This is the
talk page for discussing improvements to the
Von Neumann universal constructor article. This is not a forum for general discussion of the article's subject. |
Article policies
|
Find sources: Google ( books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL |
Archives: 1 |
![]() | This article is rated B-class on Wikipedia's
content assessment scale. It is of interest to the following WikiProjects: | ||||||||||
|
He observed that -
This citation is probably best derived from the writing contained within the text due Jeffress, Cerebral Mechanisms in Behavior - The Hixon Symposium: pages 30-31 contain the text "typical traits which appear in connection with mutation, leathally as a rule, but with a possibility of continuing reproduction with a modification of traits." Clearly, the italicised portion of this quote is a satisfactory definition for the term open-ended evolution. William R. Buckley 20:17, 11 March 2007 (UTC)
What is this? Is this a turing-machine in a celular automata? Could someone explain this further? 84.59.213.173 ( talk) 01:34, 31 January 2008 (UTC)
These two articles ought to be merged because they both discuss the same subject matter. —Preceding unsigned comment added by Maverick1701 ( talk • contribs) 17:42, 1 March 2008 (UTC)
In the proceedings volume for the conference Automata 2008, which is searchable on Amazon at the URL
http://www.amazon.com/Automata-2008-Theory-Applications-Cellular-Automata/dp/1905986165/ref=sr_1_1?ie=UTF8&s=books&qid=1211820017&sr=8-1, see page 501
the topic of von Neumann self-replicators is discussed. One of the conclusions presented in the paper is the fact of insufficiency, vis-à-vis self-replication, of the Pesavento design, as presented in the journal Artificial Life. The point is that one cell of the configuration cannot be constructed by the configuration, and so it is not a self-replicator. It would seem to me that use of the Pesavento design as an example self-replicator is to the detriment of Wikipedia. William R. Buckley ( talk) 16:52, 26 May 2008 (UTC)
Ferkel: Is there no means to obtain better quality images; something that shows individual cells sufficient to determine state? William R. Buckley ( talk) 03:37, 16 August 2008 (UTC)
In your work on Pesavento, has it occurred to you that other, quite simple means exist to both render the design a self-replicator, and without need to rely upon collision of the construction arm with tape as a means to avoid the otherwise troublesome matter of constructing a certain cell? Indeed, all one need do is to extend by quite some time the delay observed by the auto-retract signal, such that it is serviced after service of the corresponding construction signal; alter the visitation order of these signals upon the construction arm, and the problem disappears. I think this is more elegant a mechanism, than is the mechanism which the Pesavento design employs. William R. Buckley ( talk) 03:48, 16 August 2008 (UTC)
It should be noted that the reasoning for the declaration in Mange et al. is that the configuration there mentioned as being *consistent with the design of von Neumann* is so judged because it uses only the 29-states of von Neumann. William R. Buckley ( talk) 04:44, 16 August 2008 (UTC)
This is to say, I do not expect that Mange et al. were intent to claim the architecture of the design is even minimally identical to that contemplated by von Neumann. Indeed, architecturally they are not at all related, even though we might forget for the moment that von Neumann's architecture is only half complete; half the details are missing. William R. Buckley ( talk) 13:08, 9 September 2008 (UTC)
Declaring NP32 the first universal constructor is perhaps going a bit too far, though it may require a bit of clever wording to correct. My complaint is that NP32 is the first implemented UC for a CA. It may be that other programs have expressed universal construction, though the fact is not recognised. I am thinking of compilers, and viruses, etc. William R. Buckley ( talk) 04:52, 16 August 2008 (UTC)
