Creativity, the Turing Test, and the (Better) Lovelace Test Selmer Bringsjord, Paul Bello  , and David Ferrucci The  Mind...
A Pleasure to Be at the Genesis Point of So Much Lunacy Logic Theorist in 1956 put us on the brink of “thinking machines”?...
Judge What color and in what style is your hair? The Turing Test
Propositions Like:  “If  x  passes TT (or TTT, for that matter), then  x  can genuinely think” are Demonstrably False <ul>...
Like Kasparov vs. Deep Blue… Joel Benjamin Murray Campbell … Kasparov
Judge What color and in what style is your hair? Designer In the TT, it’s  Really  Judge vs. Designer I can handle tha- uh...
Lady Lovelace, Who Gave the One Objection Turing Took Seriously in the 1950 Paper, Had It Right “Computers can’t be creati...
Turing’s First Response <ul><li>“ Ah, but the Analytical Engine was a universal Turing Machine!… </li></ul><ul><li>“ Some ...
Turing’s Second Response (recently reiterated by Moravec) <ul><li>“Lady Lovelace was never surprised by a program which sh...
Judge (= Designer) Designer (= Judge) The Lovelace Test o S How did it do that?
Toward  a Definition of Passing LT (We’re working on the formal definition) <ul><li>Computer system  S , designed (or unde...
But you built this system called BRUTUS over a period of seven years, and BRUTUS is creative, no?
BRUTUS is… <ul><li>Reverse engineered </li></ul><ul><li>Might well reveal the creativity of its  creators , Bringsjord & F...
Objection from Child-Machine <ul><li>Turing responded later in his seminal paper that Lady Lovelace’s objection would be s...
Rebuttal <ul><li>ANNs equivalent to logical deduction. </li></ul><ul><li>If judge/designer follows training of ANN, or ass...
So What About Oracle Machines? <ul><li>After all, by Copeland’s lights, Searle can’t run his CRA on an O-machine. </li></u...
The Answer <ul><li>“ Free Will” in the agent causation sense. </li></ul><ul><li>This is the way to get the kind of true au...
Upcoming SlideShare
Loading in …5
×

Creativity, the Turing Test, and the (Better) Lovelace Test

734 views

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
734
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
12
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Creativity, the Turing Test, and the (Better) Lovelace Test

  1. 1. Creativity, the Turing Test, and the (Better) Lovelace Test Selmer Bringsjord, Paul Bello  , and David Ferrucci The Minds & Machines Laboratory Department of Philosophy, Psychology & Cognitive Science Department of Computer Science RPI Troy NY USA IBM T.J. Watson Research Center (D.F.) [email_address] www.rpi.edu/~brings
  2. 2. A Pleasure to Be at the Genesis Point of So Much Lunacy Logic Theorist in 1956 put us on the brink of “thinking machines”? The celebretary pyrotechnics of New Year’s Eve have died away, and we are left with the sobering fact that AI has given us systems with the conversational punch of, perhaps , an infant.
  3. 3. Judge What color and in what style is your hair? The Turing Test
  4. 4. Propositions Like: “If x passes TT (or TTT, for that matter), then x can genuinely think” are Demonstrably False <ul><li>“ Could, How Could We Tell If, and Why Should – Androids Have Minds?” in Android Epistemology </li></ul><ul><li>Searle’s CRA/Jonah mono savant WRCCB </li></ul><ul><ul><li>Copeland: “Person x declares that argument A is invalid” Therefore “A is invalid” This is invalid! </li></ul></ul><ul><li>“ The Zombie Attack on the Computational Conception of Mind” PPR </li></ul><ul><li>… at last count 15 separate deductive arguments </li></ul><ul><li>But, today: The Structure of the Turing Test </li></ul>
  5. 5. Like Kasparov vs. Deep Blue… Joel Benjamin Murray Campbell … Kasparov
  6. 6. Judge What color and in what style is your hair? Designer In the TT, it’s Really Judge vs. Designer I can handle tha- uh, it can handle that one.
  7. 7. Lady Lovelace, Who Gave the One Objection Turing Took Seriously in the 1950 Paper, Had It Right “Computers can’t be creative. For to be creative is to originate something. But computers originate nothing. They merely follow the programs given to them.”
  8. 8. Turing’s First Response <ul><li>“ Ah, but the Analytical Engine was a universal Turing Machine!… </li></ul><ul><li>“ Some computers can think for themselves. Let M be such a machine… </li></ul><ul><li>“ Ergo, she was wrong – because the AE can perfectly simulate M !” </li></ul><ul><li>Problem: Lovelace would never have accepted the existence of such an M . </li></ul><ul><li>Indeed, where are such machines, even today? </li></ul>
  9. 9. Turing’s Second Response (recently reiterated by Moravec) <ul><li>“Lady Lovelace was never surprised by a program which she then had to debug.” </li></ul><ul><li>But </li></ul><ul><ul><li>In syntactic errors no origination </li></ul></ul><ul><ul><li>In semantic errors ditto </li></ul></ul><ul><li>Camry to quasi-SUV story </li></ul><ul><li>Still, respecting their intuitions… </li></ul>
  10. 10. Judge (= Designer) Designer (= Judge) The Lovelace Test o S How did it do that?
  11. 11. Toward a Definition of Passing LT (We’re working on the formal definition) <ul><li>Computer system S , designed (or understood from the standpoint of a designer) by j , passes LT if and only if </li></ul><ul><li>S outputs o ; </li></ul><ul><li>S ’s outputting o is not the result of a fluke hardware error, but rather the result of processes S can repeat; </li></ul><ul><li>3 j cannot account for how S produced o . </li></ul>
  12. 12. But you built this system called BRUTUS over a period of seven years, and BRUTUS is creative, no?
  13. 13. BRUTUS is… <ul><li>Reverse engineered </li></ul><ul><li>Might well reveal the creativity of its creators , Bringsjord & Ferrucci </li></ul><ul><li>You can judge for yourself if you read the book AI and Literary Creativity: Inside the Mind of Brutus, A Storytelling Machine </li></ul><ul><li>Utterly fails the Lovelace Test </li></ul>
  14. 14. Objection from Child-Machine <ul><li>Turing responded later in his seminal paper that Lady Lovelace’s objection would be surmounted by the education of a “child-machine.” </li></ul><ul><li>A child-machine would be an impenetrable black box that would pass LT. </li></ul>
  15. 15. Rebuttal <ul><li>ANNs equivalent to logical deduction. </li></ul><ul><li>If judge/designer follows training of ANN, or assimilates corresponding deduction, there is no saying the system thinks for itself. </li></ul><ul><li>If the child-machine goes beyond connectionism toward something else (processing beyond Turing Limit), would that do the trick to pass LT?… </li></ul>
  16. 16. So What About Oracle Machines? <ul><li>After all, by Copeland’s lights, Searle can’t run his CRA on an O-machine. </li></ul><ul><li>But the problem – which Copeland seems unaware of – is that oracle machines, once made precise (through e.g., “Weyl Machines”) can be shown to be mere mindless symbol manipulators by trivial extension of CRA. </li></ul>
  17. 17. The Answer <ul><li>“ Free Will” in the agent causation sense. </li></ul><ul><li>This is the way to get the kind of true autonomy Hofstadter has realized is essential for being truly creative </li></ul><ul><li>Bring back the “Free Will” Chapter in (1992) What Robots Can and Can’t Be ! </li></ul>

×