It's easy to detect a vast amount of child porn, simply because the vast majority of child porn sharing is just a few thousand different files, shared over and over again. A few hundred thousand files if you count the non-nude and nude "child modelling" stuff, too.
You can catch a majority of such content just by running a dumb hash over the bits, even though the detection rate suffered a little thanks to dumb smartphone users who will screenshot everything instead of sharing the files, thus creating "new" content.
If you add some simple "content hashing", like https://pypi.org/project/dhash/ or Microsoft PhotoDNA https://www.microsoft.com/en-us/photodna you can catch an astounding number of content shares, something which really looks great in press releases. My guess would be that the vast, vast, vast majority of those 18M reports that letter mentions came from automated engines detecting the "common" content with dumb bit or content hashes. But you really didn't do much except annoy and scare some online pedophiles* who shared the same 10, 20, 40 year old content* and maybe even put a few of them away for good or caused them to kill themselves. And yet, you almost never actually prevented ongoing child rape and other abuses.
The problem however is that you absolutely cannot catch new content that way. Content that isn't widely shared but shared between two people or in rather small groups. These small groups have elaborate vouching and proving systems in place, and are hard to infiltrate, going as far as having to prove yourself by sharing a picture of you victim holding a sign with a time stamp and some passphrase.
facebook aimed their AI at porn so they might catch some of this new content by accident; it's porn after all. Then again, my hopes aren't that high, seeing Microsoft's bing fucked up even removing the known child abuse content from search results even as the company runs the PhotoDNA database I mentioned before. And google's AI is "clever" enough to think black teens are gorillas, but they are supposed to catch child porn?
Also, those dedicated child abuse content producers do read the news and know not to use facebook or twitter. Most of them do, anyway.
There is essentially three types of pedocriminals I have observed: the aforementioned part time online pedos who share the same old child abuse content but do not actually produce it or abuse children in the real world, a group of active pedophiles who do not care about getting caught because they live in places where they do not actually have to fear the police investigating them, and a cautious group who might use common public services to make initial contact and talk a little in code but will immediately exchange tox ids etc or at least link to "how to setup qtox over tor"* guides (I saw a few "groomers" do just that), and never share anything incriminating over a public service. My guess is that they are be behind tor or a VPN even when using those public services.
And there is a small number of dumb fucks too who will get caught easily and who end up in the press, and, of course, a probably quite large group of pedophiles who just abuse children but do not tell anybody about it, especially not on the internet.
The majority of people you'd actually want to neutralize because they engage in ongoing child abuse you will not catch on facebook or twitter or whatever. I sometimes liken it to catching a lot of drug users, but not the dealers let alone the drug producers.
Not that my experience dealing with the police is much better. The average time for them to get back to you is a few months if they are from the West, and usually never in other parts of the world, even if they have dedicated groups and/or tip lines. Maybe sometimes they do investigate without acknowledging the reports you sent, I don't know, but it is my feeling that probably not. I don't want to blame the individual police officers and detectives who have to deal with this shit, it's probably all due to lack of resources and institutional support.
But to shit a bit more on the police: I saw some chats with essentially reverse-grooming, where a "girl" started off telling how "wet" she is and how much she likes dicks and only then that she is only "13". After learning about "her" age the dude then usually quickly leaves, but often only after having posted some personal information already. I later learned that such "girls" are either the police "child grooming" units, or blackmailers. I don't even want to know how many of the grooming cases that actually go to court are a result of this tactic, catching stupid horny wannabe pervs, while the actual groomers go free. Burning resources like this instead of using them to catch the truly evil and nasty people again seems to be a political decision.
Source: personal experience from having to deal with this shit occasionally, and having caused some pedophiles to go to prison.
* There have been a few studies that found that the group of people consuming child abuse media and the group that actually molests and rapes children doesn't have too much overlap, contrary to what one might expect.
A lot of the "common" content is rather old, like digitized VHS tapes and magazine scans, or stuff like "1st st" (the producer of which went to prison like 10 years ago) or the East European nude "modelling" stuff from the early 2000s, or webcams from a time when webcams had half a megapixel. There is some newer stuff, like "Rbn" and "Rgirls", but that's the exception. And if you try to fill in the blanks and google any of this, you're stupid and/or evil; don't.
* I have had a few tell that to my face, taunting me to report them to their local police.
I am singling out tox/qtox because at least at in the recent past it seems to have been the goto tool for security conscious pedophiles, based on my personal observations.