It will require great care to properly handle the misconceptions implied by the simple definition of universal construction. Part of the reason is that no vNCA Garden of Eden configuration is passive. The best analogy comes from Douglas Hofstadter, who in Godel, Escher, and Bach: An Eternal Golden Braid mentions the phonograph player, which cannot play all possible phonographs: at least some of those phonographs will cause the player to break (think of standing waves in the player, whose energy increases by constructive interference with the waves emitted by the phonograph upon scratching it with a needle); there is no phonograph player that can play all possible phonographs. Similarly, there is no universal constructor; all constructors are limited in their ability to construct, and this includes configurations which are not GoE's. The key element in constructibility is the presence and nature of signal which is external to the constructor. William R. Buckley ( talk) 18:10, 17 August 2008 (UTC)
Of course, the better definition is like that which applies to the mechanism of computation: a UC (UTM) can construct (compute) any constructible (computable) construction (computation) that can be constructed (computed). William R. Buckley ( talk) 08:07, 21 August 2008 (UTC)
there are some assertions that VN's machine guarantees the possibility of mutation and evolution by natural selection. this seems like a pretty silly claim ... although a replicator with a changeable information component is necessary for evolution, it seems quite reasonable to suppose you can have a brittle self-replicator that won't function with any changes. The important fact seems to be that VN has proved the possibility of self-replicators at all, and the dubious assertion that this is a robust replicator capable of evolution, while interesting, seems to detract from the important point rather than improve the article. comments? 131.172.99.15 ( talk) 03:02, 21 August 2008 (UTC)snaxalotl
The implied minority of viable mutations falsifies the absolute negative in the last sentence. I'll try to find a cite-able reference to support my objection, then I'll include that. Voland0 ( talk) 00:49, 19 October 2018 (UTC)
VN (see page 85 and beyond of fifth lecture in Theory of Self-Reproducing Automata), very clearly talks about evolution that can be robust in his conceptual scheme. This is done by introducing an Automaton D which is not involved in the basic function of self-replication and whose function does not disturb the basic performance of the universal constructor (A) + universal copier (B) + operating system (C). If there is a mutation in the D part of the description, say D`, then the system (A + B + C + D) + Φ(A + B + C + D`) will produce (A + B + C + D`) + Φ(A + B + C + D`). VN further proposed that non-trivial self-reproduction should include this “ability to undergo inheritable mutations as well as the ability to make another organism like the original”, to distinguish it from “naive” self-reproduction like growing crystals. So the basic self-replication scheme (A + B + C) + Φ(A + B + C) should be unaffected by D, which can be seen as the extra functioning of an organism's phenotype---all the things that make it more or less fit and thus selected for. See references in text for Brenner and Rocha. I added materials to help make this more clear in article. -- PlaceboOracle ( talk) 18:38, 14 May 2020 (UTC)
It is interesting that while the concept of configuration size minimisation is discussed, especially considering the emphasis given Run-Length Limited encoding, yet it seems that such mechanisms do not yield the smallest configurations, nor the shortest tapes. NCA configurations having 3343 cells, a tape of 44,150 cells, for 8830 5-bit instructions, and 3574 cells in the configuration, 37,780 cells in the tape, for 9445 4-bit instructions are known. William R. Buckley ( talk) 05:54, 15 September 2008 (UTC)
It is incorrect to call the NP29 configuration a universal constructor. Indeed, none of the published *self-replicators* for vNCA (29 states)and NCA (32 states) is a universal constructor. Not one of them can construct, for instance, the Real-Time Crossing Organ, as given by Gorman. Properly, these configurations are all general constructors; they may construct any passive configuration.
Not only have I given two vNCA configurations which are self-replicators, I have also given a total of four NCA configurations which are also self-replicators; one of the latter is the quickest self-replicator for NCA, which has both nearly smallest corpus (a sister configuration is about 100 - 200 cells smaller) and the shortest tape yet given. Use of Golly will quickly reveal the two flathead designs. I am still waiting for a reply from Nobili, as to whether or not he can give a smaller configuration, or a shorter tape; one expects that time to self-replication is dependent upon these two parameters.
