Of course things have always been changing, but the current rate of extinction hasn't been equalled since the four or five largest mass extinction events. It's not an ordinary situation in life's history. Ultimately it depends what you care about but it's inarguable that the natural world is being greatly impoverished in ways that could take millions of years to recover from (e.g. in terms of species numbers)
I think one nuance that empasizes his point is that within millions of years, and likely much sooner (as we are already rather 'overdue'), there will almost certainly be another genuine mass extinction event that will once again strip most life from this planet.
The notion of rates is also quite a bit more contentious than many appreciate. The extremely high rates in modern times are based on modeling with rather debated assumptions. This is how you get claims of hundreds of extinctions per day even though there have only been about 800 'verified' extinctions in the past 400 years [1]. And, notably, a number of those verified extinct species ended up showing up elsewhere later on!
This is exactly what confuses me -- we're modeling both the inputs and outputs, and not with particularly robust validation or verification.
Where is the utility?
We've established that things are changing. Sure. But if you gave someone a 'stock predictor' of similar quality, few people would choose to trust it with money. But in reality, we dedicate billions of dollars of resources collectively to the outputs of biome modeling.
