Artificial Intelligence Has Greatly Outgrown the Turing Test as a Measure ofits Capabilities
In the world of artificial intelligence (AI), the Turing Test, first proposed by the renowned mathematician and computer scientist Alan Turing in 1950, has long been the benchmark for measuring an AI's ability to exhibit intelligent behaviour indistinguishable from that of a human. However, a new framework is challenging this traditional test, offering a more comprehensive evaluation of an AI's general intelligence.
This new AI Classification Framework, inspired by the Theory of Multiple Intelligences proposed by psychologist Howard Gardner in 1983, was recently introduced by Chris Saad, the former head of product development at Uber. The framework, an 8-part evaluation, considers the Turing Test and its language-testing ability as only one part of the overall evaluation.
The Theory of Multiple Intelligences identifies 8 separate intelligences: logical-mathematical, linguistic-verbal, visual-spatial, musical-rhythmic, bodily-kinesthetic, interpersonal, intrapersonal, and existential. Under this new framework, ChatGPT, the popular chatbot, has a long way to go before being considered truly "intelligent."
When applied to ChatGPT, it shows average human intelligence in logical-mathematical and linguistic-verbal intelligence but scores N/A in all other categories. This suggests that while ChatGPT can hold a conversation and solve simple mathematical problems, it struggles with other aspects of intelligence, such as visual-spatial, musical-rhythmic, and existential intelligence.
The Turing Test, despite its historical significance, has faced criticism over the years. Critics argue that the test primarily focuses on language ability and ignores other critical aspects of intelligence. Proposals for amending or replacing the Turing Test have been around for years, and the AI Classification Framework is one such proposal.
In this new framework, AI is rated on a 1 to 5 scale in each intelligence category, with "1" being essentially non-existent or infant-like and "5" being super-intelligence. While some have argued that ChatGPT has passed the Turing Test, it has a lot of studying to do if it wants to pass the final exam called human consciousness.
The Turing Test involves a human participant asking questions to both a human and a computer, with the computer passing if it successfully tricks the questioner into thinking it's a human. Many more programs have purported to pass the Turing Test since then, with Google's AI LaMDA passing the test and controversially convincing a Google engineer that it was "sentient."
The first attempt at passing the Turing Test came in the mid-1960s with a chatbot named Eliza. In 2014, the first AI reportedly passing the Turing Test was Eugene Goostman, a program designed to simulate the responses of a 15-year-old Ukrainian boy.
The impact of the Turing Test and the AI Classification Framework on the field of AI and its development cannot be overstated. The Turing Test, named after its creator, has pushed the boundaries of AI research, while the new framework offers a more holistic approach to evaluating AI intelligence.
Darren Orf, the author of this article, lives in Portland and writes/edits about sci-fi and how our world works. His previous work can be found at Gizmodo and Paste.
