“Of course, machines cannot think as people do. A machine is different from a person. Hence, they think differently. The interesting question is, just because something thinks differently from you, does that mean it is not thinking?”
This quote from the movie, “The Imitation Game” stuck with me long after my popcorn was devoured, and the end credits rolled.
More recently, I again pondered the quote as I contemplated the recent advancements made in terms of AI chatbots such as ChatGPT and Google’s Bard.
The ‘Then’: History of the Turing Test
The “imitation game” as conceptualized by Alan Turing in 1950, has long been used as the benchmark for testing artificial intelligence, more specifically the ability of a machine to be as “human as possible” in terms of its responses and thus demonstrating its so-called “thinking”.
The premise essentially is that if the machine manages to convince the participants that it is human, it has passed the test.
In testimony to the gamesmanship of programmers, a chatbot named Eliza was already created as early as 1966. It was cleverly named after Eliza Doolittle from the play “Pygmalion”.
In the play, a professor of phonetics accepts the challenge to transform the speech of an every-day flower seller to such an extent that she would be able to pass muster as a duchess. Not unlike the premise of the Turing Test!
Years later, the very first Turing Test Competition was held by Hugh Loebner on 8 November 1991. Throughout the years of the competition being held, significant deviations were made from the original Turing Test.
Initially it was easy to detect the bots in the competition as the mistakes were very noticeable but as time passed, the bots became more sophisticated and programmers arguably even more innovative.
The ‘Now’
In 2014, a chatbot called Eugene Goostman was alleged to be the first successful chatbot to pass the test. Its programming was designed to mimic that of a Ukrainian boy.
However, this assertion was not without some controversy as it was alleged that the grammatical errors and lack of knowledge were incorporated as part of the backstory that supported the “boy’s” age and limited language ability.
Be that as it may, these competitions have now become an annual event with the chatbot Mitsuku (“Kuki”) being considered as the most successful due to its prolific winning streak. It has won the Loebner prize in 2013, 2016, 2017, 2018 and 2019.
Google’s LaMDA was also alleged to have passed the Turing Test in 2022. Although, this assertion was also not met without controversy.
As a follow up, Google also presented an AI chatbot known as “Bard”, that is based on the language model used by LaMDA.
More recently it has also been submitted that ChatGPT is the newest candidate that passed the Turing Test.
AI bots such as Bard and ChatGPT, present exciting possibilities in terms of the natural language models and machine learning that is utilized. Time will tell in what ways the respective AI bots will distinguish itself in the market.
In part 2 of my blog, I will discuss the Turing Test and my opinion on its best use in the modern era.