The future of AI?

[vorlon13]

The Star Trek: TNG episode with the particle fountain and the exo-comps highlighted an approach that might be useful at some point: software that can rewrite itself (m/l). They even noted the process frequently went awry as I suspect if such a feature (tee hee) were possible, that would be a frequent result.

I also note that when Data was describing how many operations per second his internals could do, it pretty much to me defined Data as non-living. Any device that is executing instructions as Data seemed to be describing could be imitated by vast numbers of people with pencil and paper doing the same processing. It might take millions of people centuries and centuries to 'boot up' a Data analog on paper, but the principle would be the same, and the rest of us would NEVER described the resulting cubic miles of pencil marks on paper as 'alive'.

[IATIA]

Our sun breathes, eats, metabolizes and even had children. So, what is alive? I would not really consider the sun alive nor a machine, but could the machine still be considered sentient at a point?

t looks like a duck, walks like a duck and quacks like a duck ... ?

[Cato]

The sun does none of the things you claim except through metaphor.

[IATIA]

What is eating besides the transformation of matter into other matter and the release of energy in the process?

[I_am_not_mafia]

I have not seen this addressed.  It could be important because after the fact, it will become a moot issue.

Isaac Asimov's "Three Laws of Robotics"

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
   
   
2. A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
   
   
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
   

[url=http://www.auburn.edu/~vestmon/robotics.html][/url]

How do you implement them?

[I_am_not_mafia]

The Star Trek: TNG episode with the particle fountain and the exo-comps highlighted an approach that might be useful at some point: software that can rewrite itself (m/l).  They even noted the process frequently went awry as I suspect if such a feature (tee hee) were possible, that would be a frequent result.

The problem is that there will always be something that does not re-write itself. It's like poking yourself in the stomach with your finger and saying that you are a self poking stomach.

I also note that when Data was describing how many operations per second his internals could do, it pretty much to me defined Data as non-living. Any device that is executing instructions as Data seemed to be describing could be imitated by vast numbers of people with pencil and paper doing the same processing.  It might take millions of people centuries and centuries to 'boot up' a Data analog on paper, but the principle would be the same, and the rest of us would NEVER described the resulting cubic miles of pencil marks on paper as 'alive'.

We can simulate neural networks on a computer and the same argument applies. You could theoretically do all those calculations by hand. There is no reason to think that something needs to be analog in order to be alive. The single best defining characteristic of life is that it has a metabolism.

[IATIA]

I have not seen this addressed.  It could be important because after the fact, it will become a moot issue.

Isaac Asimov's "Three Laws of Robotics"

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
   
   
2. A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
   
   
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
   

How do you implement them?

Myself, I would make a C++ algorithm that ran under an interrupt system and hard code it into the boot ROM.

[I_am_not_mafia]

How do you implement them?

Myself, I would make a C++ algorithm that ran under an interrupt system and hard code it into the boot ROM.

But it takes intelligence to recognise a human being and a law and to predict the consequences of laws. You can't put that in ROM.

[IATIA]

Myself, I would make a C++ algorithm that ran under an interrupt system and hard code it into the boot ROM.

But it takes intelligence to recognise a human being and a law and to predict the consequences of laws.

Just a proper algorithm is needed.

You can't put that in ROM.

Why not?  ROM is just memory that cannot be changed.  To change that memory would require replacing the ROM chip or the microprocessor which has the embedded ROM.

[Alex K]

To hardcode these restrictions on the AI into a ROM the problem is that you'd need to know how exactly, in the artificial neural network, the concepts of Humans, injury etc are represented. That's not at all obvious in a deep learning system and may be impossible to determine uniquely let alone beforehand.