[ExI] My comments on Dario Amodei's essay

[Ben Zaiboc]

On 15/09/2026 06:20, Adrian Tymes wrote:
> On Mon, Sep 14, 2026 at 1:23 PM Ben Zaiboc via extropy-chat
> <extropy-chat at lists.extropy.org> wrote:
>> It would be very hard, probably impossible, to prohibit using AI to develop biological weapons. There are so many things this could include. A scenario that comes immediately to mind is using AI to determine the genetic markers that uniquely identify Han chinese (if this isn't already known), separately using AI to develop a means of making the relevant genetic markers readable by an external mechanism (such as an engineered virus), separately using AI to develop a way to render viruses inactive in the presence of such a marker.

> FWIW, when I ran this by Ais, they said that last step would be much
> harder to get past safety filters than the other steps.


Fine. Is it so hard to imagine breaking this down into smaller steps that would each avoid any safety filters?

This is a general principle. Almost anything dangerous can be broken down into component steps that are each not dangerous. There are very few exceptions (probably nuclear weapons, probably not much more, if anything).

There are parallels with 3D printing projectile weapons. Notice how that's not mentioned anymore? The issue of banning it has been quietly dropped. Why? Because it's not possible. Yes, you could ban it in theory, but just about anyone could easily get around the ban. You can't ban printing a hollow cylinder. You can't ban printing a shape that could act as a trigger guard,  etc., etc.

So the problem is, not filtering what AI can do, but somehow restricting what people can do in terms of stitching together separate innocuous things into a larger, dangerous thing. That doesn't have to involve AI at all.

I don't think that's possible.
-- 
Ben