In my book on Genetics, I briefly explained why CRISPR couldn't do the job we needed of husbanding our genes. This is a more expanded explanation. There is a difficult part that I originally solved intuitively but I figured out a fairly easy way to explain that part better. Basically the reason Artificial Selection is a better method for husbanding genes is that it is cheaper, safer and will do the job. It might also be more ethical in some cases.
First off the biggest and nearest problem is that we have to replace the natural selection effects that have been removed as humans have progressed. Numerically that especially means the effect of medicine such as vaccines. Other factors matter as well. Humans don't really have predators anymore. Better nutrition has greatly reduced childhood mortality. Smaller families mean less opportunity for natural selection to act though that is a little dubious in ways. The point is that the reduction in natural selection is the main problem compounded by increased mutations due to older parents. Artificial selection will be a very similar process to natural selection with a similar outcome. Using it will allow more "selection" than is needed to avoid genetic load. plus increase the frequency of "superior" genes - for health, beauty, and brains. CRISPR is a totally different process than natural selection. It is a change in the genetic sequence. It is closer to mutations and most mutations are not just failures, they are mostly lethal or nasty.
The usual argument I have used is that artificial selection will be cheaper than CRISPR because CRISPR would have to follow the exact same process as artificial selection plus more expensive steps. Both processes would have to start with a genetic analysis of multiple zygotes. You would have to know what genes you are working with at the start. With artificial selection, the next step is implantation. With CRISPR the next step is to make a CRISPR molecule that will do the genetic editing. Implement that process and then repeat the genetic analysis to select a zygote in which is worked correctly and then do implantation.
At this point, the big unknown is about CRISPR. No one knows how difficult the problem of "untargeted effects" will be. That is how often is the CRISPR process going to make unintended genetic changes besides the targeted change. CRISPR is like magic, but it is also like organic chemistry and in organic chemistry there generally are not "clean reactions". CRISPR is extremely complicated organic chemistry. Consider three scenarios. (1) CRISPR is developed to work perfectly with no untargeted effects. That's great and the potential is amazing. We still need this quickly though. In a couple of generations, we are going to see some nasty effects of genetic load if we aren't already seeing them. The analytic process to develop a knowledge of artificial selection will be the same one needed to use CRISPR. We should get started ASAP. (2) CRISPR regularly leads to a small number of untargeted effects. Each one of those represents an artificial de novo mutation. That is a problem. Still, it might be very useful if used to make a high-value change or two. That does have great potential. It appears that small genetic changes can make valuable changes. It raises an ethical question because you might have a person with an improved gene but also a new broken gene. Presumably, that could be removed in the next generation or two. Like my original description though, this would not compensate for the reduction in natural selection so artificial selection would still be required. The process would be more expensive than artificial selection because it might have to be done on numerically more zygotes to try to get one that was viable. Of course, artificial selection could be done at the same time. It is a numbers, technique, and cost equation that currently has unknowns. Then you need to ask the cost-benefit, particularly the benefit beyond what simple artificial selection could do. (3) CRISPR regularly leads to a large number of untargeted effects. Again, then the process is creating random de novo mutations. It might be impossible to safely use. This would not surprise me but no one knows right now... Well, some smart molecular geneticist might actually be able to make a very good educated guess.
In that discussion of possibilities, there is always the unknown "number" of untargeted effects - semi-random artificial mutations. So how many are too many? It's probably a binary progression. If there is one broken gene for every edited gene, then it might be a fairly safe process to use. If there are four genes broken for each one edited then you are getting to a problem area. Also, does it always cause four errors or would the same process rarely but sometimes create eight broken genes and sometimes it doesn't cause any. Then artificial selection could fix up that problem just the way it fixes up natural mutations. They are just excluded.
The point is that there is more than one unknown that would determine the usefulness of CRISPR. I'm biased. I've said that like survival, I am conservative. My bias is towards artificial selection because it will do what I am trying to accomplish, that is to allow humans to survive as they transition to a new ecology where they can survive and develop long term. Artificial selection will do what is needed. So the question comes up of what do you want to accomplish with CRISPR? It looks like there are practical potentials. In terms of using it on non-human species, CRISPR is going to make some biotech companies truckloads of money... and maybe a trainload or two for some. Using it on humans is a completely different moral question. Not only could it cause problems, it could cause hidden problems. Your subjects might just become victims. It's not going to become ready for prime time for generations. We need to compensate for the reduction in natural selection far sooner than that and the technology is not a maybe, it's being done and it is what nature does. CRISPR might be useable in terms of very small numbers so that the failures are very small numbers, but then you can't possibly hit the numbers you need in order to replace natural selection. The failures may be unavoidable and significant and ugly. That is all why in terms of survival I describe the path I do of using artificial selection. Even if CRISPR didn't have real potential dangers it would be much more difficult and expensive. Natural selection effects everyone and we're going to need this available for everyone. Artificial selection can do the job and fairly cheaply, not just more safely.