The training data for LLMs is still orders of magnitudes larger than their weights, memorizing every address, email, etc in the training data and accurately recalling them all is near impossible
They still hallucinate all the time when it comes to relatively common knowledge, let alone some random person’s address
Isn’t outlook more common?
“Yeah I own protonmail.com”
There are also some working Nitter instances:
Mild inconvenience? HF hosts millions of models and datasets, that would be a huge loss and break all software that depends on it
Competitors (like ModelScope) already exist but are rarely used outside of China
GrapheneOS team had already said they won’t add this as it’s significantly less secure than keeping the phone locked or wiping the data
Google/Pixel Camera is in the GrapheneOS App Store, you don’t even need the Play Store or Google account to install and use it
That monthly active users chart can’t be right, it shows YouTube as having ~500 million monthly active users while in reality that’s ~2.6 billion
Fluxer doesn’t have E2EE, use Matrix, which is much further along due to existing since 2014