP complete, and so forth. I'm trying to clean up this imprecise usage by not wikilinking to the terms. The words are perfectly fine English usage, so I am not necessarily altering them. Charles, that is a non-sequitur, about the "inelegance" of the human brain. I think Metasiquares made a good point, um 14 years ago.--FeralOink (talk) 10:34, 3 August 2021 (UTC)

intro

The intro currently reads:

The distinction was originally made by Roger Schank in the middle 70s to characterize the difference between his work on natural language processing from the work of John McCarthy, Alan Newell and others whose work was based on logic (Prolog, Soar, etc.).[1] The "frames" introduced by Marvin Minsky in 1975 were also considered "scruffy" at the time.

Since casual readers will not understand the difference between Schank's work and McCarthy's, this first sentence is a poor description. Can someone familiar with the details of the differences add some qualification to "his work on natural language processing"? E.g., ", which was more ad-hoc" or "which was based on approximate heuristic solutions". This will set up an opposition to "work ... based on logic". —Preceding unsigned comment added by 129.2.175.74 (talk) 20:38, 11 February 2008 (UTC)

I've fixed this, which blossomed into a fairly complete history of the term. ---- CharlesGillingham (talk) 03:29, 29 June 2009 (UTC)

Well-known neats and scruffies

Are the individuals listed in this section self-described as being in these categories, or has the label been attributed to them by others? Each individual should have proper sourcing for the label. If they all called themselves that, then the section might be more suitably renamed "Self-described neats and scruffies". "Well-known" is not an encyclopedic term and should be avoided, like the word "obvious". (Obvious to whom? Well-known to whom?) Robert K S (talk) 20:45, 23 February 2008 (UTC)

The word obvious isn't used in the article is it? The reason the title “well-known” is used, is that they have been defined by the nature of their own work. The terminology links in with the two main philosophical viewpoints of strong and weak AI with both Neats and Scruffies having a specific point of view on both. Scruffies - believe that Strong AI can only come about from emergent processes, and not an ontological representation of knowledge. Essentially Scruffies differ from Neats in that they believe human like intelligence can only come about from representing the underlying system of intelligence, rather than trying to qualify high-level mental process as logic or something similar. Although it might not appear particularly encyclopaedic the list of names helps to define the difference between the two points of view in a way that is more tenable for people not familiar with AI. Can you think of something different we could do to better illustrate the difference? MattOates (Ulti) (talk) 13:47, 24 February 2008 (UTC)

Most of the researchers listed are now in the history section, with some citations as to whether they are "neat" or "scruffy". The lists are basically correct, I think. Not sure about David McAllester, Daphne Koller or Steve Grand. The first two make sense, from their short Wikipedia biographies. Steve Grand seems wrong to me, because artificial life (i.e. "let's use an elegant evolutionary algorithm and a elegant machine learning algorithm so an intelligent machine creates itself!") seems inherently neat to me. However, if Steve Grand's version of artificial life is "let's painstaking study and generalize the systems and subsystems and circuits and sub-sub-circuits of animal nervous systems to create an intelligent machine", then he's scruffy. ---- CharlesGillingham (talk) 03:38, 29 June 2009 (UTC)

There needs to be reliable sources that say these people are scruffy or neat else we fall foul of WP:OR or WP:SYN. I looked at several people's article and there was no mention of scruffy or neat. pgr94 (talk) 09:57, 23 June 2011 (UTC)

I'm knocking out the ones that aren't mentioned in Crevier, McCorduck or Russell and Norvig. ---- CharlesGillingham (talk) 20:29, 23 June 2011 (UTC)

More links, please

Please, add more links to related articles, because I was looking for a related philosophic/AI problem, but could only remember "neats vs. scruffies". This would be a good general improvement for AI articles on this site. Erudecorp ? * 00:31, 11 December 2008 (UTC)

Why not call it "slobs vs snobs" instead?

It seems like a better term than "neats vs scruffies". No offense, but that sounds pretty ridiculous. —Preceding unsigned comment added by 24.60.157.218 (talk) 16:17, 2 April 2010 (UTC)

lol! I prefer slobs vs snobs too, as it is alliterative and easy to remember! Unfortunately, the neat vs scruffy terminology is explicitly referenced by those active in the field in the 1970s through 1990s, so we need to stick with that.--FeralOink (talk) 10:37, 3 August 2021 (UTC)

Rename

The current article title Neats vs. scruffies is unnecessarily adversarial by virtue of the word "versus". I strongly recommend renaming this article to Neats and scruffies in order to retain a NPOV, as this fits more with established style, and the nature of the article is contentious enough as it is. — Preceding unsigned comment added by 75.139.254.117 (talk) 22:44, 2 October 2016 (UTC)

DONE! And thank you for your suggestion.--FeralOink (talk) 10:38, 3 August 2021 (UTC)

How is machine learning neat?

Machine learning is only provably correct for the known examples it was trained for. If that is not an adhoc approach to AI, then I don't know what is. Big data is the epitome of a scruffy. No model, just data, not formalism, besides fitting a curve/model to the given data. It is the exact same approach that scruffies follow: abstracting from examples for specific sub tasks.<unsigned>

Just because some mathematical methods are employed, like optimization for a sub-problem, i.e. curve fitting, does not make the approach itself neat.

Obviously, scruffies also use mathematically rigorous approaches, when employing provably correct algorithms, such as searching trees, or certain signal processing approaches.

So far, the only valid "neats", are those doing GOFAI: they use a minimal model and deduce everything based on it, with no added assumptions or axioms along the way.

