Neats and Scruffies

The terms 'neat' and 'scruffy' describe two contrasting approaches to AI research, originating in the 1970s with Roger Schank. Neats use a single formal paradigm like logic or neural networks and verify via rigorous mathematics, while scruffies use diverse methods and rely on incremental testing. Modern AI combines both approaches, as machine learning applications require hand-tuning despite mathematically rigorous algorithms.

Neats and scruffies In the history of artificial intelligence https://en.wikipedia.org/wiki/History of artificial intelligence AI , neat and scruffy are two contrasting approaches to AI research. The distinction was made in the 1970s, and was a subject of discussion until the mid-1980s. 1 cite note-FOOTNOTEMcCorduck2004421–424, 486–489-1 2 cite note-FOOTNOTECrevier1993168-2 3 cite note-FOOTNOTENilsson198310–11-3 "Neats" use algorithms based on a single formal paradigm, such as logic https://en.wikipedia.org/wiki/Logic , mathematical optimization https://en.wikipedia.org/wiki/Mathematical optimization , or neural networks https://en.wikipedia.org/wiki/Neural network machine learning . Neats verify their programs are correct via rigorous mathematical theory. Neat researchers and analysts tend to express the hope that this single formal paradigm can be extended and improved in order to achieve general intelligence https://en.wikipedia.org/wiki/Artificial general intelligence and superintelligence https://en.wikipedia.org/wiki/Superintelligence . "Scruffies" use any number of different algorithms and methods to achieve intelligent behavior, and rely on incremental testing to verify their programs. Scruffy programming requires large amounts of hand coding https://en.wikipedia.org/wiki/Hand coding and knowledge engineering https://en.wikipedia.org/wiki/Knowledge engineering . Scruffy experts have argued that general intelligence can only be implemented by solving a large number of essentially unrelated problems, and that there is no silver bullet https://en.wikipedia.org/wiki/Silver bullet that will allow programs to develop general intelligence autonomously. John Brockman https://en.wikipedia.org/wiki/John Brockman literary agent compares the neat approach to physics https://en.wikipedia.org/wiki/Physics , in that it uses simple mathematical models as its foundation. The scruffy approach is more biological https://en.wikipedia.org/wiki/Biology , in that much of the work involves studying and categorizing diverse phenomena. a cite note-chomsky-5 Modern AI has elements of both scruffy and neat approaches. Scruffy AI researchers in the 1990s applied mathematical rigor to their programs, as neat experts did. clarification needed https://en.wikipedia.org/wiki/Wikipedia:Please clarify 5 cite note-FOOTNOTERussellNorvig202124-6 They also express the hope that there is a single paradigm a "master algorithm" that will cause general intelligence and superintelligence to emerge. 6 cite note-FOOTNOTEMcCorduck2004487-7 But modern AI also resembles the scruffies: 7 cite note-FOOTNOTEDomingos2015-8 modern 8 cite note-FOOTNOTERussellNorvig202126-9 machine learning https://en.wikipedia.org/wiki/Machine learning applications require a great deal of hand-tuning and incremental testing; while the general algorithm is mathematically rigorous, accomplishing the specific goals of a particular application is not. Origin in the 1970s edit /w/index.php?title=Neats and scruffies&action=edit§ion=1 The distinction between neat and scruffy originated in the mid-1970s, by Roger Schank https://en.wikipedia.org/wiki/Roger Schank . Schank used the terms to characterize the difference between his work on natural language processing https://en.wikipedia.org/wiki/Natural language processing which represented commonsense knowledge https://en.wikipedia.org/wiki/Commonsense knowledge artificial intelligence in the form of large amorphous semantic networks https://en.wikipedia.org/wiki/Semantic networks from the work of John McCarthy https://en.wikipedia.org/wiki/John McCarthy computer scientist , Allen Newell https://en.wikipedia.org/wiki/Allen Newell , Herbert A. Simon https://en.wikipedia.org/wiki/Herbert A. Simon , Robert Kowalski https://en.wikipedia.org/wiki/Robert Kowalski and others whose work was based on logic and formal extensions of logic. 2 Schank described himself as an AI scruffy. He made this distinction in linguistics, arguing strongly against Chomsky's view of language. a cite note-chomsky-5 The distinction was also partly geographical and cultural: "scruffy" attributes were exemplified by AI research at MIT https://en.wikipedia.org/wiki/MIT under Marvin Minsky https://en.wikipedia.org/wiki/Marvin Minsky in the 1970s. The laboratory was famously "freewheeling" and researchers often developed AI programs by spending long hours fine-tuning programs until they showed the required behavior. Important and influential "scruffy" programs developed at MIT included citation needed https://en.wikipedia.org/wiki/Wikipedia:Citation needed Joseph Weizenbaum https://en.wikipedia.org/wiki/Joseph Weizenbaum 's ELIZA https://en.wikipedia.org/wiki/ELIZA , which behaved as if it spoke English, without any formal knowledge at all, and Terry Winograd https://en.wikipedia.org/wiki/Terry Winograd 's b cite note-11 SHRDLU https://en.wikipedia.org/wiki/SHRDLU , which could successfully answer queries and carry out actions in a simplified world consisting of blocks and a robot arm. 10 cite note-FOOTNOTECrevier199384−102-12 SHRDLU, while successful, could not be scaled up into a useful natural language processing system, because it lacked a structured design. Maintaining a larger version of the program proved to be impossible, i.e. it was too scruffy to be extended. 11 cite note-FOOTNOTERussellNorvig202120-13 citation needed https://en.wikipedia.org/wiki/Wikipedia:Citation needed Other AI laboratories of which the largest were Stanford https://en.wikipedia.org/wiki/Stanford , Carnegie Mellon University https://en.wikipedia.org/wiki/Carnegie Mellon University and the University of Edinburgh https://en.wikipedia.org/wiki/University of Edinburgh focused on logic and formal problem solving as a basis for AI. These institutions supported the work of John McCarthy, Herbert Simon, Allen Newell, Donald Michie https://en.wikipedia.org/wiki/Donald Michie , Robert Kowalski, and other "neats". The contrast between MIT https://en.wikipedia.org/wiki/MIT 's approach and other laboratories was also described as a "procedural/declarative distinction". by whom? https://en.wikipedia.org/wiki/Wikipedia:Manual of Style/Words to watch Unsupported attributions Programs like SHRDLU were designed as agents that carried out actions. They executed "procedures". Other programs were designed as inference engines that manipulated formal statements or "declarations" about the world and translated these manipulations into actions. citation needed https://en.wikipedia.org/wiki/Wikipedia:Citation needed In his 1983 presidential address to Association for the Advancement of Artificial Intelligence https://en.wikipedia.org/wiki/Association for the Advancement of Artificial Intelligence , Nils Nilsson https://en.wikipedia.org/wiki/Nils Nilsson researcher discussed the issue, arguing that "the field needed both". He wrote "much of the knowledge we want our programs to have can and should be represented declaratively in some kind of declarative, logic-like formalism. Ad hoc structures have their place, but most of these come from the domain itself." Alex P. Pentland and Martin Fischler of SRI International https://en.wikipedia.org/wiki/SRI International concurred about the anticipated role of deduction and logic-like formalisms in future AI research, but not to the extent that Nilsson described. 