Research  in  Shape  grammars, computational  design, computer  ad  design  and  design  theory. Is  the  director  of  the  Shape  Computation  Lab, a  research  group  that  explores  how the  visual  nature  of  Shape  can  be  formally implemented  with  new  technologies  to enable  new  paradigms  in  computer  ad  design. Design  automation,  visual  scripting,  and  creative  design. I'm  really  excited  about  this  speaker  today. He  both  in  the  College of  Computing  and  in  the  School  of  Design. So  let's  welcome  Dr. Things  Sec  1212,  can  you  hear  me  well, Hopefully  the  action  will  not  be  too  strong. Hopefully  at  least  as  strong  as  the  name. Okay,  delighted. Let  me  check  here  my  time. Delighted  that  I'm  here  with  you  Today  I  will  talk  about the  project  that  I  think  you will  hopefully  find  it  fascinating. It  is  the  work  of  the  last, definitely  ten  years  of  work  that  you  do  at  the  lab. But  in  many  ways, it's  really  the  whole  life sort  of  worth  of  work  to  make  it  happen. It  is  not  easy  what  you  will  be  seeing  today. The  presentation  is  called  Shape  Machine. From  software  to  practice, I  will  outline  the  problem, the  software  they  have  really  put together  to  tackle  these  issues. And  then  the  exciting  things that  are  happening  here  at  Georgia  Tech, from  architecture, from  design  students  and computing  students  that  they  take  this  on. Let's  see  how  do  that.  Yeah,  sometimes analogies  are  good  to  start  a  topic. Of  course,  they  fail  if  you  push  them  too  much, but  they're  good  to  start  a  visual  analogy. You're  all  very  familiar,  of course,  with  word  processing. The  idea  of  finding  things  that  you  want  in a  text  because  you  want  to  find  them, the  rest  of  the  document  because  you  want  to  change them  in  whatever  you  want  to  do  with  that, But  once  you  circle  it  in  a  PDF  or  a  Word  or  anything, because  there  is  the  database  behind, they  can  find  all  these  instances. It's  not  anymore  the  thing  that  you  did  on  the  screen, but  it's  really  the  underlying  representation that  allows  it  to  be found  everywhere  in  the  document  and  to  the  Internet. In  the  Internet  for  that  matter. So  that  sort  of  structure  of  letters  to  words, to  sentences  to  what  have  you  has revolutionized  word  processing  and  made  a  big  thing, I  don't  know,  50,  60  years  ago  from  various  professions. Note  that  notaries  were the  first  profession  that  actually  got the  word  processing  because  they  were  able to  change  contracts  within  minutes. They  would  just  change  some  names, addresses,  and  they  would  make  some  good  money  for  that. So  what  would  that  mean? That's  the  analogy  really, if  we  had  the  same  kind  of thing  with  the  characters and  all,  and  the  words  and  everything. But  what  would  it  mean  if  we  had other  media,  other  symbols? If  we  had  a  plan,  for  example, Could  we  circuit  details  of  that? Could  we  find  instances  of  that? Could  we  change  things  like  that  in architectural  design  or  basically  in  every  profession, every  discipline  that  uses  sakes  as a  major  venue  for  recording  and  communicating  ideas. What  if  it  is  chemical  engineer? What  is  electrical  engineer? Circuits  whatever  it  is  with the  same  simplicity  and eagerness  and  that  you  can  do  with  word  processing. Here  is  mechanical  engineering. Finding  relationships  of  finding  parts  of  things. Anything  really  that  stands  out  to  you, not  to  the,  not  to  the  database, but  really  to  your  own  eyes,  that  sort  of  thing. Let  me  see  if  this  will  work  right  there.  All  right? Okay.  