That's actually an excellent point; I was really talking about "external" engineers, rather than internal ones, but yeah, the attrition (or even loss of agency) amongst internal engineers is another huge problem. A closely related problem google's created is that there's been a big loss of self-determination - google likely has the engineers capable of tackling something like that, but they've sunk into more of a standard corporate leadership model where these engineers really aren't allowed to work on something like this - instead of having a free-floating surplus pool of them, freely able to attach themselves to different projects, they're mostly hired on a need-based rationale.
If you're doing a standard corporate model (in the interest of cost control), where you've got exactly the number of engineers you need to maintain/develop certain products, then you've got no flex/slack in the system - everyone you've got is necessary for the projects they're on, and can't be pulled off without hurting those projects.
I remember reading about one of Google's last big bets - which was the major, AI-driven upgrade to google translate, and what was terrifying to me about reading the article was the artificial darwinian survival model applied to it. Firstly - the folks who wanted to work on it had to beg management for buy-in to get anyone the clearance to work on it, and then second of all, they were given a deadline where they had to produce meaningful results within 1-2 year's time, or the whole thing would have the plug pulled.
If we weren't right at the cusp of that technology wave cresting, and ML-driven translation being viable, it'd have gotten axed and set us back by quite a few years. What we need is a company willing to genuinely "no strings attached" commit resources to things like this - because maybe something simply can't be ready for another 5 years. But if it is on a 5 year timeline, we really want them to just put in the 5 years of work so we can have it - instead of cutting all funding and turning it into a 15 year delay.
Because again - if someone like google isn't doing it - nobody else will. The whole cost-cutting rationale on R&D stuff is really a bet that someone else can do the hard work in the meantime, and then once it's close to being ready, you, the company, can hop onboard and cash in. But if you're the only one with the means (be it cash, or in google's case: data sets) to do that hard work, and you're not doing it, then we're all screwed - it means nobody's doing it.