Although humans have not yet achieved this superpower (neural quantum can “entangle” anyone?), information posted on social media reveals the living conditions of about two billion people, or a quarter of the world’s total population A personal information may be used for analysis. What we lack is an efficient way to analyze this information and make it work.
In many applications, software has been used as an aid, especially for expressing personal opinions. Tools exist today to quickly establish conversations and social posts between consumers and businesses. With these tools, through open, authentic conversations, businesses can engage more users, understand their needs, and keep an eye on them.
However, there is still a lot of data analysis work to be done in the middle, especially data aggregation. The opinions summarised by social networks and software tools are largely drawn from the most influential viewpoints in each industry. We don’t do analysis. The prior art collects people’s opinions without exploring the reasons behind the opinions. The reason is that it is difficult to sort out the “why” because the reasons behind are often not obvious, but require complex reasoning or bold assumptions to find.
We can effectively parse word and phrase trends, but not deep sentence understanding. Capturing only current hot topics is far from pinpointing how a particular group of people feels about said topic, and the reasons behind their opinions. Conversations, and the people who carry them out, are three-dimensional and convey much more than words themselves.
The most efficient algorithms and the best practitioners fail to find the subtle connections between conversational people, topics, and causality. In the recent US presidential election, for example, pollsters, academics and analysts have not predicted who will win, and they are busy explaining why. Could the answer be hidden in billions of social posts?
What if we could use artificial intelligence to investigate and draw conclusions? Imagine having a command-driven AI product around you. For example, digital research assistants read and comprehend thousands of posts per second and then summarize key information.
Is artificial intelligence up to the task? Not yet
Current AI systems can grab headlines, but current AI systems are limited in scope and can only be used for tasks that seem laborious. Uber and other companies are working to train cars to sense their surroundings and make autonomous decisions to turn or avoid pedestrians, but cars never know how to grow wings and fly. Google has developed an artificially intelligent robot that beats humans at a highly complex game, but also cannot answer historical questions about the game or learn how to play other games on its own.
Today, artificial intelligence remains a misnomer. In the Oxford English Dictionary (Google), intelligence is defined as “the ability to acquire and apply knowledge and skills”, and AI currently falls short of this standard. In the words of Tom Davenpor, a researcher at MIT’s Digital Economy program, a thought leader in the field, “Deep learning is not deep learning.” Or, to quote another expert, Oren Etzioni (Allen Institute for Artificial Intelligence Institute for AI), “artificial intelligence, that is, performing simple mathematical operations on a large scale [of data].” Today’s artificial intelligence enhances the capabilities of computers, but it does not meet the standards of artificial intelligence that humans expect .
In the future, however, will AI meet this standard?
Fortunately, AI research is moving in the direction of “deep understanding.” Historically, artificial intelligence has primarily used the Turing test to assess a machine’s ability to think like a human. Now, researchers are being challenged even more, such as requiring the AI to pass the Winograd Schema Challenge.
The Winograd Model sheds light on the current state of artificial intelligence and proves that we may not be as far from true artificial intelligence as it seems to laymen. An example of this at this year’s O’Reilly AI Conference is as follows.
• The big ball hits the table directly because it is made of Styrofoam.
• The big ball hits the table directly because it is made of steel.
In the above two sentences, what does “it” refer to? Seven-year-olds know the answer. But for machines, finding the answer is difficult. Therefore, it is also quite difficult for the machine to find the answer from each Weibo.
When artificial intelligence learns and understands with similar human intelligence, and when artificial intelligence can effectively process massive amounts of complex data beyond the limits of human brains, artificial intelligence has truly reached the level of general intelligence.
Tests like the Winograd model could advance AI’s understanding of impact and connections. However, simple basic language understanding is only an early stop on the road to full-blown artificial intelligence, and there is still a long way to go before it reaches the point where it can autonomously acquire and apply information.
Our benchmark is only a slight improvement over the Turing test, and there is still a lot of work to be done.
GIPHY App Key not set. Please check settings