That  idea  that  I  just  explained  to  you when  about  four  years  ago  we  went  around  and  said, we  have  a  solution  around  this  topic. Cdl,  Creative  Destruction  Lab, the  Create  here  at  Georgia  Tech. The  Georgia,  the  GRA, they  had  an  interest  in  the  work  that  we  did. We  tried  to  put  together a  video  that  communicate  these  ideas. You  see  here,  for  example, that  drawings  in  the  PDF  are  entirely  useless. You  cannot  really,  you  cannot  really query  them  in  any  way,  nonsevere. We're  claiming  back  then  that  was  really  our  prototype, that  we  can  actually  have  ways  of  making  them  work  in the  very  same  way  that  word  processing  works. So  you  see  here,  for  example, the  user  selects  a  particular  thing. This  is  a  column, and  the  design  tells  you  therefore. And  the  software  says  that  there are  four  instances  or  it's a  special  relationship  between  a  column  and  the  wall. And  the  software  would be  clever  enough  to  tell  you  that  there  are two  instances  of  this  particular  relationship of  the  column  to  the  wall. So  that's  the  first  question  of  the  day. Why  is  that  that  defined  in the  replace  operations  that  are  so essentially  in  Word  or Excel  have  yet  to  be  implemented  in  Cad. Why  am  I  here  and  sort  telling you  this  staff,  what's  going  on  here? It  seems  very  important, don't  you  think,  right? Why?  So  that's  really  the  first  topic. Well,  there's  a  good  answer  to  that, because  shapes  have  no  unique  representation. It  sounds  bizarre  just  to  say  that. What  do  you  mean?  Like  we  have all  these  incredible  drawings,  models. We  have  Autocad,  we  have  all  this  stuff. What  do  you  mean  they  have  no  unique  representations? Well,  in  order  to  demonstrate  that  those  of  you  that you  work  with  any  sort  of  digital  graphics, you  know  exactly  what  I'm  talking  about. But  in  general,  we  made,  again, this  video  to  demonstrate the  ridiculousness  and  the  complexity  of  this  argument. Here  we  pretend  we  have  a  user  that  wants  to find  a  vector  file  to  work  with. So  rather  than,  you  know, sort  of  downloads  the  document  from  the  web, tries  to  open  it,  it  looks  like  a  plan,  it's  fantastic. But  then  when  she  tries  to work  in  the  instances  you  see  something  that perhaps  some  of  you  recognize  broken  lines.  It's  a  mess. It's  an  absolute  mess. Most  of  the  time,  the  architects,  designers, they  make  a  big  cup  of coffee  and  they  clean  up  the  drawing. That's  what  they  spent  a  good  time  in  the  office, taking  a  thing  entry,  working  it  entirely. And  here  you  see  when  you  rotate  the  drawing, you  might  even  have  even  bigger  problems like  asymptotes,  like, you  know,  mistakes  of  the  translational  schemes between  the  vector  graphics, or  mistakes  of  the  users, or  who  knows  whose  mistakes. It's  an  absolute  mess,  right? Imagine  now  here,  once  you  go  through this  particular  sort  of  codification, what  you  see  as  an  arc  is  an  arc. What  you  see  as  a  line  is  an  R  and  all  that. Again,  this  was  our  attempt  back  then to  like  three  years  ago  to  show  what  does  that  mean? And  of  course,  once you  assume  that  there  is  something  like  that, the  drawings,  they  can  have  a  unique  representation, then  you  can  actually  quit  them. And  this  is  really  where  the  biggest  problem  exists. Unique  safe  representation  does  not  mean  structure. This  is  really  a  mindset  that  you  have to  change  your  work  about  your  world, about  how  you  look  at  stuff. You  do  not  need  to  have  well defined  structures  to  search  things. In  fact,  that  what  makes  safe say  what  makes  designers  different. And  you  all  know  from  your  friends, you  know  if  you  are  designers, you  are  slightly  different  from  the  engineers. And  all  related  and  different here  you  see  how  we assume  that  you  can  select  parts  in  design, you  can  you  can  edit  these  parts. And  the  interface  here  in  Rhino  in  three  D,  modeling, sort  of  nerves  modeler  and now  the  machine  would  look  at  all  the  instances. Right?  