As web servers get overwhelmed by LLM bots, some operators are resorting to programs that demand visiting web clients to perform some relatively expensive computation to be granted access to the website. This is called Proof of Work, but when that computation doesn't yield any useful results, it might as well be called Proof of Waste. Why not get clients to compute something useful and valuable, so that LLM scrapers become an essentially infinite supply of computing power, and our servers pay for themselves and for the creative works we put in them?
Picture computations like those performed by protein folding simulations, electromagnetic signal analysis, and any other highly-distributed computing architectures.
Let's make providers of "compute bites" on the web so that any website can redirect to it any attempted accesses without Proof of Work, and get redirected back with the Proof of Work once they have earned it, by performing the required computation.
With an open architecture, websites and users can choose what kinds of computations they're interested in participating in, which ones offer more favorable terms, etc.
There could even be markets to which paying customers submit fragmentable computations, for pieces to be submitted to users (or third parties) to perform to earn tokens, that can then be presented as Proof of Work to gain access to participating websites.
Upon validating the tokens with the provider, websites accumulate microcredits that can eventually be claimed, and pay for the server, for the production of works published therein, for the operator's living, whatever.
All of this potential is currently being turned into Proof of Waste. Let's make it Proof of Useful Work instead!
Why go through all this trouble?
LLM bots have plenty of computing power. A lot more than cryptocurrency miners. And their computing power is growing fast.
With a race based on computing power, the required computing power will have to be increasingly turned up to the point that legitimate users will be overpowered and locked out (programmed obsolescence much?), while the seemingly infinite funding available to LLM training keeps growing.
It's not a race we can win, unless we get them to row our boat.
If we use their computing power to perform useful computing in exchange for access to our cultural works that they wish, and/or use that computing power to fund our cultural works, we may be able to fund more powerful servers that can keep up, and to grant access to welcome users by other means.
Indeed, servers may offer legitimate (identified or anonymous) users subscriptions with a reasonable amount of access tokens.
This access control open architecture should even enable users to configure powerful computers to perform the Proof of Useful Work computations, or even to outsource it to compute services providers, and then use the earned tokens on battery-powered or slower computers.
In order to avoid enshittification, we need an open architecture implemented with freedom-respecting software throughout: users shouldn't have to run software under someone else's control, not even on their browsers. No mandatory JavaScript, please: web redirects should be enough, given browser(plugin)-based support for token management.
We need multiple providers of compute jobs and tokens, and multiple providers of compute power, and freedom-respecting programs that can be installed to earn tokens, and to make them available to authorized browsers.
Tokens from different providers should be exchangeable, and users and websites must be able to choose favorable providers or start their own, instead of getting locked to specific providers. Then tokens will tend to the value of the computation performed to earn them.
Any takers?
So blong,