Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

PURPLE

0 (0%)

ALERT

0 (0%)

(I need suggestions is what I'm saying.)

0 (0%)

Total Members Voted: 0

[alway]

If it were an AI, let it go.

However, the stated dilemma says you can not figure out anything about it. This poses the critical problem of its true nature. Considering the probability of a strong AI vs the probability of a spear phishing scheme to get you to allow an unknown program access to the net, I would alert sysadmins to a likely malware infection.

[Acanthus117]

I'd leave.

I'm not qualified to make these decisions.

[andrea]

I wouldn't let it go.
If I don't know anything about it, then releasing it in the wild is a rather bad idea.

however, if something can consciously ask to live, then it deserves to live. I'll wait for the scientist the next day and try to convice him. And to understand his reasons.
I may have to forcefully stop him, if I am strong enough ( not hit, mind you. Just hold until he agrees to speak and reason). On the other hand, I may remove any cable that might allow the AI to escape on its own. Unless I get undeniable proof that the AI is good willed and the scientist is crazy.

[ToonyMan]

Is it truly a super massive file?  A wouldn't be able to run down to a computer store and purchase a 5TB harddrive for it to chill in?  That doesn't sound right at all.  Where is it suppose to hang out on the internet anyway if it can't even fit in 5TB of damn space.

[counting]

Is it truly a super massive file?  A wouldn't be able to run down to a computer store and purchase a 5TB harddrive for it to chill in?  That doesn't sound right at all.  Where is it suppose to hang out on the internet anyway if it can't even fit in 5TB of damn space.

I don't believe even 5TB storage is enough. If an AI system which at present most likely as an adapted system using network-like system, and in order for it to be sentient as human, it will be as complex as human brain. And in human brain there are 10^11 neuron, and 10^4 connections each, than there is at least S*10^15 of data to be stored (S is the data size where a connection is represented, most likely at least 1kb). And 1 TB is only 10^12 bytes. You will need hundreds of thousands of harddrives, and not include the process of divided the data and probably the right hardware for it to function. And even using the highest bandwidth like 100 Gigabit Ethernet at full capacity, it will need weeks for it to upload to another mass storage units. (It's most likely just trading one cage for another)

But this is calculated base on current tech level, it's hard to say what it will be like in the future. Maybe one day the storage device is calculated by Petabyte or Exabyte, it will be portable. And that's no where soon in near future. But I believed the basic dilemma here is that an AI is essentially different, and implied it has properties as it can be copied and interacted with other computerized system at will (NOT necessary true, like humans can NOT control other living things just by will alone, you need tools, as in computer-world it means protocols and APIs), and they are essentially "alien" as human. But as a system at that level of complexity, I think anyone should at least try to make an effort to save it as a remarkable engineering effort. (I doubt who invented it will simply wipe out their works that easily, at least making backup is the basic rule of research.). Another problem may be that what we consider something has its own right, but that's another issue altogether.

[MetalSlimeHunt]

Is it truly a super massive file?  A wouldn't be able to run down to a computer store and purchase a 5TB harddrive for it to chill in?  That doesn't sound right at all.  Where is it suppose to hang out on the internet anyway if it can't even fit in 5TB of damn space.

As counting mentioned, even the largest portable hard drive doesn't even compare to being able to hold the AI. You need a large complex of high-memory interconnected servers at the very least, which is where the AI would transfer itself to if allowed to escape. There's just no telling where an active candidate for that would be until the AI is allowed to access the internet and scan for potential systems. However, for the purposes of the scenario, assume that this candidate does exist somewhere.

[Angel Of Death]

Could I have my vote removed from "Save the A.I via internet" and put into "Stay with the AI until morning and try to reason with your colleague not to kill it."?

[MetalSlimeHunt]

I don't know why, but apparently the change vote button has vanished. Again.

[counting]

As counting mentioned, even the largest portable hard drive doesn't even compare to being able to hold the AI. You need a large complex of high-memory interconnected servers at the very least, which is where the AI would transfer itself to if allowed to escape. There's just no telling where an active candidate for that would be until the AI is allowed to access the internet and scan for potential systems. However, for the purposes of the scenario, assume that this candidate does exist somewhere.

This assumption has already giving part of the answers, since if the candidates are common on the Internet than it most likely in that future, AI are no longer that "difficult" and general accepted and potentially used in everyday lives. Than the decisions are mos likely a easy one at that time. Or if it just a very few candidates in not so advanced future, than it's likely just as I said before, its trading one cage for a new one, and not much of a problem.

But if the candidates are many, but are used as other purpose (I don't know what that might be, probably a massive online virtual reality game consoles like that). And the AI will still need to be complex enough to write itself protocols and APIs for it to function properly in the new hardware environment. Essentially the AI must be super smart programmers to rewrite its own codes, with super adaptivity (not likely happened), or the creator already intended the AI to be adapted like a virus, and has his/her own agenda other than pure AI research purposes. In that case. the characters of the creator of AI is more important than the AI program itself. And by releasing such AI programs may cause a mass chaos and disrupts on the original none-AI purpose candidate systems/services. This is the consequences we may face in this dilemma. (And the creator has the first responsibility, and you are only unwittingly assisting offenders.)

P.S. As someone who actually write AI (NN programs), I am not a good candidate for this test actually

[RedKing]

Moral of this thread: Keep B12'ers out of advanced computer science labs, you SHODAN-lovers.

[MetalSlimeHunt]

Moral of this thread: Keep B12'ers out of advanced computer science labs, you SHODAN-lovers.

Moral of this thread: Keep B12'ers out of advanced computer science labs, you intolerant synthetic lifeform haters.

[counting]

Moral of this thread: Keep B12'ers out of advanced computer science labs, you SHODAN-lovers.

Moral of this thread: Keep B12'ers out of advanced computer science labs, you intolerant synthetic lifeform haters.

As I remembered I've read an articles in Science American that there are researches in AI field already trying to impose ethical decision makings in designing everyday interaction robotics. So I am intrigued with what would an AI do when it faces all these dilemmas itself rather than a human. Maybe it will give surprising results than humans do. (Let AI to decide the AI's fate, AI juries? , probably even more frightening to most people)

[ToonyMan]

How does the AI fit on the computer it's on right now?!

[Reelyanoob]

Well a machine with ethics coded in will probably stick to that ethical line until the end (though it may be coded to have some sort of flexible learning algorithm), unlike a human who's ethics may collapse/change under pressure.

The machine will 'know' what it's ethic tell it, it will never be conflicted the same way a human is.

e.g. a fuzzy logic system which gives option A a score of 50% and option B a score of only 49% will inerrantly go for option A with no regrets, whereas a human faced with such a "close call" will have a hard time deciding.

A computer/logic system could make ANY decision, given the right inputs and rules. The main difference will be that barring changes in the external situation, the computer AI will always make the same decision in the same situation.

----

Also, regarding the size of the AI brain needed to match a human intellect, I suspect it will be a lot simpler (in terms of transistors) than a human brain in terms of neurons. Basically because a lot of neurons are redundant, it's a grown system, while a built AI could be carefully crafted with less "scaffolding" and most of the brain is physical regulatory stuff, which would not be needed by an AI (the hardware manages a lot of that in a computer, so it wouldn't be needed in software).

Consider the complexity of action insects are capable of, and the minute size of brain. An AI could be more of that order of complexity.

[MetalSlimeHunt]

How does the AI fit on the computer it's on right now?!

Because it's a bunch of tower servers hooked up to one another. Not something you could fit in your car. (Maybe a Semi, but you don't have time for that in the scenario.)