He means the most common definition of truth is 'The best predictive model' (in the scope of the problem). I would add the principle of parsimony, truth is 'The

He means the most common definition of truth is 'The best predictive model' (in the scope of the problem). I would add the principle of parsimony, truth is 'The simplest available formulation of the best predictive model'. In both cases you cannot know that no better predictive models exists (and hence if your truth is absolute), even if you experimental error is zero (which is essentially impossible in most sciences unless you grossly overfit), since there may be unobserved phenomena as well.