You  see  the  instances  and  before  we clean  them  up  and  look  now all  the  intersections  are  cleaned  up. The  very  same  way  that  you  would  use  a  character in  Word  and  you  would  change  your  name  for  another  name. Here  you  use  that  and  of  course, once  you  have  a  drawing,  you  don't need  to  look  up  a  details. But  what  makes  this  drawing  a  drawing  or  a  plan? Because  we're,  they're  standing  at  their  walls, their  staircases,  their  windows. Well,  if  you  understand  them, the  computer  can  understand  them  too. That's  really  the  idea  without having  them  restructured  in  advance. I  guess  there's  so  many  different  ways from  talking  about  that  in  my  classes, but  one  way  to  think  about  it  is  that  you  have put  the  carriage  before  the  horse, so  to  speak,  in  computer  graphics, and  then  we're  trying  to  come up  with  solutions  around  it. Is  this  something  new? Again,  the  oddity  of  me coming  here  in  2023  and  telling  you about  instances  and  drawings seems  very  essential  to  be  on  the  side. Here  is  an  excerpt  I  really  like  from  Malcolm  Call, Executive  Director  in  Auto  Desk  2006, writing  in  one  of  the  most  sort  of  popular  magazines, Architectural  Design,  Architectural  Design,  AD. And  talking  back  then  about  the  incredible  input, influx  of  programming  in  design. And  he  says  in  2006, he  describes  his  experience  20  years  before  that. In  86,  when  he  was  at  UCLA  working  with  this  guy, George  Stein,  at  UCLA, he  was  his  student,  and  I  was  also  George  student. So  he  says  back  then, you  know,  there  was  something  interesting. It  was  called  Safe  Grammars. A  desk.  We  knew  about  them, and  we  knew  that  there  was  something  wrong  with  the  way the  whole  system  has  been  set up  because  it  creates  these  black  boxes. However,  down  there,  the  keys  of  death, the  better  formal  structures had  no  radio  application,  no  need  to  use  them. That  lack  of  interest,  because  it's  complicated. Is  it  because  the  whole  industry went  on  the  pattern  recognition  on  the  image, on  the  pixel  based  stuff. Who  knows?  A  variety  of  reasons  why  this  thing  happened. What  he  was  referring  to  was  this  is the  milestone  of  all  the  work that  we  will  be  presenting  today. This  is  the  1975  books of  the  two  dissertations  that  came  out, one  from  George  Tiny  and  the  other  from  James  Gibbs. Very  close  friends  in  their  '20s,  They  wrote  this. They  got  the  Best  Paper  award  for  their  idea. It  was  a  popularization  of  the  Chomsky's  sort  of  work, but  using  directly  saves  as  opposed  to  symbols. Since  then  they  keep  on  reworking. Yeah,  this  idea  here, I  like  that  this  is  really  the  first  paragraph, the  absolute  first  paragraph, of  James  Gibbs  book  on  the  safe  grammars  and  their  uses. Safe  grammars  provide  a  means  for the  recursive  specification  of  saves. The  formalist  for  safe  grammars  is designed  to  be  easily  usable  and understandable  by  people  at the  same  time  to  be  adaptable for  using  computer  programs. A  big  stake,  really  a  big  claim,  right? Understandably,  those  of  you  that  you  have  an  issue with  programming  or  he  has  taken  you  some  time, perhaps  you  sympathize  with  this  sort  of  view. This  is  from  2015, about  40  years  since. The  original  book  that's  George  Stein  himself. He's  still  around  his  at  Mit.  He  still  works. He  had  a  book  last  year,  highly  recommended. You  will  see  the  featuring  in the  rest  of  the  presentation. This  is  by  making  grammars, and  George  still  uses  his  rules. The  left  hand  side,  right  hand  side  by  a  pencil. So  what  is  the  basic  core  of  the  argument  we  have? Save  every  pair  of  shapes  have  a  rule. Now,  how  can  I  use  a  rule? When  do  I  use  the  use  in  a  better  sense? When  do  I  use  a  rule  if  there  is a  transformation  that  makes  the  rule  the  same. But  I'm  actually  looking  for  part  of  the  design  itself. It  doesn't  need  to  be  there  before. If  it  is  for  me,  if  I  search  to  find  it, then  that's  all  I  need. If  in  fact  there  is  a  Sarsation  matches  that, subtract  that  transformation  and add  the  corresponding  right  hand  side. It  straightforward. Let's  look  at  it  pectoral, it  always  makes  more  sense. You  see  here  the  rule  says, if  you  find  a  lower  case  K, connect  the  endpoints  with  this  funny  sort  of sap  and  polyline  and  apply  to  the  two  squares. Mind  you,  there  is no  lower  case  in  these  two  squares.  I  drew  two  squares. That's  the  whole  idea,  right? I  don't  need  to  anticipate  who  will  look  at  what  for, what  they  would  look  in the  model  if  there  is  a  shape  and  yes, for  example,  it  is  why it  is  because  I  can  embed  it  into  that. I  can  embed  it,  I  sort  of  move  it  around. You  can  do  it  with  a  trace  paper  or  you  can  do  it with  computers  and  you  can  embed  it. And  if  you  can  embed  it,  you  subtract  it. And  then  you  add  the  next  piece of  the  line  notes  here  that  all  of  that  stuff, you  don't  need  to  specify  them  in  advance. And  then  of  course,  then  in  the  end use  them  and  make  a  new  shape  like  that. Now  you  would  expect  that  this  would happen  for  all  instances  in  the  design. And  here  you  see,  for  example, because  the  symmetry  of  two  squares  is  eight, for  those  of  you  that  like  and  enjoy  symmetry  theory, there  are  eight  different  partitions. So  the  computer  or  the  human  might  find eight  matches  and  might  have eight  possible  designs  to  work  with. What  that  thing  is  is  that  it  is  the  typical  use, especially  all  of  you,  that  you  do computer  graphics  at  very  well. If  I  have  a  shape  for  example, and  apply  transformation, it's  straightforward.  That's  really  what  we  do. Computer  graphics,  I  have  a  shape, I  transform  it, and  I  have  something  exciting  to  work  with. Here  is  the  inverse  of  it. I  have  two  shapes  and  I  try  to  see whether  there  is  a  transformation that  matches  the  one  to  the  other. It  is  not  always  the  case. It  might  be  that  there  is  no transformation  that  can  do  that. If,  for  example,  you  look  at this  K  and  I  put  this  kin  to  square, so  you  can  actually  see  that  a  bit  better. You  can  switch  your  eyes  back  and  forth  between the  outline  and  the  shape  in  between. If  you  start  with  a  K,  there  are  this  one  shows  here. There  are  18  different  kinds  of Ks  that  we  can  actually  talk  about. Now  if  you  look  only  about  the  outside  the  boundary, the  original  square,  it  can be  translated,  can  be  rotated. It  can  be  scaled.  All  of  these  are  squares. Or  it  can  be  stretched,  become  rectangle,  can  become  bus. It  become,  become  a  parallelogram. Can  become  a  trapezoid  or  a  quadrilateral. And  all  of  them  underneath  the  reflections, we  can  be  freakingly  precise  about  what  we  mean.  Find  x. There  are  18  different  things  that we  can  find  on  this  particular  case, and  for  all  of  them,  of  course. Zillions  of  transformation  that  we  need  to  match. Now,  the  actual  algorithm  can  be  there  forever. Those  of  you  lab  geometry,  you  know  it  in  fact, that  this  is  really  something  quite standard  to  see  how  can  you calculate  transformations  from  the  one  to the  B  here  is  from  1986. Not  a  very  particularly  great  one, but  actually  let's  show  that this  was  published  and  all  that. If  I  want  to  see  if  that  what  is the  transformation  that  makes  that  right? Then  first  you  translate  it. You  translate  the  large  one  there. In  this  particular  algorithm, you  move  everything  to  the  Cartesian  origin. Then  they  should  have  a  rotation  that  they don't  have,  and  then  a  scale. And  then  you  move  it  back  to  show  that  this  shape, it  is  part  of  that,  right? That's  really  the  algorithm.  The  algorithm  is  there. What  we  have  done  here,  I  mean, the  work  that  I'm  presenting today  is  really  the  work  with my  Phd  students  and  primarily  with  this  phenomenal  guy, Kurt  Hong,  who  finished  his  Phd  32  years  ago, and  now  he's  the  University  of  Kansas. We  have  this  presentation  that says  how  many  this  is  really  gained. Straightforward,  but  it's  nice  to  see  it  in  a  table. What  kind  of  information  do  you  need to  determine  whether  two  shapes  are  related? You  need  basically  one  point to  determine  whether  there  is  an  isometric. Imagine  you  have  a  tiling. You  want  to  find  the  tile,  you  want  to  put  it  together. All  you  need  is  one  point  of  the  tile, and  you  just  move  it  around. You  move  the  whole  thing  by  one  point. You  need  two  points  for  similarities, three  points  for  affinities, and  four  points  for  linearities. These  are  different  types  of  informations. All  I'm  trying  to  say  here  is that  this  is  the  bare  minimum. Of  course,  in  everything  out  there can  be  a  thousands  of  points. You  have  to  check  now. Yeah,  all  this  about  transformations, but  we're  missing  something  very  important. Tapes  are  unruly,  they  are  behaving, they  are  different  from  everything  else. Yes,  of  course  you  hear  transformations that  make  them  exciting  and  all  this  stuff, but  in  themselves  they  are  unruly, deeply  unruly,  so  much  deep. That  is  an  incredible  discussion  and historicity  on  the  relationship  of  symbolic  media,  text, numbers,  music  and  notes  to the  actual  shape  out  of  lines  and  planes  and  everything. So  here  is  again  from  George  Stine,  his  new  book, Shapes  of  Imagination,  Highly  recommended, fantastic  reading  to  those  of  you  that  you  write. You  enjoy  this  discussion  today  so  that  if  you  have two  Ls  and  you  have  structured  them  as  such, see  what  happens  in  the  flickering  of  the  eye  in the  moment  once  you  start relating  them  in  particular  ways. Here  is,  for  example,  when  you start  bringing  the  two  Ls  together, we  have  a  square  in  between  that  pops  out. If  you  shift  them  even  more  like  along  the  diagonal, we  have  this  really  bizarre  situation  with  three  squares. Some  eyes.  I  see  your  eyes. Actually,  your  eyebrows  go  like  what. But  now  I  hope  you  can  see  that. Then  if  you  really  make  it  even  more, suddenly  becomes  a  gestalt  square  with a  cross  or  four  squares  or  five  squares. You  start  counting  squares  here. Here  you  have  like  other  shapes, you  have  the  two  small  squares, four  reticle  around  the  sever  in  between. Here's  a  Greek  cross.  You  see what  I'm  trying  to  say  here? This  is  really  a  metaphor  for  creativity  in  design. Those  of  you  do  design, you  understand  when  you  fool  around  with  the  stuff, you  know  by  the  moment  you  sift  things  around, new  ideas  come  either  you  can  see them  yourself  by  rotating  the  paper  around, or  your  instructor  can  see  that, or  your  friends  can  see  that or  enjoying  your  favorite  beverage. You  can  change  your  mindset  and you  can  see  them  in  different  ways. But  that's  the  whole  point, changing  representations  like  that. There  are  two  problems  with  this  thing. Everybody  enjoys  it,  everybody  talks  about  it. How  do  you  go  about  it? That's  another  story.  That's  the  thing. This  is  where  the  autocad  put  like the  kiss  of  death  says  we  cannot  do  that. The  first  one  is  that  this  is computationally  very  difficult  if you  really  want  to  find  this  peer, I  mean  this  is  the  outline  of  a, per,  a  large  column, so  to  speak,  in  a  classical  plan  by  Palladium. Here  you  see  the  actual  shape that  you're  searching  here  is the  underlying  infinite  construction  lines, or  the  equation  lines  or descriptors  that  the  shape  is  embedded. If  you  want  to  find  out  how many  ways  you  can  do  that  here, if  you  look  only  for  concrete  copies, there  is  1,000  searches  you  need  to  make. Why?  