[ExI] AI enhancing / replacing human abilities
pharos at gmail.com
Tue Apr 4 21:57:01 UTC 2023
On Tue, 4 Apr 2023 at 21:56, Gadersd via extropy-chat
<extropy-chat at lists.extropy.org> wrote:
> I concur. In an adversarial environment it is almost never optimal from the perspective of one group to halt progress if the others cannot be prevented from continuing.
> The AI safety obsession is quite moot as any malicious organization with significant capital can develop and deploy its own AI. AI safety can only achieve the goal of preventing low-capital individuals from using AI for malicious reasons for a time until the technology becomes cheap enough for anyone to develop powerful AI.
> I am not sure how much good prolonging the eventual ability for any individual to use AI for harm will do. We will have to face this reality eventually. Perhaps a case can be made for prolonging individual AI-powered efficacy until we have the public safety mechanisms in place to deal with it.
> In any case this only applies to little individuals. China and others will have their way with AI.
Interesting thought, that 'friendly' AI means malicious use by the human owners.
In the past, the main worry was AI running amok and destroying
humanity. So the 'friendly' AI design was developed to try to ensure
that humanity would be safe from AI.
But how can we protect humanity from humanity?
Nations and corporations will be running the powerful AGI machines,
controlling economies and war machines.
Personal AI will probably have to be much less capable
in order to run on smartphones and laptops.
But there will be plenty to keep the population amused. :)
More information about the extropy-chat