12 cite note-14 Scruffy projects in the 1980s edit /w/index.php?title=Neats and scruffies&action=edit§ion=2 The scruffy approach was applied to robotics by Rodney Brooks https://en.wikipedia.org/wiki/Rodney Brooks in the mid-1980s. He advocated building robots that were, as he put it, Fast, Cheap and Out of Control https://en.wikipedia.org/wiki/Fast, Cheap and Out of Control , the title of a 1989 paper co-authored with Anita Flynn. Unlike earlier robots such as Shakey https://en.wikipedia.org/wiki/Shakey the robot or the Stanford cart, they did not build up representations of the world by analyzing visual information with algorithms drawn from mathematical machine learning https://en.wikipedia.org/wiki/Machine learning techniques, and they did not plan their actions using formalizations based on logic, such as the ' Planner https://en.wikipedia.org/wiki/Planner programming language ' language. They simply reacted to their sensors in a way that tended to help them survive and move. 13 cite note-FOOTNOTEMcCorduck2004454–459-15 Douglas Lenat https://en.wikipedia.org/wiki/Douglas Lenat 's Cyc https://en.wikipedia.org/wiki/Cyc project was initiated in 1984 https://en.wikipedia.org/wiki/Cyc Overview , one of the earliest and most ambitious projects to capture all of human knowledge in machine readable form; it was described as "a determinedly scruffy enterprise". 14 The Cyc database contains millions of facts about the world, each of which are used to add to the complexity of the overall system, in an attempt to create AI that can understand common-sense knowledge https://en.wikipedia.org/wiki/Commonsense knowledge artificial intelligence . The Society of Mind edit /w/index.php?title=Neats and scruffies&action=edit§ion=3 In 1986 Marvin Minsky https://en.wikipedia.org/wiki/Marvin Minsky published The Society of Mind which advocated a view of intelligence https://en.wikipedia.org/wiki/Intelligence and the mind https://en.wikipedia.org/wiki/Mind as an interacting community of modules https://en.wikipedia.org/wiki/Modularity or agents https://en.wikipedia.org/wiki/Intelligent agent that each handled different aspects of cognition, where some modules were specialized for very specific tasks e.g. edge detection https://en.wikipedia.org/wiki/Edge detection in the visual cortex and other modules were specialized to manage communication and prioritization e.g. planning https://en.wikipedia.org/wiki/Planning and attention https://en.wikipedia.org/wiki/Attention in the frontal lobes . Minsky presented this paradigm as a model of both biological human intelligence and as a blueprint for future work in AI. This paradigm is explicitly "scruffy" in that it does not expect there to be a single algorithm that can be applied to all of the tasks involved in intelligent behavior. 15 Minsky wrote: What magical trick makes us intelligent? The trick is that there is no trick. The power of intelligence stems from our vast diversity, not from any single, perfect principle. 16 As of 1991, Minsky was still publishing papers evaluating the relative advantages of the neat versus scruffy approaches, e.g. “Logical Versus Analogical or Symbolic Versus Connectionist or Neat Versus Scruffy”. 17 cite note-FOOTNOTELehnert1994-19 Modern AI as both neat and scruffy edit /w/index.php?title=Neats and scruffies&action=edit§ion=4 New statistical https://en.wikipedia.org/wiki/Artificial intelligence Statistical and mathematical approaches to AI were developed in the 1990s, using highly developed formalisms such as mathematical optimization https://en.wikipedia.org/wiki/Optimization mathematics and neural networks https://en.wikipedia.org/wiki/Artificial neural network . Pamela McCorduck https://en.wikipedia.org/wiki/Pamela McCorduck wrote that "As I write, AI enjoys a Neat hegemony, people who believe that machine intelligence, at least, is best expressed in logical, even mathematical terms." 6 This general trend towards more formal methods in AI was described as "the victory of the neats" by Peter Norvig https://en.wikipedia.org/wiki/Peter Norvig and Stuart Russell https://en.wikipedia.org/wiki/Stuart J. Russell in 2003. 18 cite note-FOOTNOTERussellNorvig200325−26-20 However, by 2021, Russell and Norvig had changed their minds. 