Artificial intelligence (AI) is all the buzz these days, with the explosion of usage of generative AI tools like ChatGPT, DALL-E, and Bard, impacting everything from content writing to art to network engineering. We have heard many people voicing concern about the unrestricted growth of AI, while others say that AI is just a tool and no replacement for human creativity. On the other side, there were reports that the US Air Force conducted a simulation where an AI drone decided to kill its operator to prevent them from interfering with its mission. President Biden met with AI experts just yesterday on the dangers of AI. So do we worry? Are we creating a HAL or a Terminator, or is this just another step in technology? How can we tell if AI is reaching consciousness?
The Turing test was developed by the British mathematician and computer scientist Alan Turing in 1950. Turing was curious about whether machines can actually think. Turing's test is between an evaluator and two subjects, a human and a machine. The evaluator asks questions to determine whether the subject is a human or a computer. Also called the "imitation game", Turing could see that computers could eventually handle such massive operations that they might be able to make decisions in a way similar to humans. Turing's work listed many of the concerns brought up by contrary views, including mathematical limitations like Gödel's theorem, differences between the human nervous system and discrete-state machines, and "Lady Lovelace's Objection", which Ada Lovelace stated “The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform.” These are similar to many of today's views, where AI is a tool, not any more of a threat than other technologies that can be used or misused, depending on the will of the user.
So have any systems ever passed the Turing Test? In 2014, a chatbot called Eugene Goostman reportedly passed the test posing as a 15-year-old Ukrainian boy. More recently, Google's LaMDA passed the test and its inventor even called it "sentient". ChatGPT took the test and did not pass, but the folks at OpenAI believe that a future version, maybe GPT-5, might be able to pass it. However, given the growing complexity of AI systems, some think that the Turing test is outdated and have proposed a new paradigm called the AI Classification Framework (ACF) that takes in eight different aspects.
Are the fears of AI grounded in true threats, or are they unfounded concerns that will never materialize? Please share your thoughts below as I leave you with a quote from Blade Runner, where the Voight-Kampff test was used to tell whether someone was artificial:
Oh, a final note... ChatGPT took the Voight-Kampff test and failed miserably!
Image citations: Scott, Ridley. 1982. Blade Runner. United States: Warner Bros.
I can't resist posting this because my involvement in machine learning back in the '80s forced me to deal with the dreaded S word.
