In 1950, British mathematician Alan Turing (1912–1954) proposed an easy way to test artificial intelligence. His idea, known as the Turing Test, was to see if a computer could carry on text-based conversations very well. Turing argued that if a computer can “deceive” a judge, it should be considered intelligent.
For decades, Turing's tests have shaped the public understanding of AI. But as technology advances, many researchers have asked whether imitating human conversations really prove intelligence, or whether machines merely show that they can mimic certain human behavior. Large language models like ChatGpt can already hold persuasive conversations. But does that mean they understand what they're saying?
in Heart is important In a podcast interview, Dr. Georgios Mapourus told host Robert J. Marks that the answer is no. In a recent paper, General intelligence thresholds, Mappouras introduces what is called Turing Test 2.0. This updated approach sets a higher standard of intelligence than simply chatting like a human. Ask if the machine can produce beyond imitation New knowledge.
From information to knowledge
At the heart of Mappouras' proposal is the distinction between two types of information: non-functional and functionality.
- Non-functional information is raw data or observations that do not lead to new insights in itself. One example is to notice an apple falling off the tree.
- Feature information is knowledge that can be applied to achieve new things. When Isaac Newton tied the falling apple to the force of gravity, he turned normal observation into science.
Mappouras claims that true intelligence is the ability to translate functional information into functional knowledge. This creative leap allows humans to build skyscrapers, develop medicines and travel to the moon. A machine that simply rearranges words or retrieves facts cannot be said to have reached the same level.
General Intelligence Thresholds
Mappouras calls this standard General Intelligence Thresholds. His threshold sets a simple task. Given existing knowledge and raw information, can the system generate new insights that are not directly programmed into it?
This threshold does not require a constant glow display. Even an undeniable breakthrough (“Genius Flash”) would be enough to demonstrate that the machine has general intelligence. Just as a person is good in mathematics but not physics, machines will need to show creativity once to prove their potential.
Creativity and open issues
One way to apply the new test is through unsolved mathematical problems. Throughout history, breakthroughs such as Andrew Wills' evidence of Fermato's final theorem and Grigoli Perelman's solution to Poincaré's speculation have made the milestones of human creativity prominent. If AI can solve unresolved problems like the Riemann hypothesis and Collatz's guess, it would be strong evidence that the system has crossed the thresholds to true intelligence.
Large-scale linguistic models already solve equations and perform advanced computations, but solving unresolved problems from centuries ago shows something much deeper.
Beyond symbol manipulation
Mappouras is also depicted in the famous “Chinese Room” thought experiment of philosopher John Searle. In the scenario, people who do not understand Chinese are sitting in a room with a rulebook for manipulating Kanji. By following the instructions, the person produces output that convinces outsiders who understand the language, even though he is not.
According to Searle, this scenario shows that a computer can look intelligent without real understanding. Mapper agrees, but goes further. For him, true intelligence is proven by not only generating output, but acting on new knowledge. If the instructions in the Chinese room included a way out, the person could only succeed if he truly understood what the words mean. Similarly, AI needs to demonstrate that it can act meaningfully on information, not just symbols.
Image credits: Top image – Can AI pass the new test?
So far, Mappouras does not believe that modern AI has passed a general intelligence threshold. Systems like ChatGpt may look impressive, but their obvious creativity usually comes from patterns of trained large datasets. They do not show the ability to generate new independent knowledge that has been disconnected from previous inputs.
That said, Mappouras emphasizes that success does not require constant novelty. The true act of creativity – an undeniable demonstration of new knowledge – will be enough. Until that happens, he is cautious about the claim that AI today is truly intelligent.
Changes in discussion
The debate about artificial intelligence is changing. In the original Turing test, machines tricked us into asking, thinking we were human. Turing Test 2.0 asks more difficult questions: Can they discover something new?
Mappouras believes this is a true measure of intelligence. Intelligence is not imitation, it is innovation. It remains uncertain whether the machine will cross that line. But if that's the case, the world isn't just talking to computers. We learn from them.
Final Thoughts: Today's System, Tomorrow's Threshold
Models such as ChatGpt and Grok are prominent in conversations, summaries and problem solving within known domains, but their strengths still reflect pattern learning from vast amounts of training data. Mappouras standards only exceed the general intelligence threshold when creating a verifiable breakthrough. This is insight that cannot be traced to previous texts and human scaffolding, such as the original solution to major open problems. Until then, they remain powerful imitators and accelerators of human work. It is impressive, useful, transformative, but is not yet a creator of new knowledge.
Additional resources
Download podcast transcript
