AI can be unpredictable, and we should be worried

Humans can be unpredictable, we should be worried.

If the idea behind AI is to mimic/create how a human thinks, then this shouldn't come as a surprise to anyone that they're unpredictable.
 
Garson007 said:
Re trolley problem: Those who do not flip the switch are retarded.
Click to expand...

So the needs of the many outweigh the needs of the few?

Now consider this situation:
In a particular hosptial there are 5 people who need organs, without them, they will die.
You as the hospital administrator, see that the healthy person who has just walked thought the door is a perfect match for the patients.

Using the exact same logic, the non-"retarded" people would therefore chop up a healthy person to save 5 people.
 
konfab said:
So the needs of the many outweigh the needs of the few?

Now consider this situation:
In a particular hosptial there are 5 people who need organs, without them, they will die.
You as the hospital administrator, see that the healthy person who has just walked thought the door is a perfect match for the patients.

Using the exact same logic, the non-"retarded" people would therefore chop up a healthy person to save 5 people.
Click to expand...

Thats not the same logic. You are actually presenting two logical fallacies. Unequal comparison and a bit of straw.
 
ghoti said:
Thats not the same logic. You are actually presenting two logical fallacies. Unequal comparison and a bit of straw.
Click to expand...

The point of the trolly and hospital problem is to show that you can't just have a simple rule such as "the more people that survive, the better" governing the actions of a system that will follow that rule to the letter.

You missed the point entirely.
 
ghoti said:
Thats not the same logic. You are actually presenting two logical fallacies. Unequal comparison and a bit of straw.
Click to expand...
Why is it different? Comes down to the same thing: you decide if one person dies to save many.
 
ghoti said:
Thats not the same logic. You are actually presenting two logical fallacies. Unequal comparison and a bit of straw.
Click to expand...

It's a ethics/moral problem, how do you teach a computer what is morally and socially acceptable, as there is no specific wrong or right answer to the trolley problem, it is very closely tied to emotions, one of the 5 people, waiting for an organ might be a family member or one of them might be a hated foe.

It isn't a logic problem to solve, but a moral one, I don't see the fallacy
 
Last edited:
I think I'll create an AI being that will do my work for me. I'll stay at home and reap the rewards.
 
konfab said:
So the needs of the many outweigh the needs of the few?

Now consider this situation:
In a particular hosptial there are 5 people who need organs, without them, they will die.
You as the hospital administrator, see that the healthy person who has just walked thought the door is a perfect match for the patients.

Using the exact same logic, the non-"retarded" people would therefore chop up a healthy person to save 5 people.
Click to expand...

Why would AI handle that any differently to the way it is currently handled? I cannot see the dilemma or complication in that scenario, as existing rules would simply determine that it would be handled in the same way that current human intelligence handles the scenario.
 
flick the switch while the train is passing. derail the train save all 6. with some luck. :D
 
In real-world applications, the scope for AI surprises is much wider. A stock-trading AI, for example, will re-invent every single method known to us for maximising return on investment. It will find several that are not yet known to us.
Click to expand...

No, it won't. The scope of what AI can and cannot do (given current technology) is fairly well understood. You couldn't train an AI to place orders since the only way to train it is against a simulator, which wouldn't even closely match reality. Online training isn't feasible, since by the time the sample set got large enough for the AI to be profitable, the investor would be bankrupt. What you can do is have an AI predict the market, and then have a human, or a non-learning based, deterministic trading execution engine use the data to place trades.

Same for all the other things - should we ever be in a situation where AIs are making life and death decisions, the technology would be far beyond what we have now, likely trained on a dataset that resembles a human's formative years, and comparing it to a machine that can play Go, would seem absurd.
 
Trolleys can be unpredictable too, and we should be worried. Many times you'll be pushing em down the aisle and suddenly it throws a wobbly and you end up in the cereal box pyramid.
 
The creators are Flawed. Then the creation can be flawed at best.
 
wizardofid said:
It's a ethics/moral problem, how do you teach a computer what is morally and socially acceptable, as there is no specific wrong or right answer to the trolley problem, it is very closely tied to emotions, one of the 5 people, waiting for an organ might be a family member or one of them might be a hated foe.

It isn't a logic problem to solve, but a moral one, I don't see the fallacy
Click to expand...
You think ethics and morals will survive the future development of the human race?
Ethics and morals will become concepts of the past.
 
A trolley is running out of control down a track. In its path are five people who have been tied to the track by a mad philosopher.

Fortunately, you could flip a switch, which will lead the trolley down a different track to safety. Unfortunately, there is a single person tied to that track.

Should you flip the switch or do nothing?
Click to expand...

And if that single person is our beloved Pres Zuma?:D
 
Until the creation starts re-creating itself

Magnum said:
The creators are Flawed. Then the creation can be flawed at best.
Click to expand...

Unless it evolves on its own, like us.. but much much faster, and more deliberate.
 
Magnum said:
The creators are Flawed. Then the creation can be flawed at best.
Click to expand...

That analogy only applies if a system cannot learn. The premise for all AI is an intelligent system that can learn from its environment to find a maximum in a problem space.

If you tell an AI to make itself better in a particular way, it will find a solution to do it that no one has ever thought of before.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X