The event on that day was a special debate, one that has never existed in human history.
The debate was organized by the tech company, IBM. It was exceptional as the debate consisted of a human on one side and AI on the other. Simply speaking, a human is reasoning with a “robot”.
The AI project on the other side of the debate has a rather straightforward name of Project Debater. It is an offline computer system. This means that it can’t acquire real-time information from the internet to rebut. It works by listening to the opponent’s speech and argue point by point with only the information available inside the machine. From this perspective, this is “fair” play as both sides must draw from previously accumulated data to form the most efficient cause and effect argument. However, some find it “unfair” in that the AI is equipped with enormous memory space. Yet we must keep in mind that AI isn’t as skillful in forming sentences or logic thinking when compared to a human.
Both debate topics were initiated by IBM (with 2 human debating for each topic). The first was whether “we should subsidize space exploration”. Here, Project Debater argue for while Nao Ovadia, national debating champ from Israel, is the opposition. The second topic was whether “we should increase the use of telemedicine”.
The highlight of this debate is that both parties – AI and human – didn’t know the topic beforehand. Thus, they weren’t able to prepare information in advance but had to draw from the knowledge in their brain or system for the argument.
In the first topic, Nao Ovadia argued that the government budget should be spent on more direct needs such as scientific research on earth rather than for space explorations. After listening to Nao Ovadia, Project Debater rebut by saying that “It is very easy to say that there are more important things to spend money on, and I do not dispute this. No one is claiming that this (space exploration) is the only item on our expense list. But that is beside the point. As subsidizing space exploration would clearly benefit society, I maintain that this is something the government should pursue.”
After both debates, spectators could vote for their preferred debater. The results were explicit that for delivery, human did better while for substance, AI won.
How teaching AI to debate will benefit human?
This project was intended for AI to thoroughly understand the use of language so that it can learn to “understandably” communicate better with a human when compared to previous virtual assistants. The challenge for Project Debater is attempting to “understand” tremendous data to form conversations that are logical and persuasive. Imagine a diagnosis system that works with doctors. It may be able to perform an imaging test and diagnose to deliver logical options to help with decision making similar to that of human assistants. An example closer to consumers may be if the system is developed for an online shopping platform. It may help us select clothes designs and patterns along with reasons why it fits us or how it fits with the current fashion trend. Another example is decision making in the business sector wherein AI can explicitly analyze the pros and cons of each option. It can simulate and become the “proposition” and “opposition” party to form an internal debate.
It may become a tool that helps us be better at systematic thinking.
Nevertheless, the biggest challenge for system development is the choice of data used as the base material. If the information is full of bias and distorted numbers, the argument by AI will surely be distorted as well.