Although Robert Mercer and Peter Brown of the IBM Yorktown Heights statistics-based research group were the stars of the conference, there is no clear sign that IBM is about to turn its research advantage into a concerted effort to develop an MT system soon. The Mercer group apparently pursues its research interests without an imposed development schedule and without fixed links with other MT developers and translations researchers.
    This uncommitted attitude was demonstrated when CMU’s Sergei Nirenburg asked Mercer how he intended to solve the problem of the fifteen percent error rate of their statistics-based system. Mercer replied, I don’t know whether MT will be a success or not.

(See  article  that this sidebar accompanied)