Turing Test: Mindless Game?

The Turing Test, originally configured for a human to distinguish between an unseen man and unseen woman through a text-based conversational measure of gender, is the ultimate test for thinking. So conceived Alan Turing when he replaced the woman with a machine. His assertion, that once a machine deceived a human judge into believing that they were the human, then that machine should be attributed with intelligence. But is the Turing Test nothing more than a mindless game? We present results from recent Loebner Prizes, a platform for the Turing Test, and find that machines in the contest appear conversationally worse rather than better, from 2004 to 2006, showing a downward trend in highest scores awarded to them by human judges. Thus the machines are not thinking in the same way as a human intelligent entity would.

Not another look at the Turing Test!

Practical application of the Turing Test throws up all sorts of questions regarding the nature of intelligence in both machines and humans. For example - Can machines tell original jokes? What would this mean to a machine if it did so? It has been found that acting as an interrogator even top philosophers can be fooled into thinking a machine is human and/or a human is a machine - why is this? Is it that the machine is performing well or is it that the philosopher is performing badly? All these questions, and more, will be considered. Just what does the Turing test tell us about machines and humans? Actual transcripts will be considered with startling results.

Assumption of knowledge and the Chinese Room in Turing test interrogation

Whilst common sense knowledge has been well researched in terms of intelligence and (in particular) artificial intelligence, specific, factual knowledge also plays a critical part in practice. When it comes to testing for intelligence, testing for factual knowledge is, in every-day life, frequently used as a front line tool. This paper presents new results which were the outcome of a series of practical Turing tests held on 23rd June 2012 at Bletchley Park, England. The focus of this paper is on the employment of specific knowledge testing by interrogators. Of interest are prejudiced assumptions made by interrogators as to what they believe should be widely known and subsequently the conclusions drawn if an entity does or does not appear to know a particular fact known to the interrogator. The paper is not at all about the performance of machines or hidden humans but rather the strategies based on assumptions of Turing test interrogators. Full, unedited transcripts from the tests are shown for the reader as working examples. As a result, it might be possible to draw critical conclusions with regard to the nature of human concepts of intelligence, in terms of the role played by specific, factual knowledge in our understanding of intelligence, whether this is exhibited by a human or a machine. This is specifically intended as a position paper, firstly by claiming that practicalising Turing's test is a useful exercise throwing light on how we humans think, and secondly, by taking a potentially controversial stance, because some interrogators adopt a solipsist questioning style of hidden entities with a view that it is a thinking intelligent human if it thinks like them and knows what they know. The paper is aimed at opening discussion with regard to the different aspects considered.

Chatterbox challenge as a test-bed for synthetic emotions

Chatterbox Challenge is an annual web-based contest for artificial conversational systems, ACE. The 2010 instantiation was the tenth consecutive contest held between March and June in the 60th year following the publication of Alan Turing’s influential disquisition ‘computing machinery and intelligence’. Loosely based on Turing’s viva voca interrogator-hidden witness imitation game, a thought experiment to ascertain a machine’s capacity to respond satisfactorily to unrestricted questions, the contest provides a platform for technology comparison and evaluation. This paper provides an insight into emotion content in the entries since the 2005 Chatterbox Challenge. The authors find that synthetic textual systems, none of which are backed by academic or industry funding, are, on the whole and more than half a century since Weizenbaum’s natural language understanding experiment, little further than Eliza in terms of expressing emotion in dialogue. This may be a failure on the part of the academic AI community for ignoring the Turing test as an engineering challenge.

Testing Turing's five minutes, parallel-paired imitation game

Purpose – The purpose of this paper is to consider Turing's two tests for machine intelligence: the parallel-paired, three-participants game presented in his 1950 paper, and the “jury-service” one-to-one measure described two years later in a radio broadcast. Both versions were instantiated in practical Turing tests during the 18th Loebner Prize for artificial intelligence hosted at the University of Reading, UK, in October 2008. This involved jury-service tests in the preliminary phase and parallel-paired in the final phase. Design/methodology/approach – Almost 100 test results from the final have been evaluated and this paper reports some intriguing nuances which arose as a result of the unique contest. Findings – In the 2008 competition, Turing's 30 per cent pass rate is not achieved by any machine in the parallel-paired tests but Turing's modified prediction: “at least in a hundred years time” is remembered. Originality/value – The paper presents actual responses from “modern Elizas” to human interrogators during contest dialogues that show considerable improvement in artificial conversational entities (ACE). Unlike their ancestor – Weizenbaum's natural language understanding system – ACE are now able to recall, share information and disclose personal interests.

Some implications of a sample of practical Turing tests

A series of imitation games involving 3-participant (simultaneous comparison of two hidden entities) and 2-participant (direct interrogation of a hidden entity) were conducted at Bletchley Park on the 100th anniversary of Alan Turing’s birth: 23 June 2012. From the ongoing analysis of over 150 games involving (expert and non-expert, males and females, adults and child) judges, machines and hidden humans (foils for the machines), we present six particular conversations that took place between human judges and a hidden entity that produced unexpected results. From this sample we focus on features of Turing’s machine intelligence test that the mathematician/code breaker did not consider in his examination for machine thinking: the subjective nature of attributing intelligence to another mind.

Human misidentification in Turing tests

This paper presents some important issues on misidentification of human interlocutors in text-based communication during practical Turing tests. The study here presents transcripts in which human judges succumbed to theconfederate effect, misidentifying hidden human foils for machines. An attempt is made to assess the reasons for this. The practical Turing tests in question were held on 23 June 2012 at Bletchley Park, England. A selection of actual full transcripts from the tests is shown and an analysis is given in each case. As a result of these tests, conclusions are drawn with regard to the sort of strategies which can perhaps lead to erroneous conclusions when one is involved as an interrogator. Such results also serve to indicate conversational directions to avoid for those machine designers who wish to create a conversational entity that performs well on the Turing test.

Effects of lying in practical Turing tests

Interpretation of utterances affects an interrogator’s determination of human from machine during live Turing tests. Here, we consider transcripts realised as a result of a series of practical Turing tests that were held on 23 June 2012 at Bletchley Park, England. The focus in this paper is to consider the effects of lying and truth-telling on the human judges by the hidden entities, whether human or a machine. Turing test transcripts provide a glimpse into short text communication, the type that occurs in emails: how does the reader determine truth from the content of a stranger’s textual message? Different types of lying in the conversations are explored, and the judge’s attribution of human or machine is investigated in each test.