19 Deep learning networks and machine learning in general require extensive fine tuning -- they must be iteratively tested until they begin to show the desired behavior. This is a scruffy methodology. Well-known examples edit /w/index.php?title=Neats and scruffies&action=edit§ion=5 Neats Scruffies See also edit /w/index.php?title=Neats and scruffies&action=edit§ion=6 Notes edit /w/index.php?title=Neats and scruffies&action=edit§ion=7 1 cite ref-chomsky 5-0 2 cite ref-chomsky 5-1 John Brockman https://en.wikipedia.org/wiki/John Brockman literary agent writes "Chomsky has always adopted the physicist's philosophy of science, which is that you have hypotheses you check out, and that you could be wrong. This is absolutely antithetical to the AI philosophy of science, which is much more like the way a biologist looks at the world. The biologist's philosophy of science says that human beings are what they are, you find what you find, you try to understand it, categorize it, name it, and organize it. If you build a model and it doesn't work quite right, you have to fix it. It's much more of a "discovery" view of the world." 4 cite note-FOOTNOTEBrockman1996 httpswwwedgeorgconversationinformation-is-surprises Chapter 9: Information is Surprises -4 ↑ cite ref-11 Winograd also became a critic of early approaches to AI as well, arguing that intelligent machines could not be built using formal symbols exclusively, but required embodied cognition https://en.wikipedia.org/wiki/Embodied cognition . 9 cite note-FOOTNOTEWinogradFlores1986-10 Citations edit /w/index.php?title=Neats and scruffies&action=edit§ion=8 ↑ cite ref-FOOTNOTEMcCorduck2004421–424, 486–489 1-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 421–424, 486–489. 1 cite ref-FOOTNOTECrevier1993168 2-0 2 cite ref-FOOTNOTECrevier1993168 2-1 Crevier 1993 CITEREFCrevier1993 , p. 168. ↑ cite ref-FOOTNOTENilsson198310–11 3-0 Nilsson 1983 CITEREFNilsson1983 , pp. 10–11. ↑ cite ref-FOOTNOTEBrockman1996 httpswwwedgeorgconversationinformation-is-surprises Chapter 9: Information is Surprises 4-0 Brockman 1996 CITEREFBrockman1996 , Chapter 9: Information is Surprises https://www.edge.org/conversation/information-is-surprises . ↑ cite ref-FOOTNOTERussellNorvig202124 6-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 24. 1 cite ref-FOOTNOTEMcCorduck2004487 7-0 2 cite ref-FOOTNOTEMcCorduck2004487 7-1 McCorduck 2004 CITEREFMcCorduck2004 , p. 487. ↑ cite ref-FOOTNOTEDomingos2015 8-0 Domingos 2015 CITEREFDomingos2015 . ↑ cite ref-FOOTNOTERussellNorvig202126 9-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 26. ↑ cite ref-FOOTNOTEWinogradFlores1986 10-0 Winograd & Flores 1986 CITEREFWinogradFlores1986 . ↑ cite ref-FOOTNOTECrevier199384−102 12-0 Crevier 1993 CITEREFCrevier1993 , pp. 84−102. ↑ cite ref-FOOTNOTERussellNorvig202120 13-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 20. ↑ cite ref-14 Pentland and Fischler 1983, quoted in McCorduck 2004 CITEREFMcCorduck2004 , pp. 421–424 ↑ cite ref-FOOTNOTEMcCorduck2004454–459 15-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 454–459. ↑ cite ref-FOOTNOTEMcCorduck2004489 16-0 McCorduck 2004 CITEREFMcCorduck2004 , p. 489. ↑ cite ref-FOOTNOTECrevier1993254 17-0 Crevier 1993 CITEREFCrevier1993 , p. 254. ↑ cite ref-FOOTNOTEMinsky1986308 18-0 Minsky 1986 CITEREFMinsky1986 , p. 308. ↑ cite ref-FOOTNOTELehnert1994 19-0 Lehnert 1994 CITEREFLehnert1994 . ↑ cite ref-FOOTNOTERussellNorvig200325−26 20-0 Russell & Norvig 2003 CITEREFRussellNorvig2003 , p. 25−26. ↑ cite ref-FOOTNOTERussellNorvig202123 21-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 23. References edit /w/index.php?title=Neats and scruffies&action=edit§ion=9 - Brockman, John 7 May 1996 . Third Culture: Beyond the Scientific Revolution . Simon and Schuster. ISBN https://en.wikipedia.org/wiki/ISBN identifier 978-0684823447 https://en.wikipedia.org/wiki/Special:BookSources/978-0684823447 Crevier, Daniel https://en.wikipedia.org/wiki/Daniel Crevier 1993 . AI: The Tumultuous Search for Artificial Intelligence . New York, NY: BasicBooks. ISBN https://en.wikipedia.org/wiki/ISBN identifier 0-465-02997-3 https://en.wikipedia.org/wiki/Special:BookSources/0-465-02997-3 .. Domingos, Pedro https://en.wikipedia.org/wiki/Pedro Domingos 22 September 2015 . The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World . Basic Books https://en.wikipedia.org/wiki/Basic Books . ISBN https://en.wikipedia.org/wiki/ISBN identifier 978-0465065707 https://en.wikipedia.org/wiki/Special:BookSources/978-0465065707 .