selforganizing - university of california, san diego

13
_J___l__lu___u_ __u__ t _ __J__________t?_ ptlsçt_at&4ltptd(dtnwobbt_tomltbyl Roreootalatem9lo Ae_6serblto_tmsss__gmunnep_tsewN8ryaTlh_eeolwgapt)anleotsdgstltatmEctetxvshlutrtullttls2_thvedllgtoepbsMsot TaaeeultceJynho__mptvnceacclb_spreeet_todcoolbLoetbelhaoll_tvnadfewnaaretleEf_eancttcrltbT_treptttttwefsbwtbslatw4httsohprsltwblltttolerr4_ToebrttaonHldNsmrbsdfnruhcht Tcdmt_teotrAtFtoasale_asaepouo_un_mlt_EtTe_rcet FthlutlgTeoyedodteydpt Toses_ecll_teeolrauHptsso4ealtnglore_weglvhlssxnuewnl4NfsEutglltchNegosrnntesltseeTvpletooctao_wtlgottsl_ho_2fnrysmoomtRemTya_eognEoyetnotttnoGeceefhlqthutl_tlf4omNeo Ltoeu4Arac_n_p HtmtnomsmtetFo41hNtycrsbnenbreteeletttetrpstelnthynasbobelt2btahnyaaaEaefrhurlgre_lAt_ltReb_st_lstlatlNne_ngcll7hmfbTtttlnt_ttgeGbauhgslhataew_Asstsrsn_otunllro_tcdthneedsouuap

Upload: others

Post on 26-Jun-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: selforganizing - University of California, San Diego

_J___l__lu___u_ __u__ t _ __J__________t?_ ptlsçt_at&4ltptd(dtnwobbt_tomltbyl Roreootalatem9lo Ae_6serblto_tmsss__gmunnep_tsewN8ryaTlh_eeolwgapt)anleotsdgstltatmEctetxvshlutrtullttls2_thvedllgtoepbsMsot TaaeeultceJynho__mptvnceacclb_spreeet_todcoolbLoetbelhaoll_tvnadfewnaaretleEf_eancttcrltbT_treptttttwefsbwtbslatw4httsohprsltwblltttolerr4_ToebrttaonHldNsmrbsdfnruhcht Tcdmt_teotrAtFtoasale_asaepouo_un_mlt_EtTe_rcet FthlutlgTeoyedodteydpt Toses_ecll_teeolrauHptsso4ealtnglore_weglvhlssxnuewnl4NfsEutglltchNegosrnntesltseeTvpletooctao_wtlgottsl_ho_2fnrysmoomtRemTya_eognEoyetnotttnoGeceefhlqthutl_tlf4omNeo Ltoeu4Arac_n_p HtmtnomsmtetFo41hNtycrsbnenbreteeletttetrpstelnthynasbobelt2btahnyaaaEaefrhurlgre_lAt_ltReb_st_lstlatlNne_ngcll7hmfbTtttlnt_ttgeGbauhgslhataew_Asstsrsn_otunllro_tcdthneedsouuapedw_eys_lfl_ag_rslNlfd__settteemhotftbattdpphr_arlhtMh_olle_Nlo__srsrtentn4netholmtotlunvabkmyctynepo Tl_arecosesNloooctelall_tysmnohtprhdvserpdounoelaGecvpepbeaenttlotoNeldl_nsrnobrpeetarnttl_te_ldtoxctanrbhhcgbtatealkT_eeoheNlwsl_om_efcpotnTutmul__wtach_t_ttettl_tshdoalhcolo_uafNtelostgaNltJcloeolrth1rneoye_ylyeyote_fr _;__t______u_____________??__?___??u_u_______??_____uu������� n , __ l ,�� l� _ _ A z __,?������_ is so broa d t hat, were i t to be dea l t w i t h w ho _ ly a t t be t e c b _ o _ o g i c a l _�, or practical level, we wou 1 d be de feate d by t be s beer quan t i ty an d j g _��������������������U�_��=,�,�'U�_�

Page 2: selforganizing - University of California, San Diego

_u__ _ atocdo_a_astenpohgnpI,yet_,ohreaocreeftopmlmcameorpttuarslrtle__chsteteeTtltmnyboao_errtcltttglNcecrouvearr4leostnue_tldh_s4nenbboumyroty,4e_Ndcsnhesauhra_ltmonsnowl_gbcsoea_rlfea1oefasxofttlr_mrcomeInaemst_athe Tgelltgyynmet tn4hhttehall_trcsreaadlcIlw_4ctayloanc_rnude_c1ttue_sopswggttrhl_tecehoaaneftl ___ tmw_ohrheusrantecrtl_eocalsa_ststslNt BontTnhar_ stllt__ntstftt(a_plNfnooTp TlNNnhsrgotoel_mdvbuteeew_nctogaeeuvsnppseeeonarNclt_anwelat_lNhs2tle_tAldhoettfatsa_l_evlnnlol_esfpwteohf)tsehsaoelNrcpb_ecplaeusotrtsehtssvulebbbsnlI_elotltdstIwoll_teamegsme_N_4sneytstwrtolhl_hctctehacaml_uvlnreyf _;__tllt1__t_11ttl_tu_J_?_________n���_ _ n,�l��256 w. Ross nsnBy _ _ _RINcI_LEs oF T_E sELF- o R G A N _ z I N G s y s T E M 2 5 7����Onentoftt OrganlZatlon''lspresent.TbuS t_et_eorYo_Org_MlZa_ioJ1 _, the methods of McG i Il an d Garner wI l l become t he 4ccepte d , J�� _ ' ' t�u_ We can get another angIe on the question by asking ''what is The treatment of "conditionaIity'' (whether by funct ions o f _,�its converse ?'' The converse of ''conditional on'' is ''not condi- many variables, by corre1ation anaIysis, by uncertainty ana Iys is, _, _ _ _�tional on'', so the converse of '_rganization'' must therefore be, or by other ways) makes us realize that the essent ial idea is t hat ! _,��reducibility''. (It is aIso caIIed ''separabiIity''.) TMs occurs, in which some sub-set of points indicates the act4al it ies. T h is way _ _�=_ mathematical forms, when what looks Iike a function of several of looking at ''con_itionaIity'' ma_es __s reaIize that it is related ! '_ _�vaTiables (perhaps very many) proves on cIoser examination to - to that of ''communication'' ; and it is, of course, qu ite p Iaus i b Ie t�U bave parts whose act_'ons are Mo_ conditiona1 on the values or the that we shouId de_ne parts as being 'coTganized't when t'com- i,��u wbat looks li_e one machine proves to be composed of two (o_ (Again the natura1 converse is that of independence, which�more) sub-machines, each of w_ch is acting independentIy oF represents non-communication.)�the otheTs. , Now ''comn_unication'' from A to B necessar i Iy imp 1 ies some ,�Questionsof''conditionality'',andofitsconverse''reducibility'; constraint, some correlation between what happens at A an d ,��methods. I shaI1 say something of such methods later. Here, ' at B, then there is no communic4tion from A to B and no con- ,�bowever, I would 1ike to express the opinion tbat the method of straint over the possible (A, B)-couples that can occ4r. Thus tbe !�Uncertainty AnaIysis, introduced by Garner and McGiIl (l956), presence of ''organization'' between variables is eq4ivalent to the _ _,�gives us a metbod for the treatment of conditionaIity that is not existence of 4 ço_strai_t in the pToduct-space o f t he poss i b i I it ies. ),��generaIity and suitabiIity for appIication to complex behavior, tended to think of _rganization as something extra, something�ties in tbe fact that it is applicable to any arbitrariIy de_ned set of _d_ed to the elementary variabIes, the modern theory, based on _�states. Its appIication Tequires neither linearity, nor continuity, the 1ogic of communication, regards organizat ion as a restr ict ion _ V_���. ;nfe d O;o d__�;on;,! !à nY, _. _;e, b, c t ,. mo ;s,S. � 'a dm, n;e d, pxnaac ltY__p d;, ;;_ e d_ ;oP,P p; c h d 'c ;n h ebe; p p_o pq, _. ;; e :, O;_. t s fc ;;; he x; b_u; gw ,. ;h_ l;,. � _ep! p Y o � ;. t g,. h _� ' Vg?�to Fisher's method of tbe analysis of variance ; yet it requires no in existence we must cleaTly go carefulIy, especiaIIy w hen _e ,�_etric in tbe variabIes, onIy the frequencies with which the various discuss with others, lest we should f411 into complete confusion. _ _,�_ combinations of states occur. It seems to me that, Just as _isher's This excursion may seem somewhat compIex but it is, I am sure, _�conception of the analysis of variance threw a _ood of 1ight on to advisabIe, for we have to Tecognize tbat the discuss ion o f oTgan i-�the compIex relations that may exist between variations on a zation theory has a pecu1iarity not found in the more objective��_ _ay give us an a1together better understanding of how to treat the pToduct sp4ce th4t I have just re ferTe d to. W hence comes t h is _�8 comp1e_ties of re1ation when the variab1es are non-metric. In p_oduct space? Its chief peculiaTity is that __r co_r__'_s more rhan��_ess; doubtless they wiIl aIso occur commonly in the brain-Iike gives us the actual, constrained sub_er. ,,�=�

