Artificial intelligence in small bites
Machine Learning Workshop – Talkies

© Lead Image © robsnowstock, 123RF.com
Natural language games highlight artificial intelligence's quest for comprehension.
While thinking about intelligence and its operation dates back at least to the Greek philosophers [1], discussion of creating an artificial brain only credibly emerged after Charles Babbage and Ada Lovelace's work on the Analytical Engine [2]. The idea that a mechanical device could be built possessing the ability to reason was further boosted by the appearance of digital computers in the 1940s and a highly symbolic language such as Lisp the 1960s [3], reducing indirection to just another set of parenthesis and thus enabling the efficient encoding and processing of large sets of interdependent rules.
Then progress hit a wall due to the impossibility of encoding in declarative fashion all the rules (some of which were manifestly yet unknown) required to reason from a premise to a conclusion. Nevertheless, interesting experiments emerged, with their results best illustrated by some games built with this design strategy.
Gaming Intelligence
MIT's ELIZA [4] is a set of simple yet clever scripts encoding a few simple strategies to answer a natural language prompt (you can try a modern version online [5]). ELIZA attempts to impersonate a Rogerian psychotherapist (a practitioner of person-centered therapy) by exploiting the reliably predictable subject-verb-object sentence structure found in the English language to mechanically turn statements into questions, or manipulate one clause into another:
[...]
Buy this article as PDF
(incl. VAT)