Machine learning is only based on added assumptions/axioms: the training data. New for each problem, no general model.<unsigned>

Yeah, I noticed that too. Not sure who introduced machine learning to the article. I'm trying to clean up, e.g. removing the jargon about scruffies just being casual hackers throwing stuff together in an ad hoc manner. I don't know enough about the people involved though. I know about the methods you mention (curve fitting, converging series, mathematical modeling) but not necessarily who did what. I don't even know whether most of these guys, the neats OR the scruffies, would be comfortable with "big data" (i.e. lots of specious results with very low cost of being wrong).--FeralOink (talk) 10:48, 3 August 2021 (UTC)

I think you are misunderstanding what the difference between neat and scruffy is -- you seem to be confusing it a little with "soft" computing vs. "hard" computing.

Neat: there is one magic bullet technique that does most of the work (such as logic in the 70s, or deep learning in the 2010s). Scuffy: AI requires solving lots of different problems, using lots of different techniques.

You said above "the only valid "neats", are those doing GOFAI". This is backwards. Almost all GOFAI was scruffy, not neat. Only methods based on formal logic were considered neat.

You also said above "I don't know who introduced machine learning into the article". The leading AI textbook, Russell & Norvig, describes modern statistical techniques, including all of modern machine learning and neural networks as neat. See the quote about "victory of the neats" in the article. ---- CharlesGillingham (talk) 07:04, 5 September 2021 (UTC)

Charles, what you describe as the leading AI textbook by Russell and Norvig, was published in 2003. Were they doing "modern machine learning" in 2003?! Also, I don't know what you mean by "modern statistical methods". Statistical methods are relatively static. There's a lot more Bayesian stuff (applications of Bayesian methodology now, in the 21st century) but the Reverend Baye (Bayes?) developed his ideas in the 19th century. Conditional probability is old as the hills. There's WP:SYNTH going on in the lead, likening neats to physicists and scruffies to biologists, especially as sourced to Noam Chomsky. I am extracting myself from further involvement in this article.--FeralOink (talk) 06:53, 8 September 2021 (UTC)

I'm sorry you feel that way -- I reached out to you because I hoped I could get your input and help.

The 2003 edition is sufficient, because this is fundamentally a history of science article. It also was the only edition available when this article was first written in 2007 or so. If someone has time, it could be updated it to the 2018 edition. It's still the most popular textbook. And it still has the "victory of the neats" bit in the history chapter.

By "modern statistical methods" I mean "modern statistical methods used in AI"-- neural networks, various classifiers, evolutionary algorithms are all "statistical methods for AI" (because they are tools borrowed or adapted from statistics). The use of these tools, in AI at least, has not been static,f but has been dynamically changing for the past 20 years, with literally tens of thousands of new papers.

There's no WP:SYNTH. I think the article accurately describes the way Schank and Minsky characterized themselves, based on the most reliable sources available. I think the physics/biology metaphor is apt. It lines up pretty well with your long quote from Abelson, at least when I read it.

Not that this counts, but, I studied AI as an undergraduate in 1979-1984 when "neat" vs. "scruffy" was still discussed. I worked on expert systems in the late 80s and 90s, long after the point where everybody had become completely sick of the idea. I have my own thoughts about this subject, but they don't appear in the article. ---- CharlesGillingham (talk) 06:55, 9 September 2021 (UTC)

Rebuild, Sept. 2021

I've rebuilt the article, because I've noticed in the comments here that many people seem to be very unclear about what the difference between scruffy and neat really was, don't seem to understand that this was only a topic of discussion 40 years ago, and especially that they seem to have trouble classifying modern AI and machine learning as "neat", may not realize the question was never completely settled and that both sides have merit.

The upshot is that all comments above don't apply the current article. If anyone who posted above has questions or issues and is still watching this page, please let me know if I've helped to clarify the subject. ---- CharlesGillingham (talk) 22:27, 7 September 2021 (UTC)

No one ever said that scruffies lack merit. I believe that the comments above, stretching back over the past twenty years on the talk page, DO have merit and should not be dismissed as irrelevant without thorough sourcing. I don't see a lot of new sourcing to the article since your rebuild. This subject is not my forte, I am merely a mathematical statistician and applied probability professional (not academic) so I will bow out of further involvement at this point.--FeralOink (talk) 06:58, 8 September 2021 (UTC)

I was just asking you to review the rewrites. The only real difference is the lede, I separated the article into sections so it reads a bit more like history, and I added the Society of Mind with Minsky's great quote. I also fixed the citation format to standard efn/sfn form.

I was under the impression that the comments above referred to aspects of the article that I fixed -- i.e., the definition.

I don't believe this article needs any new sources, because it already has the most reliable sources available: Russell & Norvig, the leading AI textbook. McCorduck's classic and insightful history of AI, and Crevier's insider's history of the field in the first four decades. ---- CharlesGillingham (talk) 01:53, 9 September 2021 (UTC)

Ongoing debate ...

It seems as if I arrived a little late to this discussion.

I do not think that there is "the victory of neats in the 1990s." In my opinion, this is similar to the concept of yin and yang; one cannot have one without the other. It would be nice to see better wording, because this is not a competition between the two types of approaches.

The argument for either approach depends on one's personality type. A person's personality type conditions one's philosophical approach too. This is a psychological truth. See Carl G. Jung's Personality Types.

I would argue that an eclectic approach is the most appropriate approach to any field of study, especially problem solving.

AndrewMcGraw1970 (talk) 14:47, 26 September 2021 (UTC)AndrewMcGraw1970

Yes, Russell and Norvig have retracted their "victory of the neats" comment in the latest edition (2021) of their textbook, because the fine-tuning of deep learning networks is more scruffy than neat. The article has been updated to be more specific. See the final paragraph of the lede. ---- CharlesTGillingham (talk) 19:30, 9 July 2023 (UTC)

Address political rally

Mca politics 154.159.237.54 (talk) 19:34, 29 August 2023 (UTC)