the turing test debate --up to nowrhorn/a/kmap/arg/artcltheturingtstdebat… · test itself (our...

7
The Turing Test Debate --Up to Now (a draft chapter prepared for the forthcoming book Turing Test Sourcebook) by Robert E. Horn Visiting Scholar, Stanford University Distinguished Consulting Faculty, Saybrook Graduate School © 2002, R.E. Horn The Turing debate as to whether computers will ever be able to think, is one of the great debates of recent times. It is as yet unresolved. Recently our team published an argumentation mapping approach to the overall debate with considerable attention to the Turing Test itself. (Horn, et. al. 1998a) This set of maps are part of a larger series that is intended to create visual navigational tools for intellectual history. (Horn, 2000a, b) The "maps" are large diagrams that connect claims, rebuttals, and counterrebuttals together so that a person unfamiliar with the debates has a convenient way of seeing the structure of the debates as well as the detailed arguments. It literally provides a “map” of the debates. Figure 1 presents an example of one of 7 of the maps in the series.

Upload: others

Post on 04-Jul-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

The Turing Test Debate --Up to Now

(a draft chapter prepared for the forthcoming bookTuring Test Sourcebook)

byRobert E. Horn

Visiting Scholar, Stanford UniversityDistinguished Consulting Faculty, Saybrook Graduate School

© 2002, R.E. Horn

The Turing debate as to whether computers will ever be able to think, is one of the greatdebates of recent times. It is as yet unresolved. Recently our team published anargumentation mapping approach to the overall debate with considerable attention to theTuring Test itself. (Horn, et. al. 1998a) This set of maps are part of a larger series that isintended to create visual navigational tools for intellectual history. (Horn, 2000a, b) The"maps" are large diagrams that connect claims, rebuttals, and counterrebuttals together sothat a person unfamiliar with the debates has a convenient way of seeing the structure ofthe debates as well as the detailed arguments. It literally provides a “map” of the debates.Figure 1 presents an example of one of 7 of the maps in the series.

Page 2: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

Figure 1. Can the Turing Test Determine Whether Computers Can Think? Map 2of the Mapping Great Debates: Can Computers Think? series. Reproduced by permission

of MacroVU Press. www.macrovu.com

This map is one of seven in the first series that together provide a complete frameworkand context for the Turing Test debates discussed in this book. The debate was, ofcourse, initiated by Alan Turing in his article in Mind (Turing, 1950). This unleashedseveral decades of debate about the many aspects of this proposition. Turing said, 'Ibelieve that at the end of the century [i.e. by two years ago] ... one will be able to speakof machines thinking without expecting to be contradicted." Well, 800 major moves inthe argument later, we are still debating the subject. I say 800 major moves in theargument because that is the count we made after diagramming this argument. Wedivided the debates into some 70 subarguments, listed in Figure 2.

My discussion in this chapter will cover a significant number of the copies listed underMap 2 (in Figure 2) “Can the Turing test determine whether computers can think? Yet,because of the intricate interrelatedness of the threads of arguments over the decades,some threads will have to be arbitrarily truncated. That is a limitation of the book as amedium for presenting such navigational maps as ours. Pare of the justification formapping arguments is that one can begin to understand the overall structure of thedebates in ways that other structures (such as prose paragraphs) can not. Another benefitof the argumentation templates is that one can quickly scan and find specific clusters ofarguments and rapidly drill down to detail.

Page 3: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

Figure 2. The subarguments of the basic question "Can Computers Think?Reproduced by permission of MacroVU Press. www.macrovu.com

Point of view of this chapter

Our maps provide a way of looking at the debate as it stood in late 1997 when the finalprepublication work was done. This chapter will thus provides an overview of the

Page 4: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

structure of the argument up to that date and – and in addition - provides a way for thereader of this book to determine several important questions:- What new lines of argument have developed since our series of maps were published?- What lines of argument have been extended by new rebuttals, counter-rebuttals, or

evidence? (The argumentation maps, in addition to displaying the intellectual historyof the subdebates, provide an easy way of seeing where the debates have stopped orpaused. You can read the last rebuttal in each thread of the argument on the righthand side of each thread.)

- What new ways of framing the arguments have been offered since then?- What claims or rebuttals have not been replied to? (And what can we infer about

that?)

In this chapter I will present the set of subquestions having solely to do with the TuringTest itself (our Map 2), not with the questions opened up by the wider Turing debate.Thus the reader who is interested in such subquestions as those concerned with theGoedel or Penrose debates or the Chinese Room arguments or the Heidiggerianperspectives will have to consult Maps 7, 4, and 3 respectively. These are obviouslyrelevant to questions taken up by various authors of chapters in this book, but it would beimpractical to discuss all major claims, evidence, and rebuttals from these maps in thischapter – for each of the maps has over 100 major moves summarized. These widerissues, of course, contradict Turing’s denial that people might legitimately debate andeven achieve wide agreement about how thinking might be adequately described anddefined by the use of his test.

Understanding the Turing Test Itself

One must first understand the test itself. We began by providing a visual illustration ofthe various versions of the test. Turing starts by envisioning a guessing game played by awoman and a man as described in Figure 3.

Page 5: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

Figure 3. The Imitation GameReproduced by permission of MacroVU Press. www.macrovu.com

Then he elaborates by replacing one of these humans with a computer, and, in our maps,we connect it with his major claim (see Figure 4). The numbered boxes show how ourmap analyzes the debate into “claims” (Figure 4, box 1) and “supports” (boxes 2 and 3)Sidebars explain the players and define essential terms.

Page 6: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

Figure 4. The Turing Test.Reproduced by permission of MacroVU Press. www.macrovu.com

This then is the paradigm about which so much has been debated. Because there are somany interpretations of the test based on so many worldviews of the protagonists in thedebate, the debates have gone off in many directions. It can be argued that the test willnever be conclusive because test results always have to be interpreted and interpretationsalways are subject to further debate (This could be argued, but, in fact, is not an argumentthat appears on one of our argumentation maps.) So, how did the debate proceed afterTuring? In the remainder of this chapter I will present sections of our maps on thesubsequent 50 years of the debate.

Page 7: The Turing Test Debate --Up to Nowrhorn/a/kmap/arg/artclTheTuringTstDebat… · Test itself (our Map 2), not with the questions opened up by the wider Turing debate. Thus the reader

Does the imitation game help determine whether computers can think?

Here then is the first of the major threads of debate.

Figure 5. How good is the imitation game?.Reproduced by permission of MacroVU Press. www.macrovu.com

MORE TO COME