Page 3: selforganizing - University of California, San Diego

________ _ _aTbobtleosdb_ N_sbeerv_Nxel_pyaxrtsfanxdhtt_ht ealnrdlntthe_ru4scts,toanrts_tbby4tttrreeaat_tl_ns_ gtbethseystwehmol4es aans ___l____;_ (((tpgo p_ y t y_ y__)gf _g _ y p_ N _N _. _ N p t N Nf h _f t N 1_!ltt1(__1lllt________���\ ' J_02 5 8��� W. ROSS ASHBY p,Tyc,p_,s oF Ty, s,L,_ogGAy,z_NG sysTEm 25g��i t e PrO UC t SP4Ce _ such as activity doTmancy swarming etc may see Mo organiz_-�represents the uncertainty of the observer. The product space may t,.o, only t,a.ec?to,_.es of tbése (unanaíysedj states !, _�ere Ore Change If the ObSe,Vef Ch4n_eS; and tWO ObSerVerS may t e xam le of the ;nde endence of cco, ani,ationt, and , _�IegitimateIY use di_erent product sPaces within which to record ,cd namI_c,tt ,_s ,_ven by tbe f,ct th,t wbetbef o, not a ,e,_ system _ U�. ,,. a. Ua eVe, me 4C U4. lng' e C:n- is organized or reducib1e depends partly on the po1nt Of VleW ta_en ,�Stralnt lS t US a re __lO, etWeen O SerVer 4n t Ing i t e prOpertIeS the obse,ve, It is we_1 known fo, instance that an or anized _�__ of any particular constraint will depend on both the reaI thing and , ,. e ,_nte,act,,ngj _,_ne,, system oF , pa,ts su_b ,s a netwo,k of ; __�On t_e ObSer Ver. _t fOIIOWS th4t 4 SUbStant'al Part Of the tbeO,Y Of pendulums and springs can be seen rrom another point of view _! _�organlzatlon will be concerned with__operties t_at a_e not intrinsic th,t of the so ca_led _cn,,m41tt coo,din,tes) in wh;ch a__ tbe�to t_e t_in_ 6_t _re _elational between o6serve_ an_ t_ing. We shaIl ,ew_ ,_dent,_ned p,,t, a,e completely sep,,,te so th4t the whole�_ see some strikin_ examPIes of this fact 1ater. ! , ,.s ,educ,Nble The,e ,_s the,efo,e notb,_ng pe,ve ;se 4bout my ins;s_ ,�tence on the relat1v1ty of organ1ZatlOn, fOr adVant4ge O t e aCt IS _�wHo_E AND pARTs ,oUtInely taken ln the StUdy Of QUlte OTdInaTy dyn4mlC SYStemS. ' _,�U ,, _ f d.. _. . . _inalIy, in order to emphasize how dependent is the organiza-�__O_ ItlOna ltY'' IS an eSSent14I COmpOnent 1n the COllCept Of ,_ n seen in a s stem on tbe obse,ve, who sees it I wi11 state the�U, Organ1Z4tlOn, So aIso ls the assumPtion that we are speaking of a ,, os,_t,_on th,t. ,_ven, whole w;th a,b,_t,,,;_y given beh4v;o, a�WbOle COmPOSed Of PartS_ ThIS aSSUmption is Wo,th a moment'S ,eat va,,_et of a,bit,a, cc 4,ts,, c4n be seen in it. for 4_1 tbat is�scrutinY, for research is d.ev.eloPin_.a theorY of dynamics that does ne,ess4,y when tbe a,b,Nt,a,y p,,t ;s p,oposed, ;s _h4t we ,ssume _, _� the given part to be coUpled to another Sultably re1ated p4Tt, SO j�, Unan4IYSed whOIe (AShbY, 1958, a). In phYslCS, of coUrSe, We that the two to ethe, fo,m a who_e ;somo, _ic with the whole�, usualIY start the description of a system by saying ''Let tbe vari- th,t was g,Nven. _o, ,_nstance, s,ppose tbe g,_ve, whole, _ of !,�' 2''", _ 10 states behaves in accordance wlth the transformatlOn: t _�_ made of n f4nctionaI parts. Tbe other method, however, deaIs ' _�_th4_JanaI d stts s s fth hl _th t __ _t _g1st___x_�_ W, Se aet _, 2t'''O e W OetWI OU eXP!C1 W _�mentlOn Of any partS tbat may be contributing to these states. g _ s g s t t x _ _ _�e YnamICS O SUC 4 SyStem Can then be de_ned and handIed lts kinem4tic graph is�mathematicaIIy ; I have shown elsewhere (Asbby, 1960, a) bow sucb _�4n apprOaCh C4n be USefUl' What I WlSh tO pOint OUt he,e iS th4t t_s__g__ l�we can have a sophisticated d___mics, of a whole as comp1ex and ? ? t _,�cross-connected as you please, that makes no ,eference to any parts _ _ , __�and that therefore does _ot use the concept oF organi2ation. Thus __x___ !, _,�tbe concePts of dYnamics and of organization are essentially and suppose we w;sh to _csee__ it as cont,;n;ng the paTt p, with __ _�indePendent, in that aII four combinations, of their presence and inteTn4_ states _ 4nd _np4_ states _ '. _?��This fact exempIi_es what I said, that ''organization'' is partly _ _ _�in the eye of the behoIder. Two observers studying the same real 1 2 !, _�i4l system a hive of bees s4y may nnd that one of them __ P ! _ Y�t ' } 1 21 j�thinking of the hive as an interaction of nfty thousand bee-parts, _�

Page 4: selforganizing - University of California, San Diego

u tms_moyfcoho,_eaar Tstlmfle_çthloroteavrb_mm_tebl_rooeafnTtaul_cnf_ccflw_re._sohrytmoas Itl_2eenl___Mtm_t_ttw_Asthtr1chel_antHthtl TweNarh2nEll__ttc31ts_hcwl__ptadNhas_eteGdtss_eEoa2tlNlhel__neE1dgIReqtmw2A_u_ el__LNt2snhtldl_op2hln_Jhs_.2_ttshk_wNcadhtlat_thsyeds_towfeomwhrsket ______1l__;tv_ ngss(ls_Ayto4sc'tstt'htheessmbotycsafthanpllfNradn9_naesla_dl_stgttml_Nhs3c4ar)_(tmephrlocp_rualNtLnwlth_hg4tttethl_rk_nleNcacgttsnhaet(nvyhblml_te9,teer42hotcvsbatfahe)veutr_Neehtnnosaee)(nm__seupndtssarseatoelv_donndmoncuanteacntwcttd(Bhoh_sfINeeleNencxr_ttthetphla___dlelelolx_4ydkl_ntfst_e_cse_m_c_wdlt9_yonhasavntcg9y(ol_thgl)___9nNosnln_susrma.eool_tmtHttuoy_aesde_nllra4lyyleNdlys.lt Illl_ll1l1_______________���_�_ 2,o , gL ,ANTz,,G sysTg, ,,, _?u,������_ F ç beJh_ves in a Iaw-abiding and mach ine- I i ke way. A Iso to be ex-�_ ' cluded _s irrelevant is any reference to energy, for any ca Icu lat ing�n _ 1 t 1 1 1 2 1t 3 2t 1 21 2 2t 3 _acbine shows th4t wh4t matters is the regularity of the behavior-�_ t 1 1 ' ' 1 W t t 1�_ _ v ''� ' ' ' ' ' The f,nd4me,t41 concept o f c cm,c h _ne t t proves to h4ve a form ;�by putting A _ F and g = _, then the new whole __ h4s t,ans_ _s that was formuIate d at 1east a century ago, bu t t h is concep t b a s _���' _,. 17 17 1 1 7 1 7 2 17 1 t 3 1, 2, 1, etc. tb,t m,.t as ,Nnte,,,l ,ta_ We and the s,4te of ,N,s su,,ound,_n �t de6,e ��2t 2t 1 21 1 t 2 2t 1t 2 1, 2, 1, etc. !_ uniqueIy the next sta�e it wi1l go to. ;�which is isomorp_iç _4_'r_ Mr under the one_one correspondence q :0 This _ennition, formal1y pTopose d _ fteen years ago ( As b by,�_ l945) has _7ithstood tbe passage o f t ime an d is now becom i__ ,�1 1 t 1 1 t 1 t 2 l t 1 t 3 1 t 2' 1, etc. gene,,lly accepted (e.g. _e_rey, 1gsg). _t 4ppears in many for_s.�_ _ p _ , etc. When the variab1es are continuous it corresponds to t he descr ip-�_ Thus subJect onIy to ce,ta;n ,eq,;,ements (e g th,t equ,_ll_b,_Na map t ion of a dynam ic system by g iv i_g a se t o f or d inar y d i _ e r e _ t i a l�'_ _nto _quij;b,ia) 4,_ d__,m,_, s_ste,, c,n be m,de ro d,.s_ray, equations with t ime _s t he in depen den t var ia b 1e. T he __n dame_ t _ l��_, v___e6wrpó;__yt._"'''M'd ''P~''S''' ''mP'' b' " Ch"n" '" 'h' _ _ _ _ _ni _nt ' ò _ej h _ s P __ _ _e _e � o gniz _d b � m _ _y _a _1ie iwo _ : _T Ys_��ar1ie _ _. O " ' " ' O ' ! _�Such a representation _y di_erential equat ions is, however, i _�I have Just used a way of tNepresenting two ccpaTts't, t'coupledt, too restricte_ for the needs of a sc ience t hat inc lu des b io Iog ica l ' _��mean by a ''machine'' in gener41? ubiqujto4s. So arises the modern de_nition, abIe to inc lu de bot b�Here we are obviousIy encroaching on what has been c41led the continuous and the discont inuous an d even t he d iscre te,�generaI system theory'', but thjs Iast discipline _lways seemed to without the sIightest loss of r igor. T he ' 'mac h ine w i t h inpu t ' ' _��and therefore tied to whatever the real wo Tld provides, oT with denned by a set S of intern_I states, a set I o f input or SurrOUn dlng !,���substant;a i eadvanc _s_ _ _h e _ _s _ de � Ca 4_ nt _ 'at _ é ha tv e n3� _a _ i _'de �t_ _ _no mw Yn i Py _ _s O õ F m ach _' n e aT e to b Ye found here ; and all intere _t;n_ t _ _ _�_ed the esse_tials of the ''machine jn general''. deviations from the co T_cept are to be foun d by t he corres_on d ing _' __',�Before the essentiaIs couId be seen, we had to reaIize that two deviation from the de_nition. _ _u_�factors must be excl__ed as 2r_eleya_r. The _rst is ccmateria1;tytt- We are now in a pos it ion to say w it hou t am b igu i ty or e v a s i o n ! _ '�

Page 5: selforganizing - University of California, San Diego

__ ________t__ _ t _ t t , 1l______?__�����?,�W_ ROSS ASHBY ? pR_Nc_pLEs oF THg sE_F_oRGAN_z_NG sysTgM 263 ?=�_, _�Conditions r. If S is a product set, so that S = IT___ say, then the wo,d suggests some eva1uation whose orjgin bas not yet been�parts i are each specined by its set of states T_. T_e "organi5ation'' conside,ed�between these parts is then spec__ed 6y t_e mappi_g _. Change _ L _n some cases the distinction betwee_ tbe 'cgood" oTg4_i2atio_�and the organization changes. In other words, the possibIe organi- and tbe _cbad_, is obvious, in the sense that as everyone i_ these�zations between the parts can be set into one-one correspondence cases wo,1d tend to ,se tbe same criterion, it wou1d not need�with the set of possible mappings of I x S into S. Thus ' 'organiza- ,;; exp1icit mentio_ Tbe brain of a 1iving organism, for instance, is ;�tion'' and ''mapping'' are two ways of Iooking at the same thing- í? usu,_1y Judged as baving a 'cgood,' organi24tion if the organi2atio_ _�the organization being noticed by the observer of the actual (whetbe, ;nbo,n o, 1ea,ned) acts so as to FurtheT the organism's '�system, and the mapping being recorded by the person wbo re- su,viva1 This conside,ation ,eadi1y genera1_zes to 41_ tbose cases�presents tbe behavior in mathematicaI or other symboIism. ., _, which the o,g,nizat_on (whetber of a cat or an autom4tic ,,�,,!; piIot or an oil re6_ery) is judged ''good'' if and onIy if it acts so as�,,GooD,, oRGAN_zAT_ o N !,_ to _eep an assigned set of variab1es, the ''essent ia 1' ' var ia b les, '� wit_n assigned Iimits. Here are all the mechanisms for homeo- '�At this point some of you, especiaIIy tbe bio1ogists, may be stasis both ;n tbe o,iginal sense of cannon and _n the ggnerali2ed�feeling uneasy; for t_s dennition of organization makes no sens ë _,om this c,_€e,ion comes tbe ,e1ated one that 4n organiza-�reference to any _se__lness of the organization. It demands on1y , t;on _s c_good,, if _t makes the system stab1e around an assigned�tbat there be conditiona1ity between the parts and reguIarity in ! equi_ib,;um somme,ho_ (1g5o) i, particu1ar bas given a wealtb 8_�bebavior. In this I beIieve the dennition to be right, for the questio_ of examp_es d,awn f,om 4 g,eat ,ange of biologica1 and mecb-�whether a given organization is ''good'' or ''bad'' is q4ite inde- an;ca_ pbenómena sbowing how in al1 c4ses t_e idea of a ccgood _,�pendent of the prior test of whether it is or is not an organization. o,ga_ization,, bas as its essence the ide4 of a number of parts so Yn,�I feel inclined to stress this poi1lt, for heTe the engineers and inte,acting as to 4chieve some given __foca1 condition''. _ wouId _,�the biologists are like1y to t_nk a1ong wide1y di_ering Iines. The ' _ike to say he,e tbat _ do not conside, that sommerho_,s contrj_ _,,�engineer, having put together some eIectronic hardware and bution to ou, subJect has yet been adequately recogni2ed. His�baving found tbe assembled network to be roaring with parasitic identi_cation of eyact1_ wbat is mea_t by coordination and '�osciI1ations, is quite accustomed to the idea of a ''bad'' organiza- integ,ation is in my opinion, on a par with caucby_s identinc4tion _�tion; and he knows that the ''good'' organization has to be ! of exactly what was meant by convergence. caucby's djscovery _�searched for. The bioIogist, however, st4dies mostly animaI was a ,e,_ discove,y, and was an enormous be_p to 1ater workers n�species that have survived the long process of naturaI seIection ; by p,oviding them witb a concept, ,igorous1y de6ned, tbat could 9'�iy So almost aIl the organizations he sees have aI__eady been selected be used again and again, in 4 vast Tange of contexts, and a1ways ' _,u�_ to be good ones, and he is apt to think of ''organizations'' as with exact_y the same meaning. somme,bo_,s discovery of how to�_ necessaril_ good. This point of view may often be true in the ,ep,esent ex,ctry what ;s meant by coo,dination 4nd integr4tion ' _�biologica1 world b4t it is most emphatically not true in the worId and good o,g4ni,4tion wil_ _ am su,e eventual_y play a simi1ar1y ' '�in which we peopIe here are working. We M__st accept that f,ndamental pa,t _n ou, wó,k ,( 1)� most organ izat ions are bad ones; His wo,k i__ust,4tes, and em pb4si2es , wbat _ want to s3 y here_ ; _�(2) the gOOd OneS h4Ve tO be SOUght for; and , f_ere is _o s_çh t___g as "good orga_izatio_'' i_ __y a6sol_te se_se. _�(3) Wh4t iS meant by ''gOOd'' muSt be cIearly de_ned, explicitly t A1ways it is Te1ative., and an org4niz4tion that is good in one _�if _ecessary, i_ ever_ case. context o, unde, one ,,ite,ion may be bad unde, ,nothe,. ' _'�What then is meant by ''good'', in our context of brain-1ike sometimes tbis statement is so obvious as to arouse no oppo-�mecbanisms and computers ? We must proceed cautiously, for the sit_o_. _f we bave ba1f _ dozen 1enses, for instance, tbat can be __��=: __�?�l=�l�, I�

Page 6: selforganizing - University of California, San Diego

______u__?__????___ t_mlhcptwmtwm__tsgchnetbgl0_salgeeb_ltseeaoautDsonaatnn1o_ytao_aorlt_ftetotltsdslcf_dlN_hty_taoyfolkstcotyutstagwmut_nh.olnndg_gwolnoeeeudtlolg_heo_ndrnnwatpoe0gclhnlsns_tyhetdn4adtlt_eohthsntl_nnNteslwrevc_datlnldntlm_trot2anwhshoo6t_anrtbgar_N__otrlnmht_csdloltermlvrNhsdhe_ot__sn_lllttyadelans__lttowtnyohll_bhtwrsse_u _nehegolw_oNwhrbns_a_afpe1e(baetselhtnnTnetltltoyNrlyllh'_lv_hr_ryoahcltzttae4o_fenslhNaerctnsbtrtfnt_eteoeclu_o1befdnlotysecrhoasrprmnshs_nclumaesethouautnlehr_nollnsoaess1antaelsftenfnnetttneydry(ertdt_kte_ednolh_sdmlteocmgtucoyeleyelutoawfdrrawxdsnaspmyll_eNleNt_nsas7spttderttc4ohbnsterresTeo1kbeyrawdtlrrhelrc_oopoeehhslbNtetpplr Adatl'aNtonbyatbcreel Tdohelhh_ret_msy__tsetdp_!_hllggNNstfsblec_aa)teoohn Uteru Tmrflrg1oe_n Bpaelotrtadvl_oeenen Aparhol_ltor_sdrennrflesnrobts)eftdenhhtdaeanupsehaaocnmtaspmc_rtuauvl_crfheeaombedlel1poNrattrdnnNdateynla_osnettadrnbhnolont__atllll_ynNNloetfe_ttt_ss_ ______l__)_l! st'tmlTNlsmogbuobtnbssnool_rpr_blcl_l_eoAwegmoocebrnournnglesceslt__eNttcseotcsoplshtdrowN14hgstsaotoeltyehlmleNlh_hfdlorpts_teszfeetplnn_aeduawowsrl_ucrotdapofne_ntlool_tlnteeetecsowtteh Tbnggcns__thrndtaeldl_r4oa_yetneoearseteynuamrs)slhoneeeesIlrredbNttat2axeawnyrpelaowllsat__a_odnt_fdethdfomele_Tct Tl_nela_ealonoprwapporth_nnl Tt_lllao_troceuytessgTntatghbpstmheaetwtahswaltwtnn_occ_ylln__oeaamvlotol__msnnzl'y_cwlunlo6eruarsoelsf4soollnwtNmdncteoluolt_pllhdoh7anll_Nnclmodleb Tnwaelbofprsn_yne7coewls_grctttlrne_ola_eoschlaG_dasttnrasdaenotshnegtycoolhuh_ada(beonlcNbotwotnssru YdeontsntalhfuhmnazhpneelaotetnasenolTNse_noustttrottemlhtwdoleadobxyln_Teelln_toeon'nahvosog Trbfbllt_t_tetnlftsoykterellgloltendyrsutlothhtms_netNrg_hrepl_lleh__tlgtangblwemals_hofdvghoetas__l_awbJreesslleNlnTdulgny_lntsd_rweal4as_cabyteevlrehyttle_l_red?eeennhIl_e_aellrnddwsothenanltbta4l_4pru_lsvseltyl_tscls__tsnsvllotsedol_lr_oalbmoentteanoylmaolal_llbNssnlnnenuasetyooeaed___yl__ _____1__________u__� U� 264 w. Ross AsHBy , _ R T N c I _ L E s o _ T H E s E L F- o R G A N _ z _ N G s y s T E M 2 6 5� = assembIed this way to make a teIc_scope or that way to make a memory? Yot at a I l, I rep ly-on Iy w hen t he e n v i r o n m e n t i s o f a� _ microscope, the goodness of an ass_mbIy obviousIy depends on type in w h ic h t he future o f ten cop ies t he p a s t; s h o u l d t h e f u t u r e�����_, once: aw ^ha Y iev _r mth e rf _cu1 _y Ór _T g _ _i ^z _ __' _n a c __ ev _s; 1 _ _th t a _ S be t " _ __ _' ' é _ _i _o n nmen _, m _moT y Yi s Sp o _ it iv _ ly d S i � _ dvan tageous. _To-�__ _ot jn tbe ''focal conditjons''. longe d co^t4ct w i t h t h is env iro_men t w i I l l e a d, o t h e r t h i n g s b e i _ g�_ M . We know, of :ou. rse, Iots of.exa.mples where the thesis is true equaI,toevo Iut ion int he d irec t _lono f d im i.n i s h. e d 1 _ e m o: Y- c a P a c i t Y.�__ an antelope has Iost its Iife by stopping to see what the h4nter's e, o f connecte dness. Is i t no t goo d t ha t a b r a i n s b o u l d b a v e i t s p a r t s�'? of the type that does, or does not, 1ead to tempoTary immobjIity when the env ir_nment is itse 1 f r ic h ly c ónnec te d. w b e n t h e e _ v i r o n-�_rom a di_erent angIe we can notice Pribram's resu1ts (1957), ' highIy re duc i b Ie, i.e. i f i ts connec t iv i ty is s m a l I ( A s h b y, l 9 6 O, d ).���'_ ' ) _ N _ ' ' ' ' '�_ assert that there is not a singIe menta1 faculty ascribed to Man As anot her examp Ie, w ba t o f t h e ' ' o r g a n i z a t i o n ' ' t h a t t h e�_ ood But cb, e th _. o U U a t O r mt a C S I_ ad. a_ C U Yt U S U ad V O O. _: e St: e S ' Sd n O !. SbI O l ' b c4 se t he e á,t h has an a tmo������U�_�_�_�

Page 7: selforganizing - University of California, San Diego

_ _?___?__J____l__ on_Ttedbbchomt estl blo_fdeha1 lth d l ttb_ d_bthbtd th b l _l t ___ tfh Nddb _kt thydt_dthtthh g hnthl2latglo_sn d_evge_gsohp. teo d_ l _ l b_th__ _ thdt do. n_ rettrh_oN__rtf 1____e___�_�l�_, ___ _u��266 w. Ross AsHBy PRINCIP__S oF THE S_LF-ORGANIZING SYST_M 267��After t_ese 4ctu4l jnstances, we can return to theor y. _t js bere more and more a�ected by one anOther aS _IamentS aPPrOaCb�_ tbat SommerhoR's formulation gives such helpful cl4ri6cation. He the other eIectrodes. In genera1 such SyStemS Can be mOTe SImPIY�shows that in aII c4ses tbere must be given, and speci_ed, nrst a characterized as ''se1f-connecting'', fOr the Ch4nge frOm lndePen-�__ set o_dist_r6ances (va1ues ofbis "coenetic variable'') and secondIy dence between the PartS tO COndltlOna!'tY Can alWaYS be Seen 4'�a goaI (his ''foca1 condition'') t, the disturbances th Teaten to drjve some form of ''connection'', eVen lf 1t lS aS PUreIY fUnCtlOn_I aS�tbe outcome outside the foc4I condition. The ''good" organi2ation th4t frOm a radlO tranSmltter tO 4 reCelVer'�_ is then of the nature of a _el_t2'on between the set of disturbances _ Here, then, iS 4 perfectly StraightfOrWard fOrm Of Self-OTganIZln_�_ but _ m,sc empbas_Nze that the,e c,n be ,o ,ss4mpt;on at '�U an t e gOa _ Chan_e the Set Of d1StUTb4nCeS, and the organizat1on SYStem i�_ _ _ . . ', ,. ' _ N _N ldwiI�Wl OU l Se C anglng, IS eVa uate a ' 1nstead of ''good''. IS POln a e Or a e a gOO '�AS I said, there is no p__operty of an organization that is good in We WiSh it to be a ''good'' one, We mUSt _rSt prOVlde 4 CrlteTlOn�any absoIute sense; aII 4re reIative to some given environment, for distinguishing between the b4d and the gOOd, and then We�or to Some given set of threats and disturbances, or to some _ mUSt enSUre that tbe apprOprIate Se1eCtlOn IS made_�given set of prob1ems. We are here approaching the second meaning of "SeIf-OTg4nlZ-�n ing'' (Ashby, l947). ''Organizing'' may have the _rst meanlng_�;, Just discussed, of ''changing from unorganized to organized''_�SEl_-ORGANIZING SYSTEMS But it may also mean c'changing from a bad oTganization to a�M . ,, _ _ _ _�_ Ope aVe nOt wear1e you y eIabori_g this relativity too gOO One , 4n lS lS e C_ e W SCUSS nOW, a? mOre 8,�__ much, but it is fundamental and js onIy too readily forgotten when fuIIy. ThiS iS the case of peculiar intereSt tO uS' fOr thIS IS the '�Y t . ' . . . . . . c,se of the s stem tb,t ch,n es itseIf f,om a b4d w,y ,f behaving _,�Un W Organl2a lOnS a are el er lO OgICa l_ . _�'?9 origin or are in imitation of sucb systems. with this in mind, we to a good. A WeI1 known exampIe iS the Ch1ld that StartS wlt a '�_ can now start to consider tbe so_c41led ccself_organi2ing,, system brain org4nization that m4kes it nre-Seeking; the_ a Ch4_ge� we must proceed with some c4,tion he,e ;f w, a,e not to 1,n_ ;, occurs, and a new brajn organization appears that makes the�confusion, for the adjective is, if used loose1y, ambiguo,s, and, if c_1d nre-avoiding. Another eXamPIe WOU1d OCCUr lf an aUtOmatlC�used preciseIy seIf-contradictory pi1ot and a pIane were so coupled, bY miSta_e, that POSltIVe�_? ' . ' a ,avati�_?, O Say a SyStem IS ''SeIf-ofganlzIng'' Ieaves open two quite ee 4C ma e e W O e errOr- ng ra er a_ e -�___, di_erent meanings. correcti_g. Here the organization is bad. The system WOU1d be�'_ There is a _rst meaning that is simp1e and u,ob;ect;on,b1e ''seIf-organizing'' if a change were uutomaticalJ_ made to tbe ?'�Tbis refers to t_e system tbat starts wjth its pa,ts sepa,ate (so feedback, changing it from positive to negative; then the WhOle _�nu tbat the behavior of each is independent of the otbers, states) a,d would have changed from a bad organiza_ion tO 4 gOOd_ CIeaTIY, '�_u whose parts then act so th4t they ch4nge towards forming connec_ t_is type of ''seIf-oTganization'' iS of PeCUliar intereSt tO US. What�_ tions of some type. Such a system is ''se1f-organjzing'' in the sense iS 1mPlIed by It?�_ tbat it changes from ''parts separated'' to "parts joined,,. An Before the q4estion is answered We mUSt nOtICe, If We are nOt ,!�example is tbe embryo nervous system, which starts wjth ce11s to be in perpetuaI danger of confusion, that no maC_ine Ca_ be�having 1itt1e or no e_ect on one another, and changes, by the seJ_-organizing in this sense. The reasoning iS SimPle. Denne the�n growth of dendrites 4nd formation of synapses, to one in w_ich set S of states so as to specify whiCb m4Chine We are talklng abOUt_ _ n,�each part's behavior is very mucb a_ected by the otbe, pa,ts. The ''organi2ation'' must then, aS I said above, be identined Wlth __ ,�Another exampIe js _ask,s system of e1ect,o1yt;c cente,s, ;n whicb _, tbe mapping of S into S that tbe b4SiC driVe Of the maChlne '�the gTowth of a _Iament from one electrode js at nrst li€tle a_ected (whatever force it maY be) imPOSeS_ NOW the lOgiCa1 re1atlOn here _,�_, by growths 4t the otber e1ectrodes ,. then the g,owtbs become is that _ determines the changes of _:-_ iS _e_ned aS the Set Of�

Page 8: selforganizing - University of California, San Diego

__ ThN_ls suggest_llon l__s l_1_Nogl.cal_t__f_or__we nnownhave th_at F __ MN l_2/d_(_).t r____x_v_ methaneN , c_rbon_tet etc_t to_sg_e_t to___ge_ther untl_I th.ey roTmed pro_ _____)___l___lt___n?______???__?___?____�������268 W. ßOSS ASI_BY , ;__, pR_Nc_pLEs oF THE sELF_oRGAN_z_NG sysTEm 2 6 9��COUPIeS (s_, sJ) S4Ch that the internal drive of the system wilI cbange, 4s we have Just seen, cannot be ascribed to a_y cause in�fOrCe State s_ tO Change to sJ. To aIlow _ to be a function of the ' the set s., so it must have come from some outside agent, acting on�State IS tO make nOnSenSe Of the whOle Concept. tbe system S as input. If the system is to be i_ some sense ''seJ_-�SlnCe tbe argument is fundamentaI in the theory of self- organjzingt', the ccseIf'' must be enIarged to include this variab1e _,�Or_anlZlng SYStemS, I maY help exp1anatlon by a parallel exampIe. and, to keep the who1e bounded, tbe cause of _'s change must be 8�NeWtOn's Iaw of gravitation says that _ = M1M2/_2, in particular, _ in s (or _).�that the fOrCe VarieS inVerSely as the distance to power 2. To _ Thus the appear_nce of being ''seIf-organizing'' can be given�POWer 3 WOUld be a di_erent taw, But Suppose it were suggested only by the machine S being co4pled to another machine (of one _,n�that, nOt the fOrCe F bUt the laW Chan_ed WIth the d_StanCe, so part): '_�that the power was not 2 but some function of the distance, _(d). _ _ =U��, and t_s represents not a law that varies with the distance but o_e _ _,_ U�Iaw covering all distances; that is, were tbis the case we would _�re-de__e the law. An41ogously, were_ in tne m3chine to be some Then the part S can be ''seIf-organ iz ing ' ' w it h in t he w hO le S +K_ ,�function of the state s, we would h_ve to ,e_denne ou, machine OnIy in this p_Ttial and strictly qua I i _e d sense can we un der- ,�Let me be quite explicit with an examp1e. suppose s hJd th,ee , _ stand that a system is ''seJ_-organ iz ing ' ' w it hou t be ing se l f- !,�states: _, b, c. If_ depen_ed on S there would be three_,s: _,, _b, ContradlCtOry. i _�_ _, say. Then if tbey are ' Sjnce no system can correctIy be said to be self-organ iz ing, an d�since use of the p_ase ''seIf-organizing'' tends to per_etUate a�_ _ _ C fundamentally confused and inconsistent way of 1oo k ing at t he ,�a b n 6 su_Ject, the phrase is probabIy better aIlowed to dle out. ! u,_�b c a n _�_, l 6 6 a THE ,poy,AyEou, G,N,RAT,oN o_ oR,AN_,AT_oN 9,,�then the tr_nSfOTm Of a muSt be under _a, and is therefore b, so , wben I say that no system c4n pToperly be said to be seIf-�the WhOIe Set Of_'S wOuld amount to the sit7gle transform_tion: organizingt the ljstener may not be satis_ed. What, he may ask, __�_ _ , of those changes that occurred a bilIion years ago, that Ied Iots oF _�_ 6 c_rbon atoms, scatteTed in littIe molecules of carbon dioxide, , __��It IS CleaTly lIlOgICaI tO ta1_ Of_ aS bein_ a functioll of S, for S_Ch teins, 4nd then we_t on to foTm those Iarge active IumpS that�_ talk WOUld refeT tO OPeTatIOnS, S4Ch aS _w(b), which c_nnot in today we call ccanima1s'' 7. Was not this process, on an isolated _ ,�raCt OCCUT. p1anet, one of ''self-oTganization'' ? And if it occurred on a l _�Ift then, nO maChIne Can prOperly be Said to _e self-organizing, p_anet4ry surface can it not be m_nde to occuT in a computeT? I�OW dO We regard, SaY, the HOmeOStat, that ,e4rranges its own am, of course, now discussing the origin of life. _as modern __�Wlrlng; Or the COmPUtet th4t WrlteS OUt ItS oWn progTam ? system theory anything to s4y on tl_is topic7. _�The neW IOglC O? meCh4nlSm enab1eS us to treat the question _t h4s a gre4t de41 to say, and some of it Aatly contradictory to ' _�rlgOrOUSIY. We Start WIth the Set S of States, and assume that _ wbat has been said ever since the idea of evoIution was __st _ __,�ChangeS, tO _ S4Y. SO We rea11y haVe a va_ia6le, _(t) say, a function considered. In tbe past, when a writer discussed the topic, he '_�

Page 9: selforganizing - University of California, San Diego

____u_____ esohyvsetTemfpdyyhy yJ(q. gyJ M_g_ steaqgfg_pypgtpghtepydp(tglcJ)tyqepugovugmyepglrgpybtgpplhnNremlaxugtalmptcya_pngpgmnNgeelfoatrhcwo4ehgmgshrp,mne)tgyaethhgowh_yehsYcayutuhgatdsnyhoosthnarawtvotgdaeeeeet lll__ll_ _N _ lN _ d d t(h l f l t._hf ft_hf l b t _ _ f f 1_d.__ _th. t N sNlel_v__e__t_Nn t_____tl1_tt____l_?___J__���_"_ ' _,���27o , w. Ross AsHBy , PRINcIPLEs oF THE SRLF-ORGANIZIYG SYST_M 27l ,��, ,nd he tben t,;ed to d;s 1a some wa that w,u_d enabl t ,_s mUCh mOre intereStin_. Here We haVe the aUtOmatlC PllOt SUCCeSS- t�_ ,a,e and peculiar event to occu,. so he tTied to djsplay that the,e fu!!Y COmba'in_ an eddY; 'he Phe"bn :ediS'r,ibU"^g h!x_!OOd, AOW i�' js some route from sa ca,bon dioxide t the mi ' t after a SeVe,e h4emOrrbage 7 t e USlneSS rm. reStOC ln_ a te: a�tben,e to the ,,te;n and so th,ou h natu,a_ se_ect,_on _ S4dden inCreaSe In COn?UmPtlOn; t e eCOnOmIC SYStem reStOrlng�_, t;on to ;ntell; ent be;_n s _ ;, that th,_s look,_n fo, s e,,_al cond,_ a diStribUtion of S4PPIies afteT 4 Sudden deStrUCtlOn Of _ fOOd ,�_ t;on� ;s U,_te w,on The t,4th ,Ns th, o os,_te_,v,, d nam,., crop; a_d it iS a m4n succesSfuIIY getting at leaSt One meaI a daY =�sy,tem gene,,tes ;ts own fo,m of ;nte11; ent l;fe ;s self_o, an,_,,_n _ dUring a Iifetime Of hardShIP and UnemPlOYment. n,u�_ ;n th;, sense _ w;_l demo,st,,te the f,ct ,_n a moment w What makeS the Cha^ge, f?Om triVlal tO 1ntere StI,6, IS SlmPIY j =_�__ d _ N h. f _ h . l _ ' ) the scale of the events ''Going to equilibrium'' is trivia1 in tbe __�u, aVe al e O reCOgnl2e IS aC IS 4t Untl reCent y We aVe a ( . .. . . _�_ no ex erience of s stems of m e di u m c m ' . ' Slmp e pen U Um_ Or e eqUl ' TIUm lS _O mOre an a ng e ''_ ,U�been l;ke the watch and the endulum and we _,ve found tbe,_, , point. BUt When the SYStem IS mOre COmPleX; Whe^, SaY, 4 COU,trY S _,�U p,ope,t;es few and t,;v;,l o, the y h,v� been l,_k e t h e d o a n, t h e 5_ economy goes back from wa,time to normaI methOdS then the _'�h4m4n be,_ng and we h ;ve fo,,d the,N, ,o e,t,Nes so ,,.ch a,_ stabIe region is vast, and much interestin_ activitY Can OCCUr t, _�,em,,kable tb,t we h4ve th,u bt them su e,,,tu,,_ onl ,Nn the WitMn it. Tbe COmPUter iS heaVen-Sent in thIS COnteXt, fOr lt '�last few ye,,s h,s the ene,a__ u, ose c o m u t e, ,N v e n_, s a enables us to b_idge the enormous concePtuaI gaP fTOm the SlmPIe _�sy,tem ,;ch enou h to be __nte,est,N, et st,_ll s,_m le en,u b ,_ and Unde,StandabIe to the COmPIeX and 1ntereStIng. TlJUS We Can ;�unde,st,nd,ble w;th th;s m,ch;ne 4s tuto, w e c, n, o w b e ,N n t o r gain a considerable inSight intO the SO-CaIled SPOnt4neOUS genera- i _�th;nk about s stems th,t a,e s,_m _e e n o u h t o b e c o m, e e n s ,. b _ e _, tion of Iife by just seeing how a somewhat SlmPIer VerSlO_ Wlll _, ,�in detail et a1su ,jch enou h to be su estive. with t i, i aPPear ln a COmPUter. , '�C4n See the trUth of the statement that ever_ isolated detefmiM_re ' ;�dY_amic sYstem o6e_iMg _nc'_anging laws _i/l develop ''orga_isms'' , "�that are ,dapted to f_eir ''environme_ts''. ' coMpET_T_oN _,�The argument is simple enough in principle. we start with the . . _ . T h t. t. _ _,�f4ct that s stems in ene,a1 o to e j i ' He,e lS an eXamP e O a SlmP er VerS'O"' e COmP' ' lOn ,�_s states a,enon __.b N l .f 1 d' h _ betweenspeciesisoftentreatedasifitwereessentiaIIybioto_ical; !�-eqUl l rla l We eXC U e t e eXtTeme CaSe�of the s ste__ jn neut,al e u;lib,ium so ,_ o,_ lt lS In faCt aß eXP,eSSlOß O a PrOCeSS O a, gTeater genera 1tY_ _,�to one ofthee u,__,Nb,,_a the, stem ,Ns' o,_, f,om al,, e,n mbe, Supposewe have a computer, forinstance,WhOSe StOreS aTe nI!ed __,'_�f st,te, to a sm__e, _n th,Ns w, ,,t ,.s e, f o, m ,., a s e l e, t ,., n ,. n _ at random with the digits O to 9. SupPose its dYnamic Iaw iS that�l the pu,e_ y ob_e,t,_,e se,se th,t ,Nt ,e_ects som e s t, t e s b _ e, v ,. n the digits are continuousIY being mu1tiPlied In PalrS, and the _ u�U them, ,nd ,et,,_n, s,me othe, st,te b st,,ck,_, to ;.t Th s as ! right-hand digit of the prod4ct going to TePlaCe the nTSt dlglt�. ' .. ' ' l t4ken Start the machine and let it ''evo1ve'' ' what wi1l happen T�y ete_mlnate SyStem gOCS tO eQUlIlb,lum, so does it select. ' ' . . ' . i! _,�we hav, e,,d ,_ , ' _ OW Un er e aWS O lS Par lCU ar WO' t 'V'n 'meS�_, se_ect. the t,4th ,_s _ust the o o,,_te. . giVes even, and odd timeS odd gives odd. But eVe_ tlmeS Odd glVeS ,�_ , ' _ l _Oe O . _�equilibrium, performs tbe corresponding act of se1ectjon , Xe enCOUn teT t e eVe_ _S t e h etter C fa_Ce O d ___�_ Now e ul_llNb,__um ,_n sl_m _e s st m _ :. SY,ViVal. So aS thIS SYStem eVOlVeS, We ShaII See t e eVenS aVOre�Nnte,ést_Nng _t . b d l h . in the struggIe steadiIy repIacing the odds in the stores and�; l lS t e Pen U Um an_1ng Ve_lCalIy; lt lS the W4tCh ' . . _�, with jts main_s ,in _n d . t _ eVen Ua Y eX e'mlna 1_g em_ _�Tod, howe,e, we _now that 'whe, t e s m _. ' But the evens are not homogeneous, and among them the ZeTOS _�and d',am,Nc p u,__,Nb,,.um and tbe st4b,._,.t a,ound ,. are best suited to survive in this Particu1a_ WOrld; and, aS We _ ___�_ _'�_�

Page 10: selforganizing - University of California, San Diego

_???_? _ tllTswtln_shh_uyoageernswsvtolhen21qetehrtucoeelssaslnalarsrywe tehruangttol _tlhf.ellqutestl_T k_ pd t ., t h l_ t t_ t t N _ tt dt thd_ t _btlb_ RdE_au__NNsgINTN Eh Avl_cARt__ET_tyhdt t_hh_ _ y B_ u_t._ t4htoletonsne ___11(_ll_tll_ __l____e_________������_RINcI___s oF THE sELF-oRGANIzING sysTE_ 273��h we sh,tl see the ze,os exte,mi,at;ng the;, fello__evens ' ' in principle, in deveIoping synthetic organisms as compIex or as�nt__ éventu,_ly they ;,he,;t th;s p,,t;cula, ea,th ' _ inteIIigent as we pIease. ;�_s an examp_e of a thes;s of ext,eme gene,al_ , In this sense, then, ever_ machine can be thought of as ''self- ;�_ty. F,om o,e po;nt of v;ew we have s;mp_y a _,e__ denned ope,atoT organizing'', for it will deveIop, to such degree as its size and�_ pl;c4t;o, a,d ,epl,cement lawj which dT;ves on towa,ds comp1exity alIow, some functiona1 structure homologous with an 1�;_;b,;4m. _n do;ng so ;t ,,,,m,,,_,,(__ sele,ts those ope,,,ds ''adapted organism''. But does this give us what we at this Confer-�N,J/J, ,e,,_s,,,, to its cbange_maki,g tendency (Fo, ence are looking for? Only part1y; for nothing said so far has ;,�_q_,ely ,es;stant to change by mult;pt;cat;on) any impIication about the organization being good or bad; the _,u,�_s p,ocess of p,og,ess,.on towa,ds the specjally ,es;sta,t fo,m criterion that would make the distinction has not yet becn intro- !, _�_, h óf ex,,emé ge,e,,_,_ty dem,nd,_,g o,_y th,t the ope,,to, (o, thé duced. Tt is tTue, of course, that the deve1oped organism, being i _,,�n phys;ca_ _aws of a,y phýs_.cal system) be dete,m;nate and uncha,g_ _ stabIe, wiI1 have its own essential variables, and it wiII show its U'�'n This is the eneral or abstr_ct oint of view The biolo ist StablI1ty by VlgOrOUS reaCtlOnS t at tend tO prCSerVe ltS OWn t ,�?, ' . ' _ existence To ,_tsel ;ts own or anizat;on w;ll ,l_a s b dennit;�seeS a specla1 case of it when he observes the march of evo1ution, ' , , , _�Nva_ of the nttest, a,d the ;,ev;tab_e eme,gence of the h;ghest be good. The wasp _nds the stinging reAex a good thing, and the _,�biologicaI functions an_ intelligence. Thus, when we ask: What leeCh nn S t e OO -SUC 1ng re eX a gOO t I__'�Nfe and _ntell;ge_ce should appe_, 7 the answe, criteria come a_te_ the organization for survivaI; having seen what�,, ;s ,ot ca,bon o, amino ac;ds o, 4ny othe, spec;at featu,e b4t only s4rvives we then see what is ''good'' for that form. What emerges *�? that the dyn_m;c _,ws of tne p,ocess shou_d be ,J,,_,,g,N,g, ;.e. th,t depends simply on what are the sYstem_ laws and from what _�the system should be isoldted. In an_ isolate_ s_stem, l,_e a,d S a e 1 S aT e , eTe !S nO ImP! a lOn a e OTg4n!2�_ ;_re((;g,,çe ,_,ey,_,,6(_ deverop (they m,y ;n degene,ate cases deveIoped will be ''good'' in any absoIute sense, or according to _�Y _ deve1op to only ze,o deg_ee) ' ' the criterion oF any outside body such as ourselvest i n�_on. How c,n we gene,ate ;nte__;gence , To summarize brieAy: there is no dimculty, in principIe, in _ ,�J . 7 _ _ deve1oping s_nthetiç organisms as complex and 4s iMtelligent as _e _?�e lCa y. IS aS O OWS. a e a ynamlC SyS em _J OSe aWS , ' __,�a,e uncha,g;,g and s;,g_e_v4_ued ,nd whose s;,e ;s so _,,ge th,t ple_se. But we must notice two fundamenta1 qualincations; _rst, t ='�_ afte, ;t h,s gone to an equ;t;b,;.um th,t ;,volves only, sm,_l their intelligence wiI1 be an adaPtation to, and a specialization _�Mu fraction of its total states this small fTactjon js still la,ge eno,gh tOWardS, their pa,tICUIar enVlrOnmentt Wlth nO lmPllCatlOn Of l _�_ t jl f d d l f h d b _ . L . validity for any other environment such as ours t and secondly ! ' J�O a OW rOOm Or 4 gOO ea O C ange an e aVlOr. et It gO On ' ' t _�fo, a long enough t;me to get to such 4n equ;l;b,;um Then e_4m;ne their inteIligence wi1I be directed towards keeping their own t ,�_j;b,ium ;n deta;t you w;11 nnd that the states o, fo,ms essentiaI variabIes within Iimits. They wiI1 be fundamentaIly _�_n be;ng a,e pecul;a,ly ab_e to su,vive 4g,i,st the ch4,ges seInsh. So we now have to ask: In view of these quaIincations,�_nduced by the _4ws spl__t the equl_l,_b,__um l_n two ca_l one pa,t can we yet turn these processes to our a dvantage ? t,�organism'' and the other part "environment'': you wilI nnd that !, _,,�u this ''organism'' is pecu1iarly able to survive against the distur-�bances from this ''environment''. The degree of adaptation and _�Nty that this o,gan;sm can develop ;s boun_ed on_y by In this matter I do not think enough attention has yet been ! ,,�he s;2e of the who_e dyn4m;c system and by the tl_me ove, wh,.ch paid to Shannon's Tenth Theorem (1949) or to the simpIer ''law�it ;s ,_lowed to p,og,ess tow_,ds equil;b,;um Th,s as _ s__.d of requisite variety'' in which I have expressed the same basIC ___�eve,y ;solated dete,m;nate dynam;c system w;ll develo o, an;sms idea (AShby, l958, a)t ShannOn'S theOTem SayS that if a COrreCtlOn- _??�_, env;,onme,_s The,e is thus no d__mcult y cbanneI has capacity H, then equivocation of amount H can be _�n,�������''_�

Page 11: selforganizing - University of California, San Diego

_uuu _ Kld_t H_lthb fthd d _bbff _ Nd T N_ t N _ _ _ _ _ ____________ phhroowcesosepas dahgyeegfnyN ythytehyma Isseyraveecpysepnhtelpeyobh_egeengoshoowgn (As_hybuypstslew96nh0oapttyc_)sonthmoa_wet _1tll_l1_t_______?u___'___?���_���274 _. goss AsHBy _�PRINCTPLES OF THE SELF-ORGANTZTNG SYSTEM��removed, 6ut _o more Shannon stated his theorem in the context . . . _ b IN d l b d__ d,� ' _ One dlreCt1on In w Ch I e leVe a great ea tO e rea I y IS-�Of te1ephOne OT SImllar commUnlc4tlon, but the foTm414tion is ? bl _ _ t d_ f t f d _�. . . . i COVera e, IS In e lSCOVery O neW ypeS O ynamIC prOCeSS. í,�JUS aS trUe O a lO OglC4 re_u atory C anne trylng to exert some _. f t mach__�sort of corrective controI He thought of the case with 4 lot of N d _ l h d d h !,� ' Spec1a1Ize , dependlng On eXaCt y W 4t pa,tS are USe an OW�meSSa_e and a 1Ittle error; the blolog1st faces the case where the the are ,o__ned to etheT But there are s stems of more net I,_ke ,�message'' is small but the disturbing errors are many and Iarge. co,st,uct__on __n wh_Nch what ha p pens can onI y be t,eated stat,Nst__ ,�The theorem can then be applied to the brain (or any other ,, c4IIy The,e 4,e p,ocesses he,e l,_ke fo, ,Nnstance the sp,ead of J�reg4Iatory and selectlve device) when it sayS that the amount of Nd N h A N f N í I N ' .t 1, 'H�. . ' epl emICS, t e UCtUatlOnS O anlm4 pOpU atlOnS OVer 4 terrl OrY,�regU atOry Or Se eCtlVe aCtlOn that tbe braln Ca_ 4chleve ls abso- _k t Th 1_� t e Sprea O WaVe- l e p enOmen4 OVer a nerVe-ne. eSe i _�Ute y OUn e y ItS CapaClty aS a Ch4nnel (AShby, l958, b). r l_ _ n l_t r d noT bad but the ex1_st i, ,__� Another way of expressing the same idea is to say that 4ny qUa__tjty ! . . ' . '. d b I _ b N Nll ' _ _ _�. ' Wlth all theIT CUrlOUS pTOpertleS, and OU t eSS t e Taln Wl USe _�O f apprOprlate Se leCtlOn demands the tranSm1SSlon or processing them should the be of advanta e w_at _ want to em h4s_Nze l __�8 of quantity K of information (Ashby, 1960, b.) T_ere is no getting ! he,e __s that they often show ve,y su,_,__s__ng 4nd pecul_Na, p,ope,t_.es. , _,�O_SeleCtiO__Or nOt_i__. such a, the tendenc ,_n e l_deml_cs fo, the outb,e,ks to occU, ;� _�I think that here we have a princip1e that we shaIl he4, much s h lN ' ' _ s m,y be J_�. . . WaVeS. UC peCU lar new prOpeTtle�, Of In t e UtUre, Or It omlnates aII work with comp1ex systems hN d _ t and th4t he mlN ht othe,w__�_ t h h I f . f 1 maC lne eSlgner Wan S,�_ en erS e SU JeC SOmeW at aS t e aW O COnSe,V4tlOn O t c l_ ve _'� energy en ters power eng ineer ing. W hen t hat Iaw nrst came in, T h e s t u d o f s u c h s s t e m s m u s t b e e s s e n t _. a l j s t a t__s t_N c aI b u t __� a bout a hun dred years ago many engineers thought of it as a h. d h h b N d_ Nd lI t _ tN ;?�. . . ' t lS OeS nOt me4n t at eaC SyStem mUSt e In 1Vl Ua y S OC aS lC.�ISappOlntment, Or It Stopped aII hopeS of perpetuaI motion. on the contrar l_�NevertheIess, it did in fact Iead to the great practical engineering no s ystem can have g,eate, em,,_enc y than the dete;m,Nn,íe when , _�,! triumphs of the nineteenth century, because it made power act,.ng as a ,eg,_,to,. so as ,egulat,_on _Ns the one func€I.on tbat !_ ;,�, _ eng ineer ing more rea I is t ic. c o u n t s b _. o I o ,. c a I I w_ c 4 n e x e c t t h a t n 4 t u, a l s e _ e c t _N o n w _N l I h 4 v e _ _�, I suggest that when the fu1I implications of Shannon's Tenth m4de the b,a__n as dete,m__nate as poss_Nble _t folIows that we c4n _ _, _�U Theorem are grasped we shall be, nrst sobered, and then heIped, conn,e ou, ,,nte,est to the lesseT ,ange ,Nnwh__chthe samp_espace,_s '�' for we shal1 then be able to focus our activities on the probIems oye,a setofmech,n,_smse,chofwhl_,hl_s,_ndl_v,_du,IIydete,m;,4te , _�that are proper1y reaIistic, and actualIy solvable. As 4 a,tl_cu_a, case t f system that dese,ves much mo,e , ' __�, I�thorough investigation is the Iarge system that is built of parts _,,�HE _uTuRE that have many states of equilibrium. Such systems are extremely I� common in the terrestriaI worId ' they exist all around us, and in ! _�ere I h4Ve COmpleted thlS blrd's-eye survey of the principles fact l_ntelll_ el,ce as we know l.t wouId be 4_most _Nm oss;bIe other_ !�t hat govern t _e se1f-organizing system. l hope I have given Justi_- w_Ns é (Ashby _g6o d j Th,.s ,_s anothe, way of ,efeT,__ng to the ! l, '_,�cation for n1y beIief that these principIes, based on the 1ogic of system _,ho_e va,,_ábles behave Ia,gejy as pa,t funct__o,s _ have í _�mechanism and on information theory, are now essentially shownejsewhe,e(Ashby _g,o ajth,tsuchsystemstendio show l1,_�comPIete, in the sense that there is now no area that is grossly _ h,b,_tu4t,_on (ex_,_nct,_o,j ' a_d _o be abje to ad,pt p,og,e,s;vejy ! ,,�mys ter i_us. Ashb _g6o d . b I. h f h Il !!�. T e,e lS re4SOn tO e 1eVe t at SOme O e We - 1�efOre I end, hOWeVer, I WOuld Ilke to lndlCate very brieAy k b t b b_ j _ I h h dNt_ _n __�. . . . , nOWn U O SCUre lO OgICa p enOmena SUC a , __�e dlreCtlOnS In W IC UtUre TeSearch seems to me to be most N , . , _ I f t _ tN f h s1.o _,�. aSSOCI4 lOn_ an ennIngS a_V O e reSO U lOn O ' , __�l e y tO e pro ta e. lo 1N _ T sl,m I and d1_rect ex ress_.ons __,���l uM�����_��! 0'�

Page 12: selforganizing - University of California, San Diego

_ _e___u_____ ___ ntawssptod_oasc_ehnooevehrrbuhnacmmeowAa_tattodottpalNnhweeIsenld_eoaptsgeennodhreyvbfea'act Tenbpe'aslrl_ybkbtmebeltaraoultllamt.saamtrehcsesoltl_uatalnaaNstltl__hcrcldntmonTatNeehfetcpar_mrmhaav_elnrrolnsNveet_eo_aeemtylerya_ceoxg_wptl_etlttep_hoaeeyohsnn_tleNwsnatehhtoeeuhvroerebrelsenea_rt Tffph_alrlIagoea_elronl__plr_snognv4yoaesrc__vlNTywonhuTsesttpd__etsoheoae_ehmualx_rstl_ttcgtsokelh_t_l_rdleedtheesl_wNyfewnaunnothg_anpeafa_vdcarxmteseotntulsNalNtpNdhoccIoeonatherfr Tpaetv4spyosnfefycorefndtseNounJh_Ntxv_tdmresmhtlcNmreteeaafynNo_rssmntsonl_nyt_,lN_n _mevsho_ptele_ts_ay4lcl_tyoberatl__IantodTlsreerpuy IhxaarntgetcmeoragoNesgfmerele_tea.lsl_dnhl IecltNtly_ctsests_ ____)__ll__ wbo_dsbT_swtl_yrye_u_meTtgto_InthnetnaNrpotalctmknIldmhttst_el_hl_hoaazm_ol_eymaecTc_nu_tstl_hlee_tps_(sotentthyetln_cooese Tselot__ttrs_ettetuaemml__ttpssmmnptbrwpdse_el)sy_ennvtc_eegcctto'7wre__e'evluNoeptndtl_hhcenr_teaeahlgl__dsocsbbsl__seh_vlenltr'eoetsa._slNtlfle_T_2swtcllntNeaaf_frtteyntahttotl_tosdohteohn Tdenltalotasayhtbhnen_sty_elteedtgftcsnhb_tmheohseytt'_erahaclc_gllb'_mtrlen_otacl_a_httnnenesheeaatct)ll.dleecc2telllh_g_JpogsaanTnrl_fteeresagoastnqvtbl_ft_cl.un_asuhmonseyIscNotenrgtslla_NtleltattJuNetemhloaotltNmocsreefgaetlryen_ly_ponaenaot_duanpsroendregp Ttbetabbarpyr_oetolmoeelyfnftponfoteomssr_d TarautelwNeee_a_eIncolfnnttdeyhyl_Nfettt _1___tt__l_l1_l(t__llll______l________e____u__?_____?u?���������f the _nultipIicity of equilibria1 states. At the moment _ 4m resources foT its inves t iga t ion. Sure ly t he t i m e h a s c o m e f o r t h e���_ t ' _�the sort of learning that Piaget has specialIy considered-may _�be an a4tom_ric process when the input comes to a system with ' _,,�many equ iI ibr ia. Be t hat as it may, t here can be l itt Ie dou bt suMMARy�that the study of such systems is Iikely to reveat a variety of . l _,�����?g_Y_;__,e___hI_tt "nth;sqeU_hl ,b;'_e;' ;th; stY;y';t , 'Ub_JeeCt_;v� h_ Os !e 1Yc,m,eY'b_';;UT�_,y we _now ,x,,,(, wh,t we me,, _ ___����unknown, although Beurle (1g56) has made a most inte,esti,g as Tigorously as t he mat hema t ic ian un ders tan ds ' ' c o n t i _ u i t y ' ' O r ' _�Start. They deSerVe extensIve lnvestlgatlon; for, with their basic COnVergenCe _ .�����that it may have some pec,lia,, and useful, p,ope,ty not ,eadily These are not carbon, water, or any ot her mateT ia l en t i t ieS !,!��! J Reference to the system with many equitibria br;ngs me to the operator that is bot h unc hang ing an d s ing le-va lue d_ _ Y e_ _�i _ second tine of investig4tion that seems to me to be in the hi_,hest operatoT forces the deve Iopment o f i ts own form o f l i f e a _ d ;, _��'' _ _ _e _ _ ;,, P; i _, ',; '_ ' ^th ge id é nt _' _ ;, Oti _ n o _ S _' _s _ _y Y _,',, _ ;at 7 ';; e _ O ' _ ' " ' S m S ' " _u _ g _e i _ i ihe forms devcloped be or use to _s ? HeTe the s ituat ion ; ?_ _�_ _ _, At the moment, ou, _now_edge of the Iivi,g b,a_n is g,ossIy _ is dominated by the b_s ic law o f requ is ite var ie ty (an d S ha n n o n ' S _ _����'ng to add yet more de€ail. B4t when we ask what happens on t he pTocess ing o f a t lea s t t h a t q u a n t i t y o f i n f o r m a t I o n, _ U t U T e _��next, practically nothing is known. yet it is these longer-term has s_arted. _,�'gnincant o,es in human behavio, _ _inalIy, l commend _s a program for researc h, t he i de_ t2 _c a t i O M i�'f the,e is one thi,g that is c,y_ng o_ t_e p_ysic_l basis o_' r_e b_ain's _1emory sro,es. Our _now Ie dge , l _,�out to be i^vesti �ated it is t he physic,l basis of the b,,ints memo,y_ _ of the brain's funct ion ing is to day gross 1y ou t o f ba lan c e_ A v a s t t!�stores. There was a time when ccmemory't was a ve,y vag4e 4nd amount is known a bout how t he bra in goes from s t a t e t O S t a t e l,�.n bg J ' o t, 'e, b è; pn h; s of d _h Y e p _ ;t g __ ;i h _ Mp ' _es Oe _ Y, ' 'á;; å i pd g be O;; _h é I! _ S; _,N s nof _ n _ e ;_ V m p ó,t _ ;t _o; __ ge; m mh ;n _ é _ w é _ _ _it j l?����������?�? _,�

Page 13: selforganizing - University of California, San Diego

___??_____? ____?____u__ _l _ __ t _ t tN t N N _ t ____l___ sbTllRhuN cwTuhf R_t E d _ l_ t t bdtl b hf b l ld h hh _ dt ll___ll1 _ ___?___?u��_ .�_ . __v�ug ' _�__, ?�_�_, . BY _ !�? _�_ 1 REFEREYCES�' !_�J l. W. Ross AsHBy, The physicaI origin of adaptation by trial and erroT, _.�__ '_ GeM. Ps_'c_ol. 31, pp. 1 3-25 (l 945). , O_ ,�2, W. Ross AsHBy, Principles of the seIf-organizing dynamic system. _. - : nu�_ Gen. Psyc_ol. 37, pp. I25-8 (l947). R w spERRy _�_,' 3. W. Ross AsHBv, An Introduclion ro C_'berJ1erics, Wiley, New York, 3rd __�; J imp. (1958, a). C_l_iornja Instit_te o_Technology ; u,�4. W. Ross AsHBy, Requisite variety and its implications for the control of j 1�complex systems, C_be,_erica, 1, pp. 83-99 (l958, b).�5. W. Ross AsHBy, The mechanism of habituation. rn: r_e Mec___n1'z4r,'on o_ ORDERLY FUNCTION VVITH DISORDERED ' _�t_ou_I1tP_ocesses. (Natl. Phys. Lab. Symposium No. IO) H.M.S.O., Lon- ' '�don (I960).�_ 6. W. Ross AsHBY, Computers and decision-making, New S_ieI7risr 7 p 746 ; !,'�lg6o,b). _�_,, _, 1. W. Ross AsHBy, The bTain as reg4latoT, N4ru,e, LoM_. 186, p. 4I3 (I96o, c). I sha1l try to make my comments as briefas possibIe, for I suspect ' _�8. W. ROSS ASHBY, Desi_n _or 4 B_ain,' t_e O_i_in o_ Ad4ptive de__vior, !, that a,yt_ng that I can contrjbute here js likely to be toward t_e _�_ Wiley, New York, 2nd ed. (1960, d). ! _ . . . .�? g. L. voN BERTALANFFy, An outline of gene,al system theo,y B,,_f _ p_,_J perlp ery O OUr SympOSlUm. W4S Un er t e mlSappre enSlOn�_,, Sci. 1, pp. I34-65 (I950). that a majority of the conferees wou1d be engineers interested in '�10_ R. L_ BEuRL_, ProPerties of a mass of cells caPable of regenerating pulses, , dev;si,g ;ngen;ous ci,cuits fo, servomechanisms and the like, who _�Proc. Roy. Soc. B2_0 pp 5S-94 (1956) > . .�ll. W.R.GAiNERand_ J McGILL The,elat;onbetween,_nfo,mat__onand wouldh4Ve a tlmely concern for the prO em O Ul Ing COmpaC�variance anaIyses, Ps_c_omerrina 21, pp. 219-28 (l956). control boxes that wouId continue to controI after being nicked�l2_ R,C,_EFFREY, SomerecentsimPli_cationsofthetheoryof6niteautomata. by a meteo;te peppe,ed by ,adiation or even pjerced by a few�_ Technical Report 2l9, Research Laboratory of Electronics, Mass4- . '. . . ' _.�chusetts_nstituteofTechnology(21Maylgsgj. U etS. lt t Slnmln, pa_ OSar yreCa1__teaWaYS�_ l3. H. S. JENNINGs, Be__vior o_ r_e Lowe_ Organisms, Yew York (19o6). remarkabIe capacity of the cerebral machinery to continue to�l4. A. _. LOTKA, _Iements o_PJlysic4l Biolo__, Williams & Wilkins, Baltimore ope,ate e_ectively and in ,elatjvely oTderly ma,ner in the face of ,,�_ ( I92S).�9 l5. _. G. MARcH and J. A. s_moN, o,g_,;z_r;o,s, _viley, New yo,k (__5g) eXtenSlVe damage tO the StTUCtUre_ ,,�_ 16. K. H. PRIBRAM, Fifteenth International Congress of psychology, Brussels __ _ As most of you know, when large Iesions are in_icted in the�(1951). ' brain or even a whoIe lobe removed Ijke the tempoTal or fTontaI,�I1, C. E. SHnNNoN and W. WEAvER, T_e M_rI_em4tic_I r_eor_ o_Comm_nic4- ' . . ' N�_o, unive,sity of __linois p,ess u,bana (1g4g) _ tbe Cerebral maC_nery may COntlnUe tO Operate ln SUC gOO __ '� ' l g. G. �om,ERHo,,, A,,l,,,_,4I B,_ ;J,g,,, oxfo,d un,.,e,s,_ty p,ess, Lond,n f,s_.on that most of us could not detect any funct;onaI change.�n, (l950)N Tt often Tequires very sensjtive tests indeed to detect even maJor�_ n .l unilateraI damage, whiIe for some brain areas the tests have yet ! _�l to be dev;sed that w;11 expose the defect. _, _�_ Another point in regard to the preservation of function after�Iarge Iesions is the fact that the nearby cortical netwoTks generaIIy ;,�continue to work e�ectiveIy right to the very edge of the damaged�tissue. _or example, the outline of a bIind spot in the visual _eld !,�produced by a cortica1 1esion may be mapped by having the subiect !, _�follow a moving object as it approaches from di_erent directions�to disappear into the blind area.�RemarkabIe preservation of function is seen aIso after di�use�damage like that found in approaching seniIity or after disease�. 2Jg _�l!'_�' t,?�;_����_ _ - _