(C) Alec Muffett's DropSafe blog.
Author Name: Alec Muffett
This story was originally published on allecmuffett.com. [1]
License: CC-BY-SA 3.0.[2]


Two thoughts on the Future of AI Regulation, Part 2: The Regulation of Public Power

2024-02-09 11:48:20+00:00

We are less than 25 years past a time where games consoles were being classified as supercomputers and subjected to export controls, just in case they were purchased in bulk by China, Russia, Iran, or any other country lacking robust McDonalds revenues, and then repurposed towards modelling nuclear bomb simulations, searching for nerve agents, or any other headline-grabbing thing that could be fearmongered to the public.

It’s not a stretch to assume that that world will return – wrapped in a flag of public safety or child protection – and that regulators will deem individual access to “too much” compute power is inequitable and clearly a societal risk because (e.g.) they may use such compute power for antidemocratic or abusive ends such as production of misinformation or deepfakes.

This is a topic which I have alluded to in a previous blogpost, where we could be heading at speed to a world where activities that are legal with paper and pencil become illegal if done by a computer, and instructions which if given to a human worker are legal will become likewise illegal if issued to an LLM.

This is the regulatory fallacy that “scale” is the problem; it’s very popular and you see it in many aspects of “Trust & Safety”; e.g. it’s true that you can choke-down the propagation of spam and scams by restricting the forwarding of message attachments to 5x other users at a single time… but that can be worked around fairly trivially by a committed “bad guy” — which simply means that the threat vector is exploited by people who are good at being bad, and the victims are never inoculated by exposure to those who are bad at being bad.

Of course nobody wants to say this quiet part out loud because the implication is that one expects people to learn to recognise and deal with shit on their own, contrary to the demands of politicians and the promises of platform PR wonks.

But anyway: in the same way that nobody needs more than 640k of memory, we should totally expect to see legislation, quite soon, telling us that “nobody should be permitted more than <so many bogoMIPS, computrons, or teraflops> at home” – and that beyond a certain limit people and organisations which do have such, will need to be registered with the state, and taxed.

Because nothing good ever came from people hacking on a computer in a garage, right?
[END]

[1] URL: https://alecmuffett.com/article/109130
[2] URL: https://creativecommons.org/licenses/by-sa/3.0/

DropSafe Blog via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/alecmuffett/