Though it is not economic yet to extract uranium from seawater, it can certainly be done, and the estimate you quote, namely that there are 12-15 million tonnes of uranium there, is too low by orders of magnitude. The figure is in fact over 4 billion tonnes. (The concentration is about 3.3 milligrams per cubic meter and the volume of the oceans is about 1.3 billion cubic kilometers.)

However, it is unlikely that we shall ever have to rely on that. On land, elements are not evenly distributed but occur in ores. Ores can be discovered – but that takes effort and money. As a result, the known reserves of valuable ores tend towards a certain multiple of the current rate of use. The exact multiple depends on all sorts of economic and political factors but for obvious reasons it would not be surprising if it were a few decades. Furthermore, controlled nuclear fusion is likely to become economic within a few decades, which makes it unrealistic to imagine the world having to rely on uranium for the indefinite future.

Therefore, it is only "safest to work with more conservative estimates" if one wants a reliable way of raising resource-depletion fears regardless of the truth of the matter. If one wants to plan rationally for the future, it is useless. Assuming that today's 'known reserves' are the only resources that will ever be used has the same logic as estimating that one will starve when the current contents of one's larder have been eaten.