In the simplest form, the Turing Test is a test to confirm if humans have developed an Artificial General Intelligent (AGI) system by facing a human against another human and the AGI system. The human facing the other two parties asks both of them a set of questions, to keep this simple we will call them the “questioner”. The questioner, after a specified time, will receive their answers, and on each response, the questioner has to decide which response is from the human and which one is from the machine.
The more “points” that are won by the machine will prove that the questioner was fooled by the machine and not the human.
However, this is exactly it. The Turing Test, created by Alan Turing, is only supposed to be a test for how easy it is for machines to fool humans. Turning did not specifically state that the test was to be used for gaining a level of intelligence. But it is to be used for demonstrating whether a machine can “think”. Furthermore, using this testing method for actual intelligence level testing invites a negative connotation with the idea that AI is able to “fool” humans. While it certainly is possible for AI to fool us, we want to focus on working with AI.
Therefore, how can we test an AI’s level of intelligence for AGI?
Some may say why not use the usual IQ test? The reason this would be a bad idea is that the IQ test is designed for humans. If we measured intelligence by the ability to per
As opposed to judging whether AGI can think as humans do, a better approach would be to assess its meaningful contributions. The creators of an AGI system can set out a meeting to deliberate around an important topic. AGI would be achieved when it makes contributions that are equivalent or better than a human. However, simply interjecting the conversation with a fact or an agreement is not a useful contribution, it has to be a valuable one where its input has been taken from the best of its knowledge; it has correctly (or better said “logically”) “read the room” and extracted the meaning of words said, body language and tonality used, along with their context.
Once the meeting concludes, the AGI team should have a clear understanding of any warnings and opportunities this instance of the AGI presents. The warnings will allow the team to add any further appropriate rules, that allow them to stay ahead of any intelligence explosion, and the opportunities are used to better understand how powerful the AGI is and whether it is safe for use within consumer, commercial or public markets.
