Longer comment:
The paper considers each parameter of the Drake as a probability distribution over some range. Notably, they assume that f_l in the Drake equation (the fraction of habitable planets that actually develop life) may vary over (at least) 200 orders of magnitude. Yes, you read that correctly.
Naturally, they thereby get a long tail of "planets essentially never develop life, even despite there being so many of them".
My main critique (as someone with no further credentials) is that the prior probability distributions they assume for the parameter distributions are more or less meaningless. Their results say "if we assume that the log of f_i (life-bearing planets that develop intelligent life) is uniformly distributed between log(0.001) and log(1), similar for other parameters, then the probability of being alone in the milky way is > 50%".
Specifically, by far the largest impact here is from assuming some distribution for f_l that covers 10x as many orders of magnitude (!) as the remaining parameters together. I'm very certain (but haven't done the math) that slightly tweaking the sigma they assume for the corresponding log-normal distribution (that sigma is 50 orders of magnitude) would drastically impact the results. They do a lot of dancing around that fact (discussing how little impact the priors of other parameters, or even the center of the log-normal distribution for f_l have), but fundamentally the result boils down to:
"We're extremely unsure how often abiogenesis happens, and this parameter dominates everything else." Followed by some math that (oversimplified) boils down to "if we set the rate of abiogenesis to effectively zero half the time, then half the time we are alone in the universe."
PS: Title should probably have (2018).