Semantics and Intelligence: Creating an Effective Public Turing Test
This is a whitepaper we drafted in partnership wth researchers at Texas A&M University as part of a follow-up to a study published last August by Messrs. Shardul Vikram, Shinan Fan, and Guofei Gu of the Computer Science Engineering department at Texas A&M University.
Messrs. Vikram, Fan, and Gu had independently posited a two-factor approach to a Turing test that was almost identical to the system we created with VouchSafe, so they were excited about the prospect of working together on a follow-up paper.
You can read the full text of this publication here: Semantics and Intelligence: Creating an Effective Public Turing Test.
This paper was written in response to a study published last August by Messrs. Shardul Vikram, Shinan Fan, and Guofei Gu of the Computer Science Engineering department at Texas A&M University called SEMAGE: A New Image-based Two-Factor CAPTCHA. The paper described a new approach to creating a Turing test based on semantic associations.
In a completely independent project, my company had developed a similar two-factor Turing test that leverages semantic relationships between objects.
In this paper we examine the thinking that lead to the design of a semantic Turing test and the evolution of a commercial product. We also contrast this product with the SEMAGE study and examine how a Turing test can be extended from using only first-order definitive relationships to embrace some of the more complex functional, contextual, and emotive relationships that humans intuit between objects.
We also examine some of the quantitative and anecdotal data drawn from our observation of actual product performance and compare that to the results of the SEMAGE study.