Artificial Intelligence Test
The Double-Blind Human-Level Cognition (HLC) Test
By now, just about anybody who has taken an introductory AI course will have heard of the Turing test, as a means to test if one is dealing with a human or an AI contributor. If you haven't heard about it, you can read up on it online. There have been many authors, who over the years, have detailed or referenced this test. Unfortunately, this test is simply nonsense, and will never deliver a reliable means to discern, if there truly exists a machine capable of abilities normally only attributed to humans. The reason for this shortcoming can be found in the simple observation, that in the original Turing test, tester and testee know that they are being tested, which indubitably leads to distortions and misconceptions with regard to the test itself. And, these distortions ultimately add up to invalidate any gleaned test result. Why? Consider for a moment, that the pharmaceutical industry has for many years now been using double-blind studies when testing new drugs. In a double-blind study, neither the doctor who adminsters a drug, nor his patient know who receives a placebo, or who doesn't. This is to avoid the placebo-effect which may distort a patients judgement with regard to the efficacy of the administered drug.
The double-blind HLC test
For exactly the same reasons, the original Turing test should take these possible distortions and misconceptions into account and be modified accordingly. But, how do you conduct a double-blind study in which the creators of AI and their human users are not aware of a test being conducted? Exactly, at that point in time when AI becomes immersive. At this moment, neither the AI's creators, nor it's human users have any longer a vested interest in pushing their personal agendas. In other words, the AI creator no longer is interested in touting/hyping his software being the best or greatest in machine intelligence, and the human user no longer feels the need to behave differently then he would if engaging with another human. This scenario can only be achieved when AI has become immersive and accepted by humans as if it was one of their own. As an example, recall the various news stories involving self-driving cars and human drivers behaving "badly" thereby forcing unnecessary accidents. Just as important as the previously stated points is the influence the media has in propagating balanced accounts of the touted AI systems. Propagating hype only raises unrealizable hopes with some and instills unnecessary fears in others. Fear mongering, - AI systems wlll become our overlords, or steal our jobs, or usurp our democracy and subjungate us in our daily decidion making process - often lacks any responsible form of narrative.
AI vs Human-level Cognition (HLC)
By now most people will have heard of systems like ChatGPT, and others from a select number of popular media outlets. The populist media hype machine has made great strides towards touting a future filled with intelligent machines deciding for us our next trip, what we should eat, wear, create the next TV shows, make movies, do our taxes, write reports, hand in our daily homework and so on and so forth. Much of this will not come to fruitition. Remember the days you were told self-driving cars will be all present, drones will deliver your next package to your housedoor, you will be living with Elon Musk on Mars and never have to pay taxes again, or machines will perform surgery on you. Well, take a long look around and what do you see? Practically, nothing. Your life progresses much as it has before, more or less.
The reason the newly touted intelligent systems hardly represent the danger they are made out to be, is simply they are not capable of human-level cognition. ChatGPT and its brethren are nothing more than examples of imitation software (ie., mimicry). Learning by imitation is well known and you can find examples of the "Monkey see, monkey do" metaphor in then animal kingdom. For example, a bear cub observing a human angler catch fish with a pole, may decide to copy this behavior by trying to extend a branch into the water, or a baboon in an open zoo may decide to snatch a persons camera/phone and shoot a selfie of himself, because he observes people engaging in this act on a daily basis. These examples may make for funny anecdotes, but they are hardly comprable to HLC.
If you still feel that ChatGPT and the other contenders are intelligent, then you might as well believe that a talking parrot is capable of HLC. And, with regard to the doom and gloom stories of machines becoming our overlords and replacing us in our jobs? Of course there will be changes. This happens all the time. Remember the 1980s as we were told that the newly arrived bank ATMs would lead to massive branch closures and bank tellers would be replaced within years. Well, it's 2023 and when I go to my house bank, I see a lot of bank employees - yes, even bank tellers - trying to service their customers. And, with regard to those highly touted, job destroying ATMs, it appears one or the other would find a better home in a museum.
