Artificial Intelligence

SWM wrote:
Besides, if people thought she was up and dressed when she wasn't, they might make assumptions that weren't true. ('Why isn't she sleeping?' 'Does she have insomnia?' 'Is something going on?' 'Another late night, doesn't she take care of herself?' 'Oh good, she's up, I can call her again if I need to.')

That is easily solved. Whenever the program is in use, an indicator could be flashing on all appropriate screens. Or, someone like Henke would select a nightie for superimposition. That would get the message across that she was asleep, without exposing too many secrets.

If we go down the path to develop truly independent and self-aware AI's there are 4 outcomes:

1. The AI's becomes so incomprehensible and alien that we have nothing in common and they just ignore/leave us.
2. The AI's works together with humanity (the benign end).
3. The AI's enslave the humanity for their own ends (the rogue AI end).
4. The AI's consider humanity to be a danger and tries to wipe us out (the bad end).

When I think about it there is a 5th outcome, the singularity event end where AI's and technology surges ahead and transform the humanity in unexpected ways. An example of this is Accelerando by Charles Stross or the ending of Heechee series by Frederick Pohl.

If and when we develop a true AI, it will spell the end of a lot things, both good and bad.

cthia wrote:I am developing an unprecedented mathematical program that also augments sections of the screen in real time. It's quite easy actually. And that's without the old-school tricks.

Any hints on what? I used to work with digital video in the 90's and we did some interesting stuff back then, like encoding DVD-quality video @1.5Mbit/s with MPEG2 MP@ML or switching between MPEG2 streams in real time with no delay or re-encoding. We also tinkered a lot with temporal noise reduction (which explains why we could encode video at a very low bitrate without loosing quality). I think Pinnacle bought all the IP when the company I worked for went belly up.

Joat42 wrote:

cthia wrote:I am developing an unprecedented mathematical program that also augments sections of the screen in real time. It's quite easy actually. And that's without the old-school tricks.

Any hints on what? I used to work with digital video in the 90's and we did some interesting stuff back then, like encoding DVD-quality video @1.5Mbit/s with MPEG2 MP@ML or switching between MPEG2 streams in real time with no delay or re-encoding. We also tinkered a lot with temporal noise reduction (which explains why we could encode video at a very low bitrate without loosing quality). I think Pinnacle bought all the IP when the company I worked for went belly up.

Not much of a hint Joat. I am very protective of this baby. Completely unprecedented. When completed, I'd like to go back in time and leave it, instruction manual included, with Einstein.

Small world. Regarding mpeg, I wrote many prototype compression algorithms in the 90's, as a very young kid. That technology spawned many offshoot applications. i.e. The computer gaming industry. In fact, with the very skills you mentioned, after your company went belly up, you should have high tailed it to California. I got rich in Silicone Valley, and they were hiring people with your skills like flies.

We've already mentioned Duhak in another of David's series. The other one would be Owl in the Safehold series.

The thing that both have in common is that they evolved to full sentience. For Duhak it happened in solitude over an very long period of time. For Owl it occurs as a result of interaction with humans over a comparatively short time.

Interesting.

Don

cthia wrote:There certainly should be a concern for security. And I considered that. But then I quickly realized that if access to the program can be attained, then the program is not the problem. General security is the problem. As long as there aren't any security issues with the program itself.

Issues such as dangling pointers, memory leaks, dead variables, bugs, defects, state.

You haven´t thought it through enough. Having the face emulation software allowed internally is an enabling technology. You don´t need any kind of special access to exploit it.

Worse however, is that if someone is using a more advanced emulator, or something similar, the person on the receiving end cannot nearly as easily realise something is being used for spoofing rather than for convenience, if it´s "ok" to use it for convenience.

If it´s not "ok" to use it at all internally, then instantly if someone see that something like that IS being used, it will cause alarms to go off, personal or tech-based or both.

This is why you NEVER EVER install certain types of software in a secure environment.

