Sorry, no, because the way you move will identify you and everything you do will be tied to that identity. That's what the article is saying. So even if you're anonymous since the beginning, companies can tie that to "Person A" and everything you do will be under this name, "Person A". It doesn't matter if you never share anything about yourself in real life because they will eventually get enough data on "Person A" that they'll know what you like and what you'll spend money on so you will be targeted with ads and be compelled to spend more of your money on whatever it is that you like. If you really want to be anonymous, you'd have to have some sort of interface that moves for you in a randomized different way each time. And you wouldn't be able to play video games with thsi interface with any skill because your movements wouldn't be tied to your real life movements. The only way to escape is to not log on.

According to the article, if you use the metaverse, even if you are only tracked using three data points -- your head and two hands -- you can be identified in 2 seconds. With more data, it would take even less time. So even with no information about you now, any time you visit the metaverse, anything you do can be associated with you, whether it's playing games with friends, shopping, or chatting on message boards. If you shared who you are in real life then that's tied to your meta profile as well, If you log in as a "different identity" it doesn't matter because your movements are the same, unless you introduce noise, which would interfere with fidelity and be disadvantageous to you. So there's really no escape.

Can you train the model on proprietary date? For example, behind the firewall at a company at the wikis, chats, and knowledge bases of a company so that when you query, you're getting data from inside the company? If not, what data is used to train GPT4? I think the larger this gets, the more garbage in, garbage out. What's the governance model for the accuracy of the data? How is it scored?