For a few decades now, scientists have been interested in Artificial Intelligence. Specifically in the 21st century, artificial intelligence (AI) is said to have progressed by leaps and bounds. But even with all this advancement in technology and increasing research being done in the field, is artificial intelligence even close to human intelligence?
To learn more about the current level of intelligence achieved by AI, researchers at MIT developed an AI system known as the ConceptNet4. The ConceptNet4 was given a standard IQ test in order to find out how “intelligent” it really was. It turned out that it has the mental capacity no more than a 4 year old child.
So what makes computers with AI more special than the traditional computers? The answer is simple. ConceptNet has the ability to, besides performing the functions of a traditional computer, find relations between different things. For example, any computer would be able to tell you that Shahid Afridi is a famous Pakistani cricket player because it would have this information saved in its memory. But ConceptNet would also find the relationship between him and the rest of the cricket team. Therefore AI cannot only show you the factual information it has, but can also find and show you the relationships between those facts.
To have an idea about the AI’s cognitive abilities the ConceptNet was given a popular test also given to young children, the Primary Scale of Intelligence Test. The ConceptNet behaved like children in the test too, i.e. it scored well in some categories and in others it didn’t score that well. As said before, it showed good results in identifying the similarities between different things and showed a good command over the English language vocabulary.
However, the AI seems to have little common sense as it performed poorly on comprehension-based questions. A simple example can be that although it can easily tell the exact boiling and freezing points of water, it cannot differentiate between hot and cold. Similarly AI systems cannot answer questions that involve human experience and intuition. Examples of such questions that AI systems would be unable to answer can be “Why does a mother love her child” or “Why do humans shake hands when they meet?”
Researchers until this time have only been able to feed hardcore facts to machines and design them to the extent that they can make connections among the information they have. But they are still to find out ways using which they can perhaps include common sense in AI systems. It is quite easy to ask a computer to recall the facts that have been feeded to it, but we haven’t been able to make it draw any sort of conclusion using the knowledge it has.
If in the future researchers can come up with a solution to this problem, then perhaps we will really be able to see AI systems worthy of being called “intelligent.” Until then they’re no more intelligent than a 4 year old.
Related Posts
