Impact of AI

xilman · 2017-01-12, 10:07

The Beeb has an article titledMEPs vote on robots' legal status - and if a kill switch is required

As expected from a news report aimed at a lay audience it's rather short on details for the topics it covers and it omits many topics which I think are important. Nonetheless I think it a good read.

henryzz · 2017-01-12, 11:19

The second law concerns me a little. What if the robot thinks a command conflicts with the first law when the human knows that not taking action does. Safeguards are needed for humans not being able to tell it to do something dangerous but humans need to be able to make the decisions in a dangerous situation.

J F · 2017-01-12, 23:15

Quote:

Originally Posted by henryzz

The second law concerns me a little. What if the robot thinks a command conflicts with the first law when the human knows that not taking action does. Safeguards are needed for humans not being able to tell it to do something dangerous but humans need to be able to make the decisions in a dangerous situation.

In your scenario, robot and human evaluate a given situation differently.
You seem to assume that the human is right, but what if he isn't?
Option to override means changing order of rule 1 and 2; no idea if
that would work any better.
Exchange the robot with a 2nd human, that always happens. Disagreement
based on asymmetric information, what course of action is the 'right' one?
Future tells, afterwards...

Robot1: I save the human
Robot2: No, I do, better suited, got upgrades
Robot1: But I have... - ohh, look, it's dead

CRGreathouse · 2017-01-13, 07:38

Quote:

Originally Posted by J F

In your scenario, robot and human evaluate a given situation differently.
You seem to assume that the human is right, but what if he isn't?
Option to override means changing order of rule 1 and 2; no idea if
that would work any better.

Mandatory xkcd:
https://xkcd.com/1613/

xilman · 2017-01-13, 11:26

One topic which was not discussed, even though I think it's important, is the matter of AI rights.

If the strong AI thesis is correct, the responses of a sufficiently advanced AI is indistinguishable from that of an intelligent organic life form. Under the strong AI model, what rights should an AI have?

Even if you don't believe the strong AI thesis, one could still argue whether they have any rights at all and, if so, what. At present some undeniably non-intelligent organic life forms have been given limited rights to survival. Consider the CITES protocols for example.

Nick · 2017-01-13, 13:16

Quote:

Originally Posted by xilman

Under the strong AI model, what rights should an AI have?

You expect there will be more than one of them, then?

xilman · 2017-01-13, 16:48

Quote:

Originally Posted by Nick

You expect there will be more than one of them, then?

I don't see why not.

Humans have reproductive rights, why not AIs?

jasong · 2017-01-15, 04:49

Quote:

Originally Posted by xilman

I don't see why not.

Humans have reproductive rights, why not AIs?

I think the question should've been,"You don't think there will be more than one of them?"

If the question confuses you, maybe reread the thread and consider your grammar.

xilman · 2017-01-15, 07:15

Quote:

Originally Posted by jasong

I think the question should've been,"You don't think there will be more than one of them?"

If the question confuses you, maybe reread the thread and consider your grammar.

There may be more than one AI and there may be more than one right, both in my view of course. One right which humans have, by and large, is the right to reproduce. AIs may als be given that right. If it is exercised there will be more than one AI.

retina · 2017-01-15, 13:46

Please define what you mean by "AI".

How much "I" does something need to have before we need to bestow rights to AIs? Smarter than the average chimp? As smart as the dimmest human? Smarter than the average human?

henryzz · 2017-01-15, 14:30

Quote:

Originally Posted by J F

In your scenario, robot and human evaluate a given situation differently.
You seem to assume that the human is right, but what if he isn't?
Option to override means changing order of rule 1 and 2; no idea if
that would work any better.
Exchange the robot with a 2nd human, that always happens. Disagreement
based on asymmetric information, what course of action is the 'right' one?
Future tells, afterwards...

Robot1: I save the human
Robot2: No, I do, better suited, got upgrades
Robot1: But I have... - ohh, look, it's dead

Long term I would expect the AI to be better. Even in some situations now it would be but there are definitely situations where humans are better than computers. Driverless cars still aren't perfect yet.

2017-01-12, 10:07	#1
xilman Bamboozled! "𒉺𒌌𒇷𒆷𒀭" May 2003 Down not across 2·7³·17 Posts	Impact of AI The Beeb has an article titledMEPs vote on robots' legal status - and if a kill switch is required As expected from a news report aimed at a lay audience it's rather short on details for the topics it covers and it omits many topics which I think are important. Nonetheless I think it a good read.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GPUs impact on TF	petrw1	GPU Computing	0	2013-01-06 03:23
GPU TF work and its impact on P-1	davieddy	Lounge	161	2011-08-09 10:27
Another Impact on Jupiter	Spherical Cow	Astronomy	24	2009-08-12 19:32
NASA's Deep Impact...	ixfd64	Lounge	5	2005-07-06 13:46
Prescott impact to prime95.	nucleon	Hardware	35	2003-07-22 06:02

2017-01-12, 11:19	#2
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 6031₁₀ Posts	The second law concerns me a little. What if the robot thinks a command conflicts with the first law when the human knows that not taking action does. Safeguards are needed for humans not being able to tell it to do something dangerous but humans need to be able to make the decisions in a dangerous situation.

2017-01-13, 11:26	#5
xilman Bamboozled! "𒉺𒌌𒇷𒆷𒀭" May 2003 Down not across 2D8E₁₆ Posts	One topic which was not discussed, even though I think it's important, is the matter of AI rights. If the strong AI thesis is correct, the responses of a sufficiently advanced AI is indistinguishable from that of an intelligent organic life form. Under the strong AI model, what rights should an AI have? Even if you don't believe the strong AI thesis, one could still argue whether they have any rights at all and, if so, what. At present some undeniably non-intelligent organic life forms have been given limited rights to survival. Consider the CITES protocols for example.

2017-01-15, 13:46	#10
retina Undefined "The unspeakable one" Jun 2006 My evil lair 1101000011110₂ Posts	Please define what you mean by "AI". How much "I" does something need to have before we need to bestow rights to AIs? Smarter than the average chimp? As smart as the dimmest human? Smarter than the average human?