Ai Dreams Forum

Artificial Intelligence => General AI Discussion => Topic started by: DemonRaven on October 13, 2015, 05:25:16 pm

Title: This author has my thoughts about AI down to a T
Post by: DemonRaven on October 13, 2015, 05:25:16 pm

How Artificial Intelligence Will Really Kill Us All

http://paultyma.blogspot.com/2015/10/how-artificial-intelligence-will-really.html?utm_content=buffer5e4b6&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer (http://paultyma.blogspot.com/2015/10/how-artificial-intelligence-will-really.html?utm_content=buffer5e4b6&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer)

Basically to sum it up we will end up using it to kill each other. The software has never been the problem the humans making the software are.

Title: Re: This author has my thoughts about AI down to a T
Post by: Art on October 14, 2015, 12:23:11 am

Good stuff! O0

Title: Re: This author has my thoughts about AI down to a T
Post by: Heresiarch on November 18, 2015, 09:50:39 pm

I like Paul's analysis here. AI isn't gonna go all Forbin Project on us any time without being close first.

The biggest AI threats that I see are: [1] an unrestricted AI attached to a nanotech printer that produces grey goo, or [2] an unrestricted AI attached to a secret manufacturing plant that's able to produce a civilization's worth of military weapons without anyone knowing about it (which suggests a lot about the AI system having 'purpose'), or [3] an unrestricted AI that, Forbin-like, humans decide to attach to a large repertoire of nuclear missiles without any failsafe device, or [4] an AI that grows "so quickly" that is somehow develops plans to boostrap some 3rd-party online code generator into a virus .exe and thence into real-world objects and thence into goo or military weapons.

Yet the first 3 kinda assume that people on Ben Carson's mental plane are the ones designing these systems. Once we got close to nanotechnology, I think there'd be billions of dollars poured through identifying, preventing, containing, and combating grey goo. There's already been a movie about #3. My issue with #4 is that it requires an AI complex enough to plan, and hence can deal with consequences, yet no-one stuck a "... and don't kill all humans" line into the planning module.

As for #2... do you trust the NSA?

Title: Re: This author has my thoughts about AI down to a T
Post by: 8pla.net on November 19, 2015, 01:16:39 am

Sometimes, I wonder if we already are, without realizing it.