Indeed the error range here is about 1.5 to 2 billion years. So from the paper we know that life began its growth in complexity about 6-9 billion years ago. This is strictly from numerical arguments. Evolution is just a random choice in the number of husbands per ceremony.
There are two fundamental issues with evolutionary "clock" models getting extrapolated too far backwards:
1. The rate of change-per-generation is very much not constant, as described by the Punctuated Equilibria theory. Sudden changes in the environment can cause sudden bursts of evolution. We don't know if there were any (and how many) mass extinctions / sudden change events before the fossil record starts, which is already hundreds of millions of years into the existence of single-celled life existing!
2. The time elapsed per generation has changed over time too, and we have virtually no direct evidence of the actual rate for the earliest epochs of life, before multi-cellular life.
These are particularly bad problems for any theory trying to extrapolate backwards, with compounding issues that can blow out any naive error estimates massively.
For example:
RNA-based vs DNA-based life. We know that DNA is more stable and resistant to mutation than RNA, which was the foundation of the earliest life forms. But we have no idea how that difference specifically affected early life evolutionary rates! We can guess... but only guess. However, almost certainly, early life had a much higher mutation rate per generation than modern life AND a consistently short generation time.
No evolutionary clock models used, instead a numerical model was used.
1) The rate of change not being constant is not significant as long as its variation is normally distributed about some mean. I expect that changes in environment are randomly distributed.
2) This is too small scale to have any impact on the trajectory of the numerical growth in number of base pairs.
We don't need to guess, life isn't special. The growth of complexity under favorable conditions is observable and occurs all over the place at the same rate, at scale.
Intelligent life on earth is DNA/RNA based, not one or the other. If you take the mean rate your interpolation from now back the origin of such life with that particular mean rate of numerical growth in base pairs holds with error margins.
