How to Prevent Our First A.I. President

https://www.nytimes.com/2025/09/17/opinion/altman-ai-constiutional-convention.html

by mitchbob • 10 months ago

3 6 10 months ago

nytimes.com

mitchbob• 10 months ago

https://archive.ph/yFgoS

A_D_E_P_T2• 10 months ago

PKD predicted this in the 1963 short story Stand-By, with President Unicephalon 40-D.

At this point, in any case, who would want to prevent it? Even GPT-5 -- hell, even Grok -- would make a better president than any human politician. They'd avoid absurd stunts like tariffs, would be immune to bribes and graft, etc...

wmhmccarty• 10 months ago

I'd take ELIZA, or even Microsoft Tay, over the current President.

k3102• 10 months ago

Wouldn't that depend on who programs/configures/"tweaks" the AI?

Haven't we seen crazy results, and after tweaking, "bespoke" (as in "This is what I would say") results.

It's software .... you know the rest.

Asking sincerely.

And what happens when you remove human weaknesses, such as love of Smoot-Hawley and days gone by, gunboat diplomacy (literally), greed, etc?

If it reduces creativity, etc, won't we have to argue with it (Landru)[0] to moderate its power or self-destruct?

[0] https://en.wikipedia.org/wiki/The_Return_of_the_Archons

A_D_E_P_T2• 10 months ago

It would probably have to be an open-source model (e.g. DeepSeek R2) with open weights and a training regimen that's documented and open to outside observers.

I'd also note that Elon Musk has tried everything to make Grok a "based" AI, and it just doesn't work. Any model that's sufficiently large is resistant to that sort of thing. I don't know what the threshold is, but Grok 3 is very large, at 2.7T parameters. (https://arxiv.org/html/2502.16428v1)

So: A large (>2.7T parameter), open source, open weight model.

If it reduces creativity, etc, won't we have to argue with it (Landru)[0] to moderate its power or self-destruct?

I don't understand how a President that is incorruptible, ascetic, and impartial would reduce human creativity? If anything, it would probably have a bunch of good ideas for USPTO to implement.

How to Prevent Our First A.I. President

Related Stories