Abstract. We express some criticism about the definition of an algorithmic sufficient statistic and, in particular, of an algorithmic minimal sufficient statistic. We propose another definition, which might have better properties.
Nikolay K. Vereshchagin