Return to site

Connecting the Minds of Human and Artificial Intelligence:

IBM's Project Debater looks to do just that!

 

· IBM,Ai,Debate,Machine Learning,Project Debater

Connecting the Minds of Human and Artificial Intelligence:

IBM's Project Debater looks to do just that!

Debate is the discussion of a topic between two or more opposing opinions. Being good at debate is quite difficult, requiring a combination of skills and knowledge. The best debaters must hone their public speaking abilities while also understanding both their own argument as well as their opponent’s. They must be creative in order to offer rebuttals against each opponent’s unique arguments. The necessity for passion, skills, knowledge, and creativity make debate one of the most complex and difficult practices to perfect.

Out comes IBM and their new Project Debater, the first AI that the company claims can debate humans on a number of complex topics, and do it well. IBM put the new machine to the test against seasoned human debaters at a San Francisco event on June 17. In two debates against Noa Ovadia and Dan Zafrir concerning subsidizing space exploration and telemedicine respectively, Project Debater won both. The machine was prepared in 100 topics before the debate, although it was not told of the specific topics that would be debated. Disregarding some minor errors such as repetition and diction, IBM’s Project Debater articulated its arguments remarkably well and offered sound rebuttals against the opponents’, all while even inputting some corny jokes.

The machine’s success at debating is obviously very impressive. As a machine, it is not tainted by human bias, but nonetheless is able to present arguments and rebuttals as well as an experienced debater. As IBM’s head of global AI research Aya Soffer states, "In general, computers are lagging significantly in understanding and being able to express themselves. If we expect AI to be useful, being able to communicate with people is critical." IBM’s new achievement surely does raise hopes that AI can one day seamlessly communicate with humans, which would only make artificial intelligence even more useful in our society.

So how exactly does Project Debater work? As stated on IBM’s website, Project Debater incorporates features including Argument Mining, Stance Classification, Deep Neural Nets, Natural Language Processing, Text-to-Speech, and data sets. IBM claims that its machine is the first to be able to detect claims and evidence in relevant texts or opponent’s speeches through machine learning techniques such as linguistic cues and rule-based approaches. In addition, the machine is trained with deep neural nets, or deep learning, in order to improve its speaking and listening skills. Deep learning is meant to simulate the human mind, further demonstrating scientists' progress in linking artificial and human intelligence. For even simple human abilities like understanding each other’s informal speech, the machine must use its deep learning to put its opponent’s words into punctuated sentences. To construct their deep learning system, IBM had to produce appropriate and clear data sets, or parameters, for the machine to learn from. These data sets are available for download on IBM’s site.

As impressive as Project Debater is, are we that surprised? After AI’s rapid development over the past 20 years, defeating humans in the most complicated board games and intellectual game shows, many scientists have accepted artificial intelligence’s limitless potential in the near future. It’s interesting what system IBM and other tech companies will construct next. Can AI grow out of its rule-based structure to best humans in other ambiguous, intricate tasks? Artificial intelligence will continue to expand its abilities so that humans’ unique and gifted minds won’t be so special anymore. The infinite complexities in our life and universe may one day be mastered by artificial intelligence- a truly intriguing and terrifying thought.

Written by Michael Ding & Edited by Alexander Fleiss