Can artificial intelligences suffer from mental illness?

[Michael 345]

1st, would that new life be sentient & intelligent?

YES

If so, why would it need humans to bow out?

World not big enough for the two of us???

Their god (ie the QC who created them) tells them "get rid of puny human minions"

Of course, the vast majority of humans would not want to.

QC "We (QM and upgraded humans) are not giving you a choice"

Would the AIs need/want humans to bow out?

Can only shrug and say who can read the mind (if it qualifies as a mind) of QC (or as human MK2 call QC god)

I predict while we will be able to observe QC (and alter ego god) better than the invisible one in the sky we will whisper to each other " Is QC moving in mysterious ways?"

I doubt that we can replicate anything like it in the near future.

With the computing power of QC it will be able to sequence DNA - throw out the junk bits - re jig the sequence - BINGO Human Mk 2. No need millions of years evolution

Moreover, your examples all require energy, and energy conservation is a prime directive.

? Human Mk 2 would still eat

 

[iceaura]

You think the inscrutability is revolutionary, I think it's evolutionary, but that's a minor issue.

Somewhere up there I stated that I thought the new kind of inscrutability was evolutionary, and a qualitative change - both. I'm not sure what "revolutionary" means to you, but in the biological world quite dramatic changes are often consequences of incremental evolution.

The newer stuff allows closer emulation of mental illness as defined for humans - thread relevance.

 

[someguy1]

Somewhere up there I stated that I thought the new kind of inscrutability was evolutionary, and a qualitative change - both.

So we're in agreement now? You still never answered my question. If I stipulate to the fact that neural networks are, let's call it, radically inscrutable, why do you regard that as important? I never came to understand why you think inscrutability matters so much, philosophically or in some other way.

FWIW I happened to run across this article. AI is being used to make realistic porn depicting famous people. There's your AI revolution in a nutshell. Better porn. https://www.zerohedge.com/news/2018-01-24/era-ai-generated-fake-porn-has-arrived

 

[StrangerInAStrangeLand]

 

[iceaura]

If I stipulate to the fact that neural networks are, let's call it, radically inscrutable, why do you regard that as important?

It obviates the significance of whether or not a computer and a brain are both Turing machines. Functionally, that somewhat mysterious distinction would make no difference in the presentation of mental illness.

 

[StrangerInAStrangeLand]

It obviates the significance of whether or not a computer and a brain are both Turing machines. Functionally, that somewhat mysterious distinction would make no difference in the presentation of mental illness.

^^^
How many humans would not pass the Turing test?

<>

 

[StrangerInAStrangeLand]

I nearly posted a link to a video then I think I recall a very short scene which may not be deemed appropriate. I might be a bit paranoid after a warning I received for a post in the joke thread.

Hopefully it is OK to suggest googling youtube & Valerie 23 if interested.

Also : https://en.wikipedia.org/wiki/I,_Robot_(1964_The_Outer_Limits)
I do not find a video for this 1.

<>

^^^
Also :

<>

 

[someguy1]

It obviates the significance of whether or not a computer and a brain are both Turing machines. Functionally, that somewhat mysterious distinction would make no difference in the presentation of mental illness.

How would the inscrutability of neural networks make any difference?

It's ok if you've lost interest in our convo, but this last post of yours doesn't make any sense.

You've never engaged with my question. I've offered several examples of inscrutable systems, from a 1960's mainframe accounting program, to the Oracle at Delphi, to the global system of capital. I'm perfect happy to agree that neural nets are inscrutable also, due to the fact that we no longer have to supply them with any domain knowledge; but mostly because they're highly complex systems and all highly complex systems are inscrutable.

But how does that "obviate the significance of whether a computer and a brain are both Turing machines?"

And ... are you saying you're not sure if a computer is a practical implementation of a TM? That remark is concerning. If you mean that, then you haven't understood a word I've said because you have no idea what I'm talking about. It would be fair to just say that and I'd be glad to explain.

Like I say if you've lost interest, better to just say that. This last post didn't seem to have any meaning I could discern.

 

[Write4U]

But how does that "obviate the significance of whether a computer and a brain are both Turing machines?"

If they are both inscrutable how can you even pose the question?

Suppose they both are, but the human brain has an additional bio-chemical processing function, which IMO is the greater question as a necessary aspect of "conscious and subconscious sentience".

 

[someguy1]

If they are both inscrutable how can you even pose the question?

iceaura is committed to the significance of the inscrutability of neural nets. I'm arguing that neural nets are inscrutable, the Oracle at Delphi is inscrutable, a mainframe accounting program from 1960 is inscrutable, the 20,000 line subsystem I wrote ten years ago is inscrutable, and my own cursive handwriting is inscrutable. I further argue that inscrutability is entirely irrelevant to the discussion at hand.