Because  this  shape  here consists  out  of  240  lines,  1234567. It  consists  out  of  70  lines  that  they go  like  this  and  consists  out of  1,125  points  of  intersection. That's  really  what  the  database  gets. Points,  infant  lines  and  lines  that  you  see. If  you  want  to  find  out  this  1234  line  shape  here, if  it's  exactly  the  same  concrete  copy, you  need  1,000  if  you  want  to  find  scale  one. 1  million  searches  in  the  computer. If  you  want  to  find  under  affinities  stretches, you  need  1  billion  calculations. And  under  affinities  1 trillion  calculation,  it's  ridiculous. Really  says  that's  something  wrong  with  this  idea. Because  also  this  is  a  ridiculously  simple  drawing. Those  of  you  that  you  have  friends  or  you  are  designers, you  know  that  the  design  specifications can  be  very,  very  complex. The  other  problem,  of  course,  is  the  aspect  of  precision. Precision  meaning  that  you  have  two  lines. You  think  these  two  lines  are  the  same. But  if  you  really  zoom, you  realize  that  these  two  lines perhaps  you  have  different  ***gths. The  details  are  not  exactly  the  same. Of  course,  then  you  say  fuzz  it. The  problem  is  that  once  you  do  it  second,  third, fourth  time  that  you  do  this  fuzziness, the  calculation  fails  and the  computer  stops  and  cannot  have  any  answer  to. That  is  all  this happens  because  of  the  floating  points  and  all  that. Again,  these  two  problems  are  the  keys  of  death  of the  auto  described  several  efforts here  we  have  documented  70. I  show  up  here  on  the  screen  with  2013  with our  first  effort  in  doing a  graph  representation  with  my  former  Phd  team. Because  we  did  it  with  that  because it  thought  like  it  cannot  be  done. So  let's  use  graphs  because  somehow  they  are more  available  and  commercial  and  robust. We  abandoned  that  and  here  we  have  2020. We  were  able  to  solve through  the  engineering  of  the  same  machine, through  the  work  of  Kurt  Hong here  some  interfaces  of  things  of the  past  that  they  have  been familiar  in  everyone  who  does  this  sort  of  work. Here  we  are,  we  announced  here  that, in  fact  we  do  have  a  software  that  tackles the  two  ridiculous  impossibilities that  I  just  told  you  about  the  lines, the  complexity  and  the  imprecision. There's  a  new  computational  technology  that brings  say  based  search  and  say, base  replace  to  the  Cam  world. How  do  we  do  it  by addressing  these  two  things  that  I  told  you  and basically  taking  into  heart what  the  formalism  talks  about. The  book  that  I  recommended,  that  again, I  will  recommend  to  look  at Georgia's  new  book,  George  Stein's  book. It  talks  about  this  embed,  embed,  fuse, embed  fuse  cycle  once  you  do  art  and  we  do exactly  the  same  in the  underlying  representation  of  the  safe  machine. Let's  see  how  we  do  that. First  of  all  with  the  complexity, we  have  a  nice  formalism  there. It  is  called  save  signature  that recognizes  that  not  all  points  are  the  same. If  you  realize  that  you can  have  different  conditions  around  points, you  can  actually  see  the  design  in  profound  ways. You  can  see  that  when  we  have  this  point  signature, we  have  some  remarkable  improvement. That's  one  thing  that  we  do. This  is  the  most  important  thing  is  this. That  says  every  time  we  apply the  rule,  the  design  changes, the  actual  representation  changes in  front  of  view  is  exactly  the  same, but  underneath  it  changes so  that  you  can  never  have  a  hiccup  in the  calculation  of  the  shifting  