I imagine many people will react only to the headline and not read the article, but:

"Meta tells The Verge that, for now, it’s not training on your unpublished photos with this new feature. “[The Verge’s headline] implies we are currently training our AI models with these photos, which we aren’t. This test doesn’t use people’s photos to improve or train our AI models,”

As someone who is familiar with the ML space, it seems unlikely that the addition of private photos will significantly improve models, as you have mentioned.