The problem with a program like this is that the normal use of the program by itself is a security risk, it doesn´t matter if in itself it has zero vulnerabilities.


If you were to use this in "your" ships, ok, then i aquire a copy. Then make an alternate version of that which mimics the original software enough that internal comm software wont see the difference, then i either need a few moles or someone to stuff some extra automated software onboard, doesn´t need to be in any of the high security areas of the systems even.

After that, i can wreak havoc in your fleets.
At least once, and probably a few times.
I don´t need more than that.

And that´s just the most blatant exploit.

What if i had those programs "fixed" to suddenly start switching people with uniforms out for nude pornstars(many options for various replacements here)? Set to do it at the right time and action, the amount of annoyance and/or temporary confusion it could cause is quite major.

Or it could be used for personal attacks on the most competent enemy leaders.


I also have a fair idea how such a program could be used for transmitting information covertly. I´ll keep the "how" for that to myself though.


And i already know that it´s "not that easy", but the point here isn´t that it easy, but that YOU are making it POSSIBLE for your enemies to exploit your own software.

Tenshinai wrote:You haven´t thought it through enough. Having the face emulation software allowed internally is an enabling technology. You don´t need any kind of special access to exploit it.

Worse however, is that if someone is using a more advanced emulator, or something similar, the person on the receiving end cannot nearly as easily realise something is being used for spoofing rather than for convenience, if it´s "ok" to use it for convenience.

If it´s not "ok" to use it at all internally, then instantly if someone see that something like that IS being used, it will cause alarms to go off, personal or tech-based or both.

This is why you NEVER EVER install certain types of software in a secure environment.

The problem with a program like this is that the normal use of the program by itself is a security risk, it doesn´t matter if in itself it has zero vulnerabilities.


If you were to use this in "your" ships, ok, then i aquire a copy. Then make an alternate version of that which mimics the original software enough that internal comm software wont see the difference, then i either need a few moles or someone to stuff some extra automated software onboard, doesn´t need to be in any of the high security areas of the systems even.

After that, i can wreak havoc in your fleets.
At least once, and probably a few times.
I don´t need more than that.

And that´s just the most blatant exploit.

What if i had those programs "fixed" to suddenly start switching people with uniforms out for nude pornstars(many options for various replacements here)? Set to do it at the right time and action, the amount of annoyance and/or temporary confusion it could cause is quite major.

Or it could be used for personal attacks on the most competent enemy leaders.

Or just abused by enlisted personnel to screw with officers or each other, either to skate out of work or get people in trouble; just by impersonating each other or people a bit further up the chain of command.

Yes you could put lockouts and require authentication to use it, have people restricted to "impersonating" only them selves. But all that make it less convinent for the woken-up use that was being proposed and still leaves unnecessary attack vectors (grabbing the auth data, people who forget to sign out, etc ,etc) to grab this software.


Yeah I just don't see sufficient benefit to let something like this on the internal network. (Obviously there are situations where it's quite useful on the bridge to screw with external comms. But that's rare enough, and public enough, you probably don't have to worry much about your communications people placing prank calls from the "Admiral", or otherwise screwing around with the ability.

Tenshinai wrote:The point i was making was much simpler though, how many nasty people would think that it´s a LOVELY idea to make copies of themselves? Clone an heir? Henchmen?

Problem is that clones are not really "copies" of oneself. A clone would be a different person, like an identical twin raised by different parents.

As for henchmen, I'm not sure what you gain by having them be clones.

And to what extent exactly can it be exploited and abused?

The main "bad" use I can see is growing people for spare parts... but we'll probably be able to grow organs alone pretty soon, so even that becomes pointless.

Joat42 wrote:The AI's enslave the humanity for their own ends (the rouge AI end).

Rouge AI?
What does colour have to do with anything?

Michael Everett wrote:

Joat42 wrote:The AI's enslave the humanity for their own ends (the rouge AI end).

Rouge AI?
What does colour have to do with anything?

Gah, was slightly dyslexic due to loss of coffee-intake...