IBM has long loved pitting humans against machines, and it was beginning to look a little one-sided, in AI’s favor, until Harish Natarajan beat Miss Debator (an AI) at the company’s Think Conference in San Francisco on Monday. [1]
Let’s Have a Debate
Thousands of hours of training for debate paid off for the champion debater, Harish Natarajan when he went up against IBM’s AI Miss Debater and won. Harish Natarajan is a European Debate winner and well versed in the components of debate. As he stated in a LinkedIn post, debating has three components “First, a debater needs to process large amounts of information and construct relevant arguments. Second, debating involves [explaining] complicated arguments in a clear and structured way. Third, you need to make those arguments matter to the audience. This requires the careful use of language, emotions, rhetoric, and examples. While a machine should excel in the first, the latter two may be challenging.” [2]
IBM’s black monolith AI and Natarajan exchanged quick retorts on pre-school subsidies, a topic picked out beforehand, for 25 minutes before the audience of 700 people hailed Natarajan the victor.
Grand Challenge
Both Natarajan and Miss Debater were given 15 minutes to prepare. Then they each gave a four-minute opening statement, a four-minute rebuttal, followed by a two-minute summary. The audience was filled with top debaters along with over a hundred journalists.
Miss Debater got her information from its database that’s filled with over 10 billion sentences from newspapers and academic journals. The event was the latest in IBM’s “Grand Challenge” series. The grand challenge series is focused on pitting humans against machines.
In the past, IBM’s computers have beaten beat chess grandmaster Garry Kasparov and two record-winning Jeopardy! contestants. Their computer lost to Natarajan in the debate, however, and Natarajan suggested that it was because “Debating is…more complicated for a machine than any of those [Go or video games], however.”
Natarajan did go on to say that it won’t be long before AI learns emotions, rhetoric, and examples, which will make them better at a debate.
Notes:
- ^“IBM AI fails to beat human debating champion.” Engadget, 12 Feb. 2019, www.engadget.com/2019/02/12/ibm-ai-human-debate-fail. (go back ↩)
- ^“IBM’s Project Debater.” 12 Feb. 2019, www.linkedin.com/pulse/ibms-project-debater-harish-natarajan?articleId=6500512818150805504&trk=aff_src.aff-lilpar_c.partners_pkw.10078_net.mediapartner_plc.Skimbit%20Ltd._pcrid.449670_learning&veh=aff_src.aff-lilpar_c.partners_pkw.10078_net.mediapartner_plc.Skimbit%20Ltd._pcrid.449670_learning&irgwc=1#comments-6500512818150805504&trk=prof-post. (go back ↩)
