IBM showcased several of its artificial intelligence (AI) technologies at its recent Analyst Insights day in London. Two examples of IBM’s technology stood out: IBM Watson Debater and PowerAI. Project Debater is still a research project, but its progress was recently showcased in a couple of debates in San Francisco and Tel Aviv that were staged to demonstrate how well the AI system can debate with an experienced human debater. Debater shows impressive abilities in the field of argument mining, an offshoot of natural language processing that focuses on extracting information from the subtleties of complex argument constructs. PowerAI, meanwhile, is an appliance AI solution incorporating IBM’s latest Power CPU with Nvidia’s latest GPU, connected with the fastest connector, NVlink2, and a deep learning software stack. The showcased customer case study proved how deep learning made easy for data scientists can make impressive discoveries.
The Debater technology took part in demonstration debates in which it was asked to persuade an audience to its view on a position in a typical professional debate setting. The overall scores, assessed by the degree by which audience members had changed their mind from before to after the debate, were close, but audience members overwhelmingly said that they learned more new information from the Debater than from the human. Debater is being trained in several domain topics, so it is not yet able to compete with a human in a completely open debate. However, within the acceptable domains, it can be given any side of the debate, for or against. This makes the technology useful beyond debating. It can serve as an assistant to someone entering a debate or negotiation, say a politician due to speak in parliament, or a CEO meeting with a business party. In this assistant role, Debater can provide arguments on both sides of a question or assertion and can enable humans to be better prepared.
IBM also showcased a customer using its PowerAI appliance. The neurologist from University College London Hospitals used PowerAI to analyze big data with deep learning algorithms. The data related to drug trial results, mining data for subsets where the trial was successful, and then analyzing the common factors within the success groups. There is a large amount of waste in new drug research because final human trials show that on average the results are poor (worse or the same as a placebo) and the drug is often shelved. What the big data and deep learning analysis showed was that by looking at the subsets of people with certain common characteristics, it is possible to find that discarded drugs can be beneficial, and the analysis of the favorable characteristics becomes important information that can save lives. The case study gave a glimpse of the future of medicine making good use of AI technology.
Michael Azoff, Principal Analyst, Information Management