One final note. In all of this work, there has been a tacit acceptance of von Neumann's assertion that self-replication is but a special case of universal construction. This is not correct, in my view. Indeed, one can demonstrate that self-replication is not possible given possession of a universal constructor (universal over passive configurations). To see this, review the paper Computational Ontogeny, Biological Theory, Vol 3 No 1. I state here the intention to demonstrate same. I will do it with the partial constructor; this is the configuration mentioned by Mange. William R. Buckley ( talk) 05:33, 16 November 2008 (UTC)
I excised this phrase "and its offspring are not self-replicable" because it is ambiguous.
One may construct a tape of the 29-state configuration, and then the statement makes sense.
Except in the following case. It is possible to construct a description (D) for machine A, which includes a description of the tape which machine A needs in order to self-replicate. In that case, construction from D does not constitute self-replication. This is true even if machine A is the machine which constructs according to the instructions in D. Here, the offspring may well be self-replicable, even if their construction does not constitute an act of self-replication. The reason is that machines derived from D will inherit tape A, not tape D.
The words *offspring* *daughter* and *construct* are synonyms for the product of construction. I would suggest that the phrase is better expressed as "and its replicants are not self-replicable" to distinguish the nature of the construct - does it describe the machine which interprets the tape, or a machine other than the one which interprets the tape (which would be a construct).
The language for this distinction in cellular automata is not well developed. William R. Buckley ( talk) 23:13, 22 August 2009 (UTC)
From my vague memories of a class in Artificial Life I took almost a decade ago, I understood this construction of von Neumann to be an attempt to produce a self-replicating Turing machine; this article appears to support that view. Could the wiki article detail how accurate that description is, i.e. how well did von Neumann fare in his attempt? Pcap ping 23:12, 1 September 2009 (UTC)
It is not exactly true to state that the traditional view of von Neumann's work in machine self-replication is that of the logical requirements for same. Indeed, it is generally understood that Hixon addressed the logical requirements for non-trivial self-replication. The opening paragraph of the section Purpose is therefore poorly constructed; Codd retained this orientation (as I recall, it is explicitly stated in his doctoral thesis), and the works of later researchers (such as Langton) were intended to address the minimisation of configuration, and simplification of underlying cellular automata, with the concomitant retention of non-triviality. After all, one can go all the way to Ulam to see that self-replication is trivially achieved in some systems. William R. Buckley ( talk) 11:18, 24 November 2009 (UTC)
Where do the numbers in the Table - Comparison of implementations - and in particular the number of time-steps for different implementations come from? Are there any peer-reviewed publications associated with these? The publications of Pesavento and Nobili that are linked from the article do not discuss practical implementations from what I can see.
Alvis Brazma (brazma@ebi.ac.uk) — Preceding unsigned comment added by 193.62.194.241 ( talk) 10:33, 7 October 2013 (UTC)
Can somebody provide a download for phi9.rle and codon3.rle? 76.232.22.106 ( talk) 18:43, 8 July 2018 (UTC)
Forum etc. —DIYeditor ( talk) 17:15, 28 August 2018 (UTC) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
The following discussion has been closed. Please do not modify it. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
I am constructing a machine that is capable of constructing any passive configuration and computing any μ-recursive function. I have estimated that it will take at least 1.3 million cells, not including any tapes. It is not an attempt to create the smallest universal computer-constructor. A configuration this large would take inordinate amounts of time and space to replicate, not to mention that it has 4 tapes. Here’s how it might replicate:
The tape uses full run-length encoding, *NOT* run-length *limited* encoding. This is mostly to minimize the length of the tape, because an overly long tape can cause slowdown. I might make a smaller one later. 76.232.22.106 ( talk) 16:08, 25 August 2018 (UTC) My comparison of implementationsNote: May include unreliable information
|