representations. This,  once  we  put  it  together, we  did  two  things  to basically  announce  to  the  academic  community that  this  is  not  the  Holy  grall  has  been actually  figured  out  for at  least  straight  lines  and  arcs. And  now  we  have  to  climb  this  incredible  mountain  up. We  start  recreating  some  of the  classic  calculations  that have  been  proposed  forever  in  the  field. And  so  this  miraculous  calculation  we saves  here  is  says  if  you  see  a  K  shape, put  that  end  like  that, this  bracket,  make  this  U  shape. And  therefore  you  go  from  the  two  squares, you  go  to  this  octagon  and  the  pentagons  around  it. Pentagons,  yeah.  But  then  again, without  really  caring  about  any  database  constraints, you  can  always  impose  your  own  structure. We're  saying,  look  at  this  shape  here. This  smaller  K,  K, or  I  don't  know,  arrow  and  the  line. Put  it  in  any  way  you  like  it  here. The  machine  calculates. We  actually  ask  it  now  to  be  under  similar  transmission. How  large  you  want  it,  how  small. All  these  end  points  are  not  in  the  shape  itself. They  are  just  matched  upon  it. You  don't  need  to  have  them  like  registered  within  that. Here,  again,  just  to  show this  incredible  flexibility  with  the  shapes, imagine  that  we  say  you  find  that  K  shape. Now  make  this, it's  very  strange  to  actually  describe  that. I  mean,  what  is  this  two  lines  floating? That  they  connect  but  once  you  do  it. We  have  the  remarkable, again,  fluctuation  of  a  design  in  front  of  your  eye. And  this  is  the  eureka  moment  that every  designer  feels  once  they  start  drawing. And  with  a  pencil  and  so  forth. Not  with  Rhino,  not  with  auto  desk, not  with  autograd,  not  with  beam, nothing  but  with  the  pencil. That's  the  idea  behind  that. And  on  the  other  hand, we  wanted  to  show  as  well  one  of  the  other, again,  flags  on  the  field  to  see  how  you  can  do  design, that  can  be  more  meaningful. Imagine  now  left  hand  side, right  hand  side.  These  are  actions. If  we're  trying  to  create  an  Irish  here, you  want  to  verbally  discuss, describe  this  rule  if  you  see  an  arc  and  another  arc, but  extend  it  in  a  way  so  that the  two  end  points  are  perpendicular  one  to  the  other. How  much,  how  long? It's  impossible.  Designers  usually, so  most  of  the  times  actually  say, I  don't  know,  I  just  did  it. Or  if  they  know,  they  just  the  point, they  point  to  drawings,  they  point  to the  stuff  behind  them  that's  here. There  is  no  need  to  describe  that  with  words. You  just  draw  them. You  model  them,  and  you  say  you  ask the  machine  to  work  them  out. This  is  two  different  examples, one  for  changing  representations  easily  on  the  fly. The  other  taking  curves and  everything  else  in  creating  something  very  specific. That's  the  question  one  like  demonstrating the  field  tutors,  where  do  you  move  on? Question  number  two,  the  day, once  you  have  something  like  that, what  happened  to  go  from a  file  that  looks  word  processing  to  actually Rhinoscript  to  plus  like using  characters  to  use  a  programming  language. Is  that  possible  to  take the  shapes  and  make  a  programming  language  out  of it  by  introducing a  logical  processing  framework  with  stage  loops, jumps,  conditionals,  and  all  that. Is  it  possible  actually to  program  shapes  where  the  line  is  not  really  a  line, but  it's  actually  a  line  of code  but  really  a  line  specification. It  makes  your  mind  go around  trying  to  sort  out  these  conditions. We  have  also  good  news  on  this  topic  too. So  far  I've  been  talking  about  these  two  things. We  have  I  guess three  ways  of  working  with  the  safe  machine or  basically  21  is the  manual  and  the  other  is  the  automated.  