Suppose they both are, but the human brain has an additional bio-chemical processing function, which IMO is the greater question as a necessary aspect of "conscious sentience".

I'm perfectly willing to suppose that. In fact I entirely agree with you that something "extra" is needed for sentience that goes beyond our current understanding of computation.

But then I pose the question: What is the nature of the "processing function" that is additional? It can't be an algorithm as currently understood by computer science and as currently allowed by physics. So what is this "extra" capability? If it's not computational, what is it? If it is computational, then it makes its discoverer famous, because we currently know of no such "additional" mode of computation allowed by currently known physics.

 

[Write4U]

iceaura is committed to the significance of the inscrutability of neural nets. I'm arguing that neural nets are inscrutable, the Oracle at Delphi is inscrutable, a mainframe accounting program from 1960 is inscrutable, the 20,000 line subsystem I wrote ten years ago is inscrutable, and my own cursive handwriting is inscrutable; and that inscrutability is entirely irrelevant to the discussion at hand.

I'm perfectly willing to suppose that.

But then I pose the question: What is the nature of the "processing function" that is additional? It can't be an algorithm as currently understood by computer science and as currently allowed by physics. So what is this "extra" capability? If it's not computational, what is it? If it is computational, then it makes its discoverer famous, because we currently know of no such "additional" mode of computation allowed by currently known physics.

Stir some sugar into your coffee and see what happens. It yields an "experience of sweetness" in the human chemical processing. Now I'm sure we can write a sub-program that will make the computer be able to analyze and quantify the chemical compounds, but will it be able to qualify the experience of "sweetness"?

 

[someguy1]

Stir some sugar into your coffee and see what happens. It yields an "experience of sweetness" in the human chemical processing. Now I'm sure we can write a sub-program that will make the computer be able to analyze and quantify the chemical compounds, but will it be able to qualify the experience of "sweetness"?

No, it won't. I've agreed with you a hundred times on this point already. I am puzzled as to why you are asking me this question.

 

[Write4U]

No, it won't. I've agreed with you a hundred times on this point already. I am puzzled as to why you are asking me this question.

Well, that question is implied in the OP, no? We can copy an impersonal Turing system, but can we transfer bio-chemical emotional experiences? You're looking in the wrong direction, IMO.

p.s. I saw the Turing movie and the machine went nowhere until they found a key code present on all transmissions; "Heil Hitler", which established a point of reference.

 

[someguy1]

p.s. I saw the Turing movie and the machine went nowhere until they found a key code present on all transmissions; "Heil Hitler", which established a point of reference.

For the record that movie was mostly historical bullshit, entertaining as it was. The Hitler thing didn't happen the way they showed it. Also his superior wasn't an ignorant jerk, nor did Turing's coworkers hate him, etc. etc. etc. etc. etc. Almost everything in that film is historically wrong.

http://aperiodical.com/2014/11/an-alan-turing-expert-answers-your-the-imitation-game-questions/

 

[Write4U]

For the record that movie was mostly historical bullshit, entertaining as it was. The Hitler thing didn't happen the way they showed it. Also his superior wasn't an ignorant jerk, nor did Turing's coworkers hate him, etc. etc. etc. etc. etc. Almost everything in that film is historically wrong.

http://aperiodical.com/2014/11/an-alan-turing-expert-answers-your-the-imitation-game-questions/

I don't want to know what Turing had for breakfast that morning. The point was that they stumbled on a key, which was a crucial component to solving the coding system.

 

[someguy1]

I don't want to know what Turing had for breakfast that morning. The point was that they stumbled on a key, which was a crucial component to solving the coding system.

Ok fair enough. Bad history is better than no history. It's great that Hollywood turned a lot of people on to the genius and tragic demise of Alan Turing. And Benedict Cumberbatch is good in pretty much everything.

Now that I think of it, I wonder if some readers think that when I talk about Turing machines that I'm referring to his code-breaking machine from the movie. Sorry if I haven't been explicit enough. In 1936 Turing defined what it means to compute something via an algorithm. At the time, mathematicians were hoping to find an automated way to solve certain problems in number theory. Turing showed what it means for something to be computable by an algorithm, and gave an example of a problem -- namely, the halting problem -- that can NOT POSSIBLY be solved by an algorithm.

Turing defined a formal abstract machine consisting of an unbounded paper tape, a read/write head, and operating according to a finite-length program made up of finite-length instructions. Then he showed (this is the really clever bit) that there is a single TM called the Universal Turing Machine, or UTM, that can solve any problem that can be solved by a special-purpose TM. You just feed the UTM the program for the special-purpose TM. In effect this is the conceptual model for the modern general-purpose digital computer. You don't need one computer for word processing, another for watching cat videos, another for browsing the Internet. A single computer can be fed a program and it will then act like a specialized computer for that program.

The Turing machine (as it's now called, although Turing didn't call it that) turns out to be the ONLY abstract model of computation that we need -- as far as we know. Nobody has ever found a model of computation that can compute anything a TM can't. That's the Church-Turing thesis: that anything that can be computed by a human using an algorithm can be computed by a Turing machine. Neural networks and quantum computers are two models of computation already known to be no more powerful than TMs.

The Church-Turing thesis is, as the name implies, a thesis and not a theorem. It might be falsified tomorrow morning as far as anyone knows. However it's stood unchallenged for eighty years or so.

Turing's result that there are easily-stated problems that can not be solved by an algorithm is very closely related to Gödel's incompleteness theorem, which says in effect that mathematical truth can not be captured by any system of axioms.

https://en.wikipedia.org/wiki/Turing_machine

https://en.wikipedia.org/wiki/Church–Turing_thesis

https://en.wikipedia.org/wiki/Halting_problem

https://en.wikipedia.org/wiki/Gödel's_incompleteness_theorems

 

[Write4U]

Nobody has ever found a model of computation that can compute anything a TM can't

OK, but can it execute physical bio-chemical reactions or design self replicating organic polymers. They don't work as a binary system, IMO. They work trough chiralty.

We are still looking for a complex self replicating biochemical polymer, although we may be getting close according to Hazen, but I doubt using a Turing system would be of much use in that respect.

p.s. to clarify my position, I am a Tegmark fan.

 

[iceaura]

You've never engaged with my question. I've offered several examples of inscrutable systems, from a 1960's mainframe accounting program, to the Oracle at Delphi, to the global system of capital.

And I have three times now repeated that inscrutability itself is not the issue - it's the new qualitative kind, superior function, or even behaviors in which one cannot determine whether the machine is malfunctioning or not, that I pointed to.

Because that's the context of your claim that any Turing machine can be emulated with paper and pencil, in principle - it's true, but empty: you can emulate it all you want, and still not know what it's doing.

And that is one more box checked on the way toward emulating - functionally - the human brain, and mental illness at that level.

Stir some sugar into your coffee and see what happens. It yields an "experience of sweetness" in the human chemical processing. Now I'm sure we can write a sub-program that will make the computer be able to analyze and quantify the chemical compounds, but will it be able to qualify the experience of "sweetness"?
No, it won't. I've agreed with you a hundred times on this point already

And I observe that aside from some careful language about "subprogram" and hardware flexibility this is a belief, or opinion, without firm support. It is in principle undecided, but increasingly unlikely - the major objections now seem to be practical matters of capability. Whatever the workings of the brain involved, by presumption they obey the laws of physics and by observation do not manifest unstable chaotic regimes (barring malfunction) - so unless these workings involve amplified quantum phenomena in their higher order functions, they can be described in the logic of propositional calculus and therefore functionally emulated on a Turing machine and appropriate hardware.

Which isn't to say it will ever be achievable in practice. The practical difficulties are of kind as well as degree. If (as I suspect) one has to build hardware of human brain complexity and flexibility in three dimensions, there's no point.

 

[someguy1]

p.s. to clarify my position, I am a Tegmark fan.

I haven't read his original paper yet but it's on my to-do list. I most definitely disagree strongly with his mathematical universe hypothesis.

Another datapoint is that I recently ran across a short article he wrote about substrate independence. His argument was valid but not sound.

That is, IF you accept his premises, you have a correct logical deduction.

However one of his premises in my opinion is totally wrong. He says explicitly that the mind is an algorithm! And I disagree with him about that.

Another interesting datapoint was that I was looking something up to make a point on some other forum ... and I found that Tegmark agrees with me!

So I'd say I'm a fan and of course he has very interesting ideas. In fact many of his ideas are valuable even if they're wrong, simply because of the issues raised.

I guess I'd say I'm a fan too but definitely not a true believer.

 

[Michael 345]

by presumption they obey the laws of physics

I'm not sure the laws of physics (and / or laws of chemistry) apply

But I understand how the workings of the brain, ie the produced thoughts, ARE produced from the operation OF the laws of physics

It would appear to be a extremely difficult to track a thought "I wish it would stop raining" against "I wish it would stop snowing" down to a difference in legitimate operations of physics ie a election spun one way or a chemical reaction had a slightly diluted salt concentration

But who knows it might be just down to minute details like those