IBM Project Debater — Transforming Quality and Speed of Human Labor

IBM Project Debater crushed the Turing Test earlier this week and left me with the impression that there is a lot of life beyond the marketing hype in IBM’s Artificial Intelligence program and in artificial intelligence itself. Hosted at San Francisco’s Yerba Buena Center for the Arts, IBM invited a number of press, analysts, and other industry experts to witness an incredible debate between AI and a top human debater. I do not use the term “incredible” lightly, but the degree of understanding of the human language of IBM’s new Project Debater AI (I think its branding will need some improvement) was nothing short of stunning. This deep language understanding enabled the AI to dig deep into research statistics, easily dwarfing the ability of any human trying to surface and qualify the best scientific studies on a topic within only 15 minutes. These studies were dead-on, containing data points that exactly addressed the individual issues within the debate.

Harish Natarajan, the human debate champion, of course showed a very high level of structure in his thoughts but did by no means outperform the AI, despite ultimately winning the competition. However, this debate was not about winning, but about demonstrating how AI can enrich and enhance human thought processes. Using the technology behind Project Debater as your assistant who constantly listens to your business arguments and customer communication, you would permanently be able to delight your clients with fresh ideas, more choices, no more missed nuances, and a generally balanced viewpoint on most issues.

I basically described this in a blog post a few years back and the principle has not even changed since I first played with latent semantic analysis (LSA) and natural language processing (NLP) in 2000.

Originally published at cloudadvisors.net on February 15, 2019.