Have you ever thought of a question and google couldn’t find the perfect answer for you? In the near future, we could expect search engines to give an answer straight away rather than giving you possible links. Companies like Alibaba and Microsoft are racing to create a powerful AI, one that could do things that only thought possible for a human being to do, thinking, comprehending and actively responding. In this article we will take a look at how it might impact our society.
Alibaba and Microsoft have built artificial intelligence that have beaten humans on a Stanford University reading comprehension test. Alibaba later stated, “This is the first time that a machine has outperformed humans on such a test.” Artificial intelligence experts at Stanford created the test to calculate computers’ reading capabilities, which are constantly growing. Alibaba’s software was the first to beat the human score of the test. Although many considered this accomplishment as a great milestone, Luo Si, the chief scientist of natural language processing at the Chinese company’s AI research group acknowledged admitted this advance in artificial intelligence will cause many to lose their jobs. The technology “can be gradually applied to numerous applications such as customer service, museum tutorials and online responses to medical inquiries from patients, decreasing the need for human input in an unprecedented way,” Si also said.
I think we first have to ask, “ How are tests designed?” What does this test measure? The Stanford test consists of questions about a set of Wikipedia articles. For example, a human or AI program reads a passage and then answers multiple questions regarding the passage. Alibaba’s network program scored 82.44 on the test on January 11, beating the 82.304 scored by the human participants by a hair. A day later, Microsoft’s AI software also beat the human score, with a result of 82.650.
I felt that the AI programs weren’t really fully comprehending what it was reading. On a basic level, there are two parts to comprehension. The first part, taking in information and being able to categories them in patterns and links. The second part, being able to use those patterns and links to answer questions that aren’t directly related what is already known. For example I give you two pieces of information, water has a large amounts of H2O molecules and H2O reflects blue light. Then, I ask you what is the color of the water? If you can only do part one of comprehension, there is no direct correlation to what I asked and what you know, so you would not be able to answer that question. If you were able to do the second part of comprehension, you will be able make a assumption that since water is mostly H2O molecules, the color of water must be blue. It is kind of like learning something is class and being able to apply that knowledge on homework that isn’t exactly like the material learned in class.
I don’t think we are at a time where, AI will start replacing humans quite yet. The AI systems mentioned in the article seem to be not capable of the second part of the comprehension. They lack the human qualities that is required for the jobs that AI was thought to replace. Furthermore, I would add that the prediction that some job will be lost to AI is a gross underestimate. As soon as AI can master the second part of comprehension, they will replace everyone’s jobs, even the researcher’s jobs. The AI will be able to make assumptions, test ideas and learn. With its superior processing power, it will only be a short time before it surpasses all the information that humans have gathered over 1000 years.
I think we will come to a point when we have to decide on how much AI we want and how much we should develop it. It is going to become an ethical question. After AI has surpassed us, what does that mean for humans? What is our purpose? Should we merger with AI’s?
Pham, Sherisse. “Reading Robots Beat Humans in Stanford Test.” CNNMoney, Cable News Network, 16 Jan. 2018, 4:16 AM ET, money.cnn.com/2018/01/15/technology/reading-robot-alibaba-microsoft-stanford/index.html.
Picture reference: