How high can AI technology score on an IQ test?

Unlike the intelligence displayed by humans, artificial intelligence (AI) is demonstrated by machines. Where a person’s intelligence is natural, a machine must instead simulate cognitive functions by studying and learning from its environment. Artificial intelligence therefore has the ability to learn and solve problems and can demonstrate this by mimicking human behavior. So this then poses the question—how do we fully measure AI, and does it have the potential to outsmart a human? To answer that question, Washington University devised an IQ test that was designed to test how well AI technology could adapt to new concepts in unfamiliar environments.

Hand of AI technology touching human hand

The university’s IQ test measured the precision, accuracy and the time it took for a machine to perform a certain task. Much like a standardized IQ test for humans, the test-taker (in this case the machine) has to answer questions on pattern recognition, problem-solving, mathematics, and logical reasoning. Results from the study are still ongoing.

Example question from IQ test Example question from a standard IQ test. In this sequence, the test-taker (machine or human) must figure out which logical sequence comes next in the series.

A further study, conducted by scientists in China, looked at analyzing the intelligence of Google, Siri, and Bing in comparison to humans. This IQ test uses a standard intelligence model, meaning it is suitable for both humans and machines to take. The goal of the test-taker is to use the knowledge (or data) they already have, but to process it in a new and innovative way. In other words, gather all the details, understand/master them, and come up with a creative solution to the question at hand.

So how did these different AI technologies perform? Using a standardized IQ scale with a maximum 160 score, Google only managed to get 47.28. Anything below 80 for an IQ test is generally considered below the average, and the scientists deemed that Google had an artificial intelligence equivalent to that of a 6-year old child. With growing investment into AI technology it can only be assumed that these IQ scores will get higher over time—however they have a long way to go. Here’s how the other machines performed:

  • Baidu (Chinese-based technology company specializing in AI) scored 32.92
  • Sogou (Chinese-based technology company specializing in search engines) scored 32.25
  • Bing (Microsoft’s leading search engine) scored at 31.98
  • Siri (Apple’s popular virtual assistant) scored just 23.94

Think you can beat these machines? Take the IQ test today and prove that AI technology has a long way to go before it can conquer human intelligence.

How artificial intelligence can benefit us

The purpose of AI is to transform how we interact with technology in the day to day—whether that’s through creating on-demand chatbots, advanced search engines, or even emergency service assistants.

The research company Gartner, for example, is working to develop “emotion AI” technology that can detect a person’s mood and feelings. This emotion-sensing technology will be built into machine assistants, enabling them to recognize facial expressions, behavioral patterns, and even their user’s tone of voice. By equipping the machine with an added ability to interpret and respond to the emotions of its user, this development brings the role of an AI assistant closer to that of a human. Through increased interaction with this “emotion AI” technology, Gartner predicts that by 2022 these machines will know a lot more about our emotional states and may even develop their own form of personality as they learn more about us. While this may sound frightening to some, ground-breaking emotion-sensing technologies such as this could prove to be invaluable to healthcare providers, who rely on the reliable mental assessment of a patient before making an informed decision on their care.

Woman using artificial intelligence

Similarly to Gartner, Huawei, a leading provider of information and communication technology, is developing a personal assistant with a unique set of AI skills that would allow it to understand and react to the emotional needs of its users. The global tech giants claim their assistant will cater to a new generation of artificial intelligence through added emotional intelligence.

For Huawei, emotion-based artificial intelligence will also continue to play a key role in shaping the smartphone market over the next five to ten years. Huawei is committed to developing smartphones that “think contextually and engage humans in dialogue to understand their needs.” They believe that AI technology will evolve beyond the role of the assistant, as its unique learning capabilities will be able to provide users with personalized experiences catered specifically to their needs.

The role of the user in advancing artificial intelligence software

Many of you will already be familiar with virtual assistants such as Siri and Alexa. These mobile operating systems are designed to perform simple tasks that can enrich the user-experience by answering questions and providing information on services, entertainment, and general information (both local, global, and historical). But their level of artificial intelligence is limited and a lot of us are all-too familiar with the ChatBot experience—where you ask a relatively simple question but get something completely unrelated in response. At times they can be helpful, but it is often very frustrating and requires a human touch.

Let’s take Alexa as an example. She knows the answers to a million questions, but her answers are not necessarily always correct. That’s why the bright minds at Amazon decided to give their customers a chance to teach Alexa and help her to become smarter. The concept is quite simple—users who want to be a part of Alexa’s education can sign up to the Alexa Answers Program which allows users to suggest their own answers. When someone then asks Alexa a question, she will be able to give them a real “user answer”. The Amazon community will be able to vote for the best answers, which will help the algorithm to phase out inaccurate, irrelevant or offensive responses.