Sample randomly within cutoff in tibble R

I have a tibble with 100 points in R, below:

preds <- tibble(x=1:100, y=seq(from=0.01,to=1,by=0.01))

And I want to randomly sample 20 observations with values less than 0.5. Currently, I can select the first 20 observations by:

number_of_likely_negatives<-20

likely_negatives <- preds %>% 
    arrange(y) %>% 
    slice(1:number_of_likely_negatives)

But how can I randomly select 20 observations with y values below 0.5?