I removed the "Constructor theory" wikilink from the See Also section. There is no published evidence of any connection between the Deutsch/Maletto Constructor Theory and von Neumann's Universal Constructor, or indeed any other concept in the field of cellular automata. If an editor wishes to add it back in, a reliable reference is needed to comply with Wikipedia policies, principally WP:NOR. David Spector ( talk) 18:54, 12 September 2022 (UTC) David Spector ( talk) 18:54, 12 September 2022 (UTC)
This is the
talk page for discussing improvements to the
Von Neumann universal constructor article. This is not a forum for general discussion of the article's subject. |
Article policies
|
Find sources: Google ( books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL |
Archives: 1 |
![]() | This article is rated B-class on Wikipedia's
content assessment scale. It is of interest to the following WikiProjects: | ||||||||||
|
He observed that -
This citation is probably best derived from the writing contained within the text due Jeffress, Cerebral Mechanisms in Behavior - The Hixon Symposium: pages 30-31 contain the text "typical traits which appear in connection with mutation, leathally as a rule, but with a possibility of continuing reproduction with a modification of traits." Clearly, the italicised portion of this quote is a satisfactory definition for the term open-ended evolution. William R. Buckley 20:17, 11 March 2007 (UTC)
What is this? Is this a turing-machine in a celular automata? Could someone explain this further? 84.59.213.173 ( talk) 01:34, 31 January 2008 (UTC)
These two articles ought to be merged because they both discuss the same subject matter. —Preceding unsigned comment added by Maverick1701 ( talk • contribs) 17:42, 1 March 2008 (UTC)
In the proceedings volume for the conference Automata 2008, which is searchable on Amazon at the URL
http://www.amazon.com/Automata-2008-Theory-Applications-Cellular-Automata/dp/1905986165/ref=sr_1_1?ie=UTF8&s=books&qid=1211820017&sr=8-1, see page 501
the topic of von Neumann self-replicators is discussed. One of the conclusions presented in the paper is the fact of insufficiency, vis-à-vis self-replication, of the Pesavento design, as presented in the journal Artificial Life. The point is that one cell of the configuration cannot be constructed by the configuration, and so it is not a self-replicator. It would seem to me that use of the Pesavento design as an example self-replicator is to the detriment of Wikipedia. William R. Buckley ( talk) 16:52, 26 May 2008 (UTC)
Ferkel: Is there no means to obtain better quality images; something that shows individual cells sufficient to determine state? William R. Buckley ( talk) 03:37, 16 August 2008 (UTC)
In your work on Pesavento, has it occurred to you that other, quite simple means exist to both render the design a self-replicator, and without need to rely upon collision of the construction arm with tape as a means to avoid the otherwise troublesome matter of constructing a certain cell? Indeed, all one need do is to extend by quite some time the delay observed by the auto-retract signal, such that it is serviced after service of the corresponding construction signal; alter the visitation order of these signals upon the construction arm, and the problem disappears. I think this is more elegant a mechanism, than is the mechanism which the Pesavento design employs. William R. Buckley ( talk) 03:48, 16 August 2008 (UTC)
It should be noted that the reasoning for the declaration in Mange et al. is that the configuration there mentioned as being *consistent with the design of von Neumann* is so judged because it uses only the 29-states of von Neumann. William R. Buckley ( talk) 04:44, 16 August 2008 (UTC)
This is to say, I do not expect that Mange et al. were intent to claim the architecture of the design is even minimally identical to that contemplated by von Neumann. Indeed, architecturally they are not at all related, even though we might forget for the moment that von Neumann's architecture is only half complete; half the details are missing. William R. Buckley ( talk) 13:08, 9 September 2008 (UTC)
Declaring NP32 the first universal constructor is perhaps going a bit too far, though it may require a bit of clever wording to correct. My complaint is that NP32 is the first implemented UC for a CA. It may be that other programs have expressed universal construction, though the fact is not recognised. I am thinking of compilers, and viruses, etc. William R. Buckley ( talk) 04:52, 16 August 2008 (UTC)