The  manual. You  can  define  the  rules  from scratch  exactly  as  you've  seen  so  far. We  actually  saw  the  derivation  when  you  pick  up a  from  the  model  itself.  These  are  the  manuals. You  select  a  rule  and  you see  what  you  want  to  do  with  that. The  other  one  is  the  automated. We  put  conditionals,  jumps,  looks, and  we  put  them  in  a  particular  framework. And  we  call  it  draw  script  as the  DS  stuff  like  we'll  talk  about  it  later  on. The  discussion  here  is  some  of the  basic  frames  to  see  how  we  went  around  it. This  is  from  the  work  of  one, does  his  Phd  in  here  in  CS, but  he  also  enjoys very  much  the  work  with  the  same  grammar. So  we  work  together  with  Professor  Varma  on  that  has put  a  whole  manual  of  starting  to  work  with  draws. Now  you  see  for  example,  who  rules. There  are  three  rules  here. Every  rule  is  in  a  particular  window, like  windows  in  computing. And  every  window  has  inside  some  information. It  tells  you  under  which  transformation. The  left  hand  side  will  go  to  the  right  hand  side. Whether  you  will  look  it,  whether  you  will  apply  it serially  or  in  a  parallel  sort of  fashion.  What  does  this  say? If  I  see  a  square,  add  another  square  in, and  make  the  exterior  square  green. If  I  see  the  same  thing, the  counter  clockwise,  the  other  clockwise, if  I  see  a  green  thing,  make  it  black. And  apply  the  stuff  and  apply  it  four  times. Applied  four  times,  and  apply  to  once  here. Applied  it  for  all  green  things because  it's  really  parallel. This  is  serial  one  by  one. The  actual  calculation  is  exactly  what  you  would  expect. It  recognizes  the  square  and  moves  and  puts another  1  second  third  here,  one  to  three. And  then  it  goes  the  other  way  around  to  complete, and  then  in  the  end  becomes  black  once  you  start. Then  having  a  framework  of  calculations. You  can  see  where  this  thing  can  go. The  rules,  now  you  can  put  them  in  their  own  world. You  can  name  them  the  framework that  is  very  rudimentary,  Just  number  them. This  is  the  rule  number  one.  This  is  rule  number  two. This  says  when  you  start  jump  to  number  one, calculate  three  times  go  back, then  two  from  the  two  jump  to  two, go  four  times,  go  back. And  then  it's  the  same  thing  that  you  saw  before. But  now  introducing  the  jumps  from  block  to  block. We  can  have  conditionals  when  the  window  now becomes  green  or  it  has  weight. If  you  start  to  the  one, if  you  see  a  square  inside, go  to  two  and  apply  that  three  times, go  back  again  if  you  don't  see  anything, which  means  see  nothing. If  you  don't  see  anything,  yeah,  yeah. That  means  you  see  something  then  you  go  here, you  apply  that  three  times,  go  back, change  the  colors  and  all  of  them,  the  same  thing. We  keep  on  working  on  this  fashion. This  is  like  the  work  now, Van  Cluster,  he  does  his  Phd  in  music  Technology. He  got  fascinated  with  our  work  as  well. He  introduced  the  pseudorandom  generators  basically  to make  it  really  to  have the  rules  firing  ad  hoc  in  any  way. Very  much  like  a  programming  language. If  you  see  a  line  here  with  two  endpoints, Jagt  it  and  make one  more  point  and  make  a  deflection  like  that. And  then  in  the  end,  all  the  lines  and  then  you start  applying  very  much like  the  Lindy  stuff  that  you  can  do, Lindy,  but  do  them  like  this. I  don't  know  if  you've  done  a  liter my  or  stuff  like  that. It  would  be  nice  to  sow  the  actual  code next  to  it  is  what  it  takes  to  program with  the  my  three  and what  it  takes  to  do  it  in  shape  machine  with  draw  script. This  is  just  a  question  here,  just  to  be  funny, you  understand  that  what  I'm  talking  about  here, hopefully  you  see  it. Profound  is  really  very  new, like  it  is  off  the  beaten  path  in  computing  and  design.