- Lehnert, Wendy C. 1 May 1994 . "5: Cognition, Computers, and Car Bombs: How Yale Prepared Me for the 90's". In Schank, Robert; Langer, Ellen eds. . Beliefs, Reasoning, and Decision Making: Psycho-Logic in Honor of Bob Abelson First ed. . New York, NY: Taylor & Francis Group. p. 150. doi https://en.wikipedia.org/wiki/Doi identifier : 10.4324/9780203773574 https://doi.org/10.4324%2F9780203773574 . ISBN https://en.wikipedia.org/wiki/ISBN identifier 9781134781621 https://en.wikipedia.org/wiki/Special:BookSources/9781134781621 . - Minsky, Marvin 1986 . . New York: Simon & Schuster. The Society of Mind ISBN https://en.wikipedia.org/wiki/ISBN identifier 0-671-60740-5 https://en.wikipedia.org/wiki/Special:BookSources/0-671-60740-5 . McCorduck, Pamela https://en.wikipedia.org/wiki/Pamela McCorduck 2004 , Machines Who Think 2nd ed. , Natick, Massachusetts: A. K. Peters, ISBN https://en.wikipedia.org/wiki/ISBN identifier 1-5688-1205-1 https://en.wikipedia.org/wiki/Special:BookSources/1-5688-1205-1 . Nilsson, Nils https://en.wikipedia.org/wiki/Nils Nilsson researcher 1983 . "Artificial Intelligence Prepares for 2001" https://ai.stanford.edu/~nilsson/OnlinePubs-Nils/General%20Essays/AIMag04-04-002.pdf PDF . AI Magazine . 1 1 . Archived https://web.archive.org/web/20200817194457/http://ai.stanford.edu/~nilsson/OnlinePubs-Nils/General%20Essays/AIMag04-04-002.pdf PDF from the original on 17 August 2020. Retrieved 22 August 2020. Presidential Address to the Association for the Advancement of Artificial Intelligence https://en.wikipedia.org/wiki/Association for the Advancement of Artificial Intelligence Russell, Stuart J. https://en.wikipedia.org/wiki/Stuart J. Russell ; Norvig, Peter https://en.wikipedia.org/wiki/Peter Norvig 2003 . Artificial Intelligence: A Modern Approach 2nd ed. . Upper Saddle River, New Jersey: Prentice Hall. ISBN https://en.wikipedia.org/wiki/ISBN identifier 0-13-790395-2 https://en.wikipedia.org/wiki/Special:BookSources/0-13-790395-2 . Russell, Stuart J. https://en.wikipedia.org/wiki/Stuart J. Russell ; Norvig, Peter https://en.wikipedia.org/wiki/Peter Norvig 2021 . 4th ed. . Hoboken: Pearson. Artificial Intelligence: A Modern Approach https://en.wikipedia.org/wiki/Artificial Intelligence: A Modern Approach ISBN https://en.wikipedia.org/wiki/ISBN identifier 9780134610993 https://en.wikipedia.org/wiki/Special:BookSources/9780134610993 . LCCN https://en.wikipedia.org/wiki/LCCN identifier 20190474 https://lccn.loc.gov/20190474 .- Winograd, Terry; Flores 1986 . Understanding Computers and Cognition: A New Foundation for Design . Ablex Publ Corp. Further reading edit /w/index.php?title=Neats and scruffies&action=edit§ion=10 - Anderson, John R. 2005 . "Human symbol manipulation within an integrated cognitive architecture" https://doi.org/10.1207%2Fs15516709cog0000 22 . Cognitive Science . 29 3 : 313–341. doi https://en.wikipedia.org/wiki/Doi identifier : 10.1207/s15516709cog0000 22 https://doi.org/10.1207%2Fs15516709cog0000 22 . PMID https://en.wikipedia.org/wiki/PMID identifier 21702777 https://pubmed.ncbi.nlm.nih.gov/21702777 . - Brooks, Rodney A. 2001-01-18 . "The Relationship Between Matter and Life" https://doi.org/10.1038%2F35053196 . Nature . 409 6818 : 409–411. Bibcode https://en.wikipedia.org/wiki/Bibcode identifier : 2001Natur.409..409B https://ui.adsabs.harvard.edu/abs/2001Natur.409..409B . doi https://en.wikipedia.org/wiki/Doi identifier : 10.1038/35053196 https://doi.org/10.1038%2F35053196 . PMID https://en.wikipedia.org/wiki/PMID identifier 11201756 https://pubmed.ncbi.nlm.nih.gov/11201756 . S2CID https://en.wikipedia.org/wiki/S2CID identifier 4430614 https://api.semanticscholar.org/CorpusID:4430614 . - Poirier, Lindsay 2025 . "Neat vs. Scruffy: How Early AI Researchers Classified Epistemic Cultures of Knowledge Representation" https://doi.org/10.1109%2FMAHC.2024.3498692 . IEEE Annals of the History of Computing . 47 2 : 7–19. doi https://en.wikipedia.org/wiki/Doi identifier : 10.1109/MAHC.2024.3498692 https://doi.org/10.1109%2FMAHC.2024.3498692 .