It will require great care to properly handle the misconceptions implied by the simple definition of universal construction. Part of the reason is that no vNCA Garden of Eden configuration is passive. The best analogy comes from Douglas Hofstadter, who in Godel, Escher, and Bach: An Eternal Golden Braid mentions the phonograph player, which cannot play all possible phonographs: at least some of those phonographs will cause the player to break (think of standing waves in the player, whose energy increases by constructive interference with the waves emitted by the phonograph upon scratching it with a needle); there is no phonograph player that can play all possible phonographs. Similarly, there is no universal constructor; all constructors are limited in their ability to construct, and this includes configurations which are not GoE's. The key element in constructibility is the presence and nature of signal which is external to the constructor. William R. Buckley ( talk) 18:10, 17 August 2008 (UTC)
Of course, the better definition is like that which applies to the mechanism of computation: a UC (UTM) can construct (compute) any constructible (computable) construction (computation) that can be constructed (computed). William R. Buckley ( talk) 08:07, 21 August 2008 (UTC)
there are some assertions that VN's machine guarantees the possibility of mutation and evolution by natural selection. this seems like a pretty silly claim ... although a replicator with a changeable information component is necessary for evolution, it seems quite reasonable to suppose you can have a brittle self-replicator that won't function with any changes. The important fact seems to be that VN has proved the possibility of self-replicators at all, and the dubious assertion that this is a robust replicator capable of evolution, while interesting, seems to detract from the important point rather than improve the article. comments? 131.172.99.15 ( talk) 03:02, 21 August 2008 (UTC)snaxalotl
The implied minority of viable mutations falsifies the absolute negative in the last sentence. I'll try to find a cite-able reference to support my objection, then I'll include that. Voland0 ( talk) 00:49, 19 October 2018 (UTC)
VN (see page 85 and beyond of fifth lecture in Theory of Self-Reproducing Automata), very clearly talks about evolution that can be robust in his conceptual scheme. This is done by introducing an Automaton D which is not involved in the basic function of self-replication and whose function does not disturb the basic performance of the universal constructor (A) + universal copier (B) + operating system (C). If there is a mutation in the D part of the description, say D`, then the system (A + B + C + D) + Φ(A + B + C + D`) will produce (A + B + C + D`) + Φ(A + B + C + D`). VN further proposed that non-trivial self-reproduction should include this “ability to undergo inheritable mutations as well as the ability to make another organism like the original”, to distinguish it from “naive” self-reproduction like growing crystals. So the basic self-replication scheme (A + B + C) + Φ(A + B + C) should be unaffected by D, which can be seen as the extra functioning of an organism's phenotype---all the things that make it more or less fit and thus selected for. See references in text for Brenner and Rocha. I added materials to help make this more clear in article. -- PlaceboOracle ( talk) 18:38, 14 May 2020 (UTC)
It is interesting that while the concept of configuration size minimisation is discussed, especially considering the emphasis given Run-Length Limited encoding, yet it seems that such mechanisms do not yield the smallest configurations, nor the shortest tapes. NCA configurations having 3343 cells, a tape of 44,150 cells, for 8830 5-bit instructions, and 3574 cells in the configuration, 37,780 cells in the tape, for 9445 4-bit instructions are known. William R. Buckley ( talk) 05:54, 15 September 2008 (UTC)
It is incorrect to call the NP29 configuration a universal constructor. Indeed, none of the published *self-replicators* for vNCA (29 states)and NCA (32 states) is a universal constructor. Not one of them can construct, for instance, the Real-Time Crossing Organ, as given by Gorman. Properly, these configurations are all general constructors; they may construct any passive configuration.
Not only have I given two vNCA configurations which are self-replicators, I have also given a total of four NCA configurations which are also self-replicators; one of the latter is the quickest self-replicator for NCA, which has both nearly smallest corpus (a sister configuration is about 100 - 200 cells smaller) and the shortest tape yet given. Use of Golly will quickly reveal the two flathead designs. I am still waiting for a reply from Nobili, as to whether or not he can give a smaller configuration, or a shorter tape; one expects that time to self-replication is dependent upon these two parameters.
One final note. In all of this work, there has been a tacit acceptance of von Neumann's assertion that self-replication is but a special case of universal construction. This is not correct, in my view. Indeed, one can demonstrate that self-replication is not possible given possession of a universal constructor (universal over passive configurations). To see this, review the paper Computational Ontogeny, Biological Theory, Vol 3 No 1. I state here the intention to demonstrate same. I will do it with the partial constructor; this is the configuration mentioned by Mange. William R. Buckley ( talk) 05:33, 16 November 2008 (UTC)
I excised this phrase "and its offspring are not self-replicable" because it is ambiguous.
One may construct a tape of the 29-state configuration, and then the statement makes sense.
Except in the following case. It is possible to construct a description (D) for machine A, which includes a description of the tape which machine A needs in order to self-replicate. In that case, construction from D does not constitute self-replication. This is true even if machine A is the machine which constructs according to the instructions in D. Here, the offspring may well be self-replicable, even if their construction does not constitute an act of self-replication. The reason is that machines derived from D will inherit tape A, not tape D.
The words *offspring* *daughter* and *construct* are synonyms for the product of construction. I would suggest that the phrase is better expressed as "and its replicants are not self-replicable" to distinguish the nature of the construct - does it describe the machine which interprets the tape, or a machine other than the one which interprets the tape (which would be a construct).
The language for this distinction in cellular automata is not well developed. William R. Buckley ( talk) 23:13, 22 August 2009 (UTC)
From my vague memories of a class in Artificial Life I took almost a decade ago, I understood this construction of von Neumann to be an attempt to produce a self-replicating Turing machine; this article appears to support that view. Could the wiki article detail how accurate that description is, i.e. how well did von Neumann fare in his attempt? Pcap ping 23:12, 1 September 2009 (UTC)
It is not exactly true to state that the traditional view of von Neumann's work in machine self-replication is that of the logical requirements for same. Indeed, it is generally understood that Hixon addressed the logical requirements for non-trivial self-replication. The opening paragraph of the section Purpose is therefore poorly constructed; Codd retained this orientation (as I recall, it is explicitly stated in his doctoral thesis), and the works of later researchers (such as Langton) were intended to address the minimisation of configuration, and simplification of underlying cellular automata, with the concomitant retention of non-triviality. After all, one can go all the way to Ulam to see that self-replication is trivially achieved in some systems. William R. Buckley ( talk) 11:18, 24 November 2009 (UTC)
Where do the numbers in the Table - Comparison of implementations - and in particular the number of time-steps for different implementations come from? Are there any peer-reviewed publications associated with these? The publications of Pesavento and Nobili that are linked from the article do not discuss practical implementations from what I can see.
Alvis Brazma (brazma@ebi.ac.uk) — Preceding unsigned comment added by 193.62.194.241 ( talk) 10:33, 7 October 2013 (UTC)
Can somebody provide a download for phi9.rle and codon3.rle? 76.232.22.106 ( talk) 18:43, 8 July 2018 (UTC)
Forum etc. —DIYeditor ( talk) 17:15, 28 August 2018 (UTC) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
The following discussion has been closed. Please do not modify it. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
I am constructing a machine that is capable of constructing any passive configuration and computing any μ-recursive function. I have estimated that it will take at least 1.3 million cells, not including any tapes. It is not an attempt to create the smallest universal computer-constructor. A configuration this large would take inordinate amounts of time and space to replicate, not to mention that it has 4 tapes. Here’s how it might replicate:
The tape uses full run-length encoding, *NOT* run-length *limited* encoding. This is mostly to minimize the length of the tape, because an overly long tape can cause slowdown. I might make a smaller one later. 76.232.22.106 ( talk) 16:08, 25 August 2018 (UTC) My comparison of implementationsNote: May include unreliable information
|
I removed the "Constructor theory" wikilink from the See Also section. There is no published evidence of any connection between the Deutsch/Maletto Constructor Theory and von Neumann's Universal Constructor, or indeed any other concept in the field of cellular automata. If an editor wishes to add it back in, a reliable reference is needed to comply with Wikipedia policies, principally WP:NOR. David Spector ( talk) 18:54, 12 September 2022 (UTC) David Spector ( talk) 18:54, 12 September 2022 (UTC)