What's more indicative of reality is the relatively constant vertical distribution, in a log scale, across any time range up to about 90 years, when the lack of data starts to play a significant role. So, in relative terms, uncertainty remains fairly constant over time. In absolute terms it actually increases. In a sense risk still does decrease over long time horizons, as the chance of earning over any given amount does increase with time; contrary to common belief though, the distribution of possible returns (another possible definition of risk) actually increases with time - or remains roughly constant on a log scale, as mentioned.
(You can find a great many opinions surrounding this by googling "time diversification".)