It’s official – Siri is a simpleton
It is true, but might be a bit mean to single out the iLeader, as researchers have put various artificial intelligence (AI) programmes to the test, figuring out what the IQ of each one is.
Telecoms.com (Banking Technology’s sister publication) reports that Liu Feng of Beijing Jiaotong University, Yong Shi from the Chinese Academy of Sciences Research Center and Ying Liu of from University of Chinese Academy of Sciences have figured out what we’ve all been wondering; how smart are AI virtual assistants. And it turns out those who are worried about the technology rising up and enslaving us can sleep easy for the moment; the average six-year-old has a higher IQ.
“This research provides a possibility of using the AI IQ test method to continually assess relevant intelligence systems and to analyse the development of the artificial IQ of various systems, allowing for the differentiation of similar products in the field of artificial intelligence,” the researchers says in the paper. “The resulting test data will have practical value in researching competitors’ development trends.”
In terms of the best and worst, Google featured at the top of the AI list, with an IQ of 47.28. Baidu’s Duer got 37.2, Chinese search engine Sogou hit the heights of 32.25, while the AI powering Microsoft’s Bing had an IQ of 31.98. Apple’s Siri was a lowly 23.94. And just to provide a bit of context, a six year-old child average IQ is 55.5, at 12 the average increases to 84.5 and at 18 it is 97.
To be completely honest, the rankings don’t really mean much at the moment, though it is a much needed reminder the AI world is still in its infancy. The benefits have been loudly pronounced by the technology companies of the world, but realistically, there is still a lot to do before we see AI make more than quirky and slightly inaccurate impacts on our daily lives.
“IQ essentially is a measurement of the ability and efficiency of intelligent systems in terms of knowledge mastery, learning, use, and creation. Therefore, IQ can be represented by different knowledge grades,” the researchers says.
The idea of introducing knowledge grades in the AI world is an interesting one, as it effectively gives an instant answer to how advanced such a programme is. The researchers have suggested this could be a scale from one to six, with each level demonstrating an increased level of independence, interactivity and intelligence.
This scale starts as simply as whether an AI programme can interact with a human, and rises all the way up to a system which “continuously innovates and creates new knowledge, with I/O ability, knowledge mastery, and application ability that all approach infinite values as time goes on”. This would be what most people would see as AI.
Aside from the age of AI, another explanation for the low IQ scores might be the limited nature of most AI programmes. AI propositions are currently being built for specific purposes, with little of zero opportunity of that particular programme broadening its talents. AlphaGo, for instance, demonstrates a high skill level for the complicated game of Go, but would it be able to beat a six year-old at Hangman? The researchers believe AlphaGo would be at the third knowledge grade.
Some people might be scared of the potential of AI, but for the moment it seems you can just threaten it with no dessert if it starts getting a bit moody.