I think in the next five years the internet will change drastically. Walled gardens, invite only, and companies that verify you're a human (I.E. use government identification).
There is now no way to tell if somebody online in a human or a robot. Even video/zoom/voice is no longer a distinguishable.
There could be a technology that cryptographically ensures that each letter/character came from a physical keypress/touch from a human when you want to interact with a human online in written form, but then again a robot could also be pressing the physical switches or interacting with a touchscreen. Perhaps this technology could handle copying/pasting too as long as the text originally came from a human.
I hope we'll develop more robust mechanisms for detecting non-human generated content (including text) that aren't only enforceable by law (e.g. labeling AI-generated content in the EU). Apple Reference Image is one of the steps in this direction.
I actually no joke had the idea of making this in Rust. Basically store snapshots of the history alongside an eventlog (binary file containing compressed chronological input events).
But the problem with this is that someone could write an agent to mimic a human. And if there is a financial incentive to prove that you're not AI, which there always will be, it is inevitable.
I've considered learning to handwrite to write my first drafts like that. But then I could use an AI to generate realistic looking scans of handwriting.
I guess I could physically mail the originals to a third party verifier. That's what it may come down to for creatives.
siriusastrebe · · focus · HN ↗
There is now no way to tell if somebody online in a human or a robot. Even video/zoom/voice is no longer a distinguishable.
rpozarickij · · focus · HN ↗
There could be a technology that cryptographically ensures that each letter/character came from a physical keypress/touch from a human when you want to interact with a human online in written form, but then again a robot could also be pressing the physical switches or interacting with a touchscreen. Perhaps this technology could handle copying/pasting too as long as the text originally came from a human.
I hope we'll develop more robust mechanisms for detecting non-human generated content (including text) that aren't only enforceable by law (e.g. labeling AI-generated content in the EU). Apple Reference Image is one of the steps in this direction.
devilsdata · · focus · HN ↗
But the problem with this is that someone could write an agent to mimic a human. And if there is a financial incentive to prove that you're not AI, which there always will be, it is inevitable.
I've considered learning to handwrite to write my first drafts like that. But then I could use an AI to generate realistic looking scans of handwriting.
I guess I could physically mail the originals to a third party verifier. That's what it may come down to for creatives.