Google AI vs Siri vs Bing: IQ tests show one is smartest by a mile

Researchers who’ve developed an IQ test for artificial intelligence report that Google considerably outsmarts Siri and Bing — but with an IQ that’s less than a six-year-old’s.

According to researchers Feng Liu, Yong Shi, and Ying Liu, Google’s AI has an IQ of 47.28 based on tests carried out throughout 2016.

Google’s score is slightly below a six-year-old human’s IQ of 55.5, but more than double Siri’s IQ of 23.9. By their measure, Siri is not as smart as Microsoft’s Bing or Baidu, which have respective IQs of 31.98 and 32.92. All AI’s IQs are significantly less than an average 18-year-old’s 97 IQ.

An IQ evaluation of several search engines in 2014 by two of the three researchers found Google had an IQ of 26.5 compared with Baidu’s 23.5.

“The results so far indicate that the artificial-intelligence systems produced by Google, Baidu, and others have significantly improved over the past two years but still have certain gaps as compared with even a six-year-old child,” they write in a new paper published on ArXiv.

The researchers also propose a model for grading the intelligence of AI systems and assessing technology in the context of the threat it poses to humans. Their model unifies AI and human characteristics around four areas of knowledge, including “input, output, mastery, and creation”.

The AI intelligence model ranges from a first-grade system, which can’t exchange information with humans, to an almighty God-like sixth-grade system feared by Elon Musk, 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”.

First-grade AI systems have the same capacity for information-related interaction with humans as stones, while second-grade systems include smart TVs and washing machines that can control program information but lack the ability to learn automatically.

The third class include computers and mobile phones, which can be programmed and upgraded, while fourth-grade systems include Google Brain, Baidu Brain, and the EU’s RoboEarth robots, due to the ability to adapt based on information exchanged in the cloud.

Humans, they argue, can be regarded as possessing “special” naturally occurring AI that conforms to fifth-grade intelligence due to their capacity for creativity, which all systems in lower grades lack.

In this scheme, the researchers argue that AlphaGo, from Google-parent Alphabet’s DeepMind, would be a third-grade system, in part because it’s still in the R&D phase. Also, despite beating humans at Go with exotic strategies, it can’t be considered as having creativity since its still relies on humans for training.

“If humans did not provide help to the program, and AlphaGo could obtain Go chess data on its own initiative, self-program, and simulate game contests to gain experience for changing its training model to win games in real contests, it might be more defensible to say that AlphaGo could innovate,” they write.

“However, as AlphaGo does not appear capable of such a development process, from a comprehensive point of view its intelligence rating is of the third grade, which is two grades lower than that of humans.”