Governments are talking about identifying AI generated results as protective measure against misuse, and thatโs right. But ๐ฎ๐ฟ๐ฒ ๐๐ฒ ๐ฎ๐ฏ๐น๐ฒ ๐๐ผ ๐ถ๐ฑ๐ฒ๐ป๐๐ถ๐ณ๐ ๐ผ๐๐ฟ๐๐ฒ๐น๐๐ฒ๐, as humans?
Globally, as human civilization, we struggle to identify the most basic entities in our world. People and Businesses.
The US doesnโt have a person ๐ฎ๐ฑ๐ฑ๐ฟ๐ฒ๐๐ ๐ฟ๐ฒ๐ด๐ถ๐๐๐ฟ๐ฎ๐๐ถ๐ผ๐ป register. Therefore address verification happens there, and across many other countries, by mailing in some random โutility billโ assuming that someone who pays for electricity in a place, also lives there.
The ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐ฟ๐ฒ๐ด๐ถ๐๐๐ฟ๐ฎ๐๐ถ๐ผ๐ป register is different in every country, in every state. And if you want to have a bit more reliable print out from it it costs you dozens of $ for a report that looks like in the 80ies. Not machine-readable anywhere. Humans are required to read and approve that. Cross-country business verification is a mess.
Tax IDs, even in a โharmonizedโ Europe can only be validated automatically for their ID, but not against a specific unique Company entity. It works via Address matching. Itโs likely and possible that wrong VAT IDs are used with similar, but wrong, company addresses. Of course humans are here to help and fix all that months or years down the line in a tax audit.
Legally required acceptance of ๐ฒ๐น๐ฒ๐ฐ๐๐ฟ๐ผ๐ป๐ถ๐ฐ ๐๐ถ๐ด๐ป๐ฎ๐๐๐ฟ๐ฒ, as required by EIDAS law since the 90ies in EU is accepted or rejected on a case by case basis, even from financial organizations like PayPal. Compared to the US thatโs still OK, where still in 2023 no common legally binding electronic signature exists and people keep ๐ฑ๐ผ๐ฐ๐-๐๐ถ๐ด๐ป๐ถ๐ป๐ด ๐ฎ โ๐ฝ๐ถ๐ฐ๐๐๐ฟ๐ฒ ๐ผ๐ณ ๐ฎ ๐๐ถ๐ด๐ป๐ฎ๐๐๐ฟ๐ฒโ ๐ผ๐ป ๐ฃ๐๐๐.
Adding to that, roughly 50% of the US Citizens ๐ฑ๐ผ ๐ป๐ผ๐ ๐ต๐ฎ๐๐ฒ ๐ฎ ๐๐ฎ๐น๐ถ๐ฑ ๐ฝ๐ฎ๐๐๐ฝ๐ผ๐ฟ๐. So no address verification system, and not even an internationally โcompatibleโ ID verification system that could be used.
Personal data shall be protected, yet a consistent and correct identification of the person to protect is unsolved in many relevant parts of the world.
Yes, itโs desirable to know if somebody used a spell-checker or a ChatGPT to write a piece of text, for some.
But is this the real problem?
Of course a perfect deep-fake of myself signing up for random bank accounts across the world must be avoided.
I personally donโt care if the content I write was written with the help of others, machine or human.
But are our current procedures to identify ourselves error-free to begin with? Have ๐บ๐ฐ๐ถ gone thru one of those phone-id procedures lately?
There are basic administrative flaws in the western world, still unsolved - way more fundamental than AI Content foot printing. And I donโt see many people talking about improving those basics.
Are we able to correctly and easily identify ourselves, as humans?
I think this is where it starts, and this is where the AI identification will need to tie in.