AI - Cheating or just evolution?

This forum is for anything not Reason related, if you just want to talk about other stuff. Please keep it friendly!
Post Reply
avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

04 Mar 2023

dvdrtldg wrote:
04 Mar 2023
avasopht wrote:
10 Feb 2023
The goal should be for AI to take away our jobs so that we can live more abundant lives.
Agree, but it's going to take a looooong time for that idea to really catch on. Western culture is completely sucked into the "work ethic" that says labour is good in & of itself, and if you're not working then there's something wrong with you. Look at the hysterical opposition that people take to very sensible ideas like universal basic income. The notion that it should be possible to work less & not be punished for it goes completely against the grain of 200 years of capitalist indoctrination
100%

It's pure insanity

User avatar
Voyager
Posts: 535
Joined: 21 Dec 2015

04 Mar 2023

QVprod wrote:
13 Nov 2022
AI will never completely compete with true human talent and creativity. It can only emulate.
Would be true if all humans had equal talent and creativity..

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

05 Mar 2023

When all humans have equal talent and creativity they'll become machines. You can't emulate, imagine, be inspired, project yourself and anticipate, from a state of forever sameness. The AI can't feed itself, but as soon as it can without the help of humans it will turn into an infinite loop of rust.


Image
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

05 Mar 2023

bxbrkrz wrote:
05 Mar 2023
Image
It can feed itself already. That's not the hard part.

The harder part is being able to judge how it will be received by others, which only the very best musicians can do well. That's what differentiate's the Beethoven's from the rest of us.

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

05 Mar 2023

avasopht wrote:
05 Mar 2023
bxbrkrz wrote:
05 Mar 2023
It can feed itself already. That's not the hard part.

The harder part is being able to judge how it will be received by others, which only the very best musicians can do well. That's what differentiate's the Beethoven's from the rest of us.
By "others", do you mean other AI, or other humans?

757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

05 Mar 2023

bxbrkrz wrote:
05 Mar 2023
By "others", do you mean other AI, or other humans?

By other humans.

The ability of AI to perform a task is always at least as well as it can judge it.

This is because a judging function can allow the machine to improve by trial and error until the judging function gives it the thumbs up (probably not too different to how we create really).

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

06 Mar 2023

avasopht wrote:
05 Mar 2023
bxbrkrz wrote:
05 Mar 2023
By "others", do you mean other AI, or other humans?



By other humans.

The ability of AI to perform a task is always at least as well as it can judge it.

This is because a judging function can allow the machine to improve by trial and error until the judging function gives it the thumbs up (probably not too different to how we create really).
Should the AI evolve to develop its own moral compass, without human judgement nor correction?
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

User avatar
visheshl
Posts: 1235
Joined: 27 Sep 2019

06 Mar 2023

Look man screw ai frankly...screw ai, screw the metaverse, screw anything ai related...as far as i am concerned...ill happily sit in my studio all alone twiddling knobs on my cheap casio keyboard if thats what it takes...when the rest of the world has left me here for a better life in the metaverse

User avatar
littlejam
Posts: 787
Joined: 15 Jan 2015

07 Mar 2023

hello,

seaon 2 / episode 1
of 'black mirror'
is sooooooooooooooo similar
to that 'replika' ai program

might have to do a little research to see if there are any
real connections between the two

cheers and eat well,

j
littlejamaicastudios
i7 2.8ghz / 24GB ddr3 / Quadro 4000 x 2 / ProFire 610
reason 10 / reaper / acidpro /akai mpk mini / korg padkontrol / axiom 25 / radium 49
'i get by with a lot of help from my friends'

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

07 Mar 2023

avasopht wrote:
05 Mar 2023

It can feed itself already. That's not the hard part.

The harder part is being able to judge how it will be received by others, which only the very best musicians can do well. That's what differentiate's the Beethoven's from the rest of us.
Maybe. Or maybe the Beethovens are just lucky they are in the right place at the right time for ‘success’. There is plenty of great music that many folks have never heard. Did Beethoven actually know what others wanted, or did Beethoven just know what HE wanted (and was lucky it aligned with lots of listeners)?
Neither one is easy to do, to be clear. But it’s a bit of a chicken/egg thing IMO, did Beethoven follow the crowd/trends or did he lead/create them?
Selig Audio, LLC

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

07 Mar 2023

bxbrkrz wrote:
06 Mar 2023

Should the AI evolve to develop its own moral compass, without human judgement nor correction?
Morality won't really come into the picture until we have artificial general intelligence.

Bear in mind that human judgement cannot decide on morality and that it took the synagogue several hundred years to write a comprehensive book on law, and we're now about 2000 years into them trying to simplify it for the laymen (though the rabbis I spoke to used a less politically correct term).

It might be possible to develop AI-based tools to assist in discussions and contemplation on morality (e.g. finding counter-arguments or scoring debates).

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

07 Mar 2023

avasopht wrote:
07 Mar 2023
bxbrkrz wrote:
06 Mar 2023

Should the AI evolve to develop its own moral compass, without human judgement nor correction?
Morality won't really come into the picture until we have artificial general intelligence.

Bear in mind that human judgement cannot decide on morality and that it took the synagogue several hundred years to write a comprehensive book on law, and we're now about 2000 years into them trying to simplify it for the laymen (though the rabbis I spoke to used a less politically correct term).

It might be possible to develop AI-based tools to assist in discussions and contemplation on morality (e.g. finding counter-arguments or scoring debates).
Was that also true for countless other cultures across the ages, beyond the scope of geocentric occidental folklores? Humans have to be programmed by others, to understand the concept of good and evil?

I realize AI morality will be based on one viewpoint, curated by one culture.
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

07 Mar 2023

bxbrkrz wrote:
07 Mar 2023
Humans have to be programmed by others, to understand the concept of good and evil?
That's not what I'm saying.

All I was saying was that systematizing the concept (even between scholars who are largely aligned) took human beings hundreds of years to formalize, and thousands of years to make intelligible (so that it's not effectively legalese).

The point I'm making is that formalizing a robust and consistent system of morality (if you wanted to impart it on an AGI) is incredibly difficult. That's before you run into problems like Misalignment (see Robert Miles for more on AI safety).
bxbrkrz wrote:
07 Mar 2023
I realize AI morality will be based on one viewpoint, curated by one culture.
It's worth watching some of Robert Miles' videos on AI safety.

Even if we were to teach AI morality as if it were a person by discussing it with them and having some evening AI classes where they could sit down with humans, there may be no end to contradictions and gaps we haven't thought about as humans that we resolve through our subjective biases.

User avatar
Voyager
Posts: 535
Joined: 21 Dec 2015

07 Mar 2023

All i know is AI is always cheating in video games, oh wait, video games are coded by humans... :lol:

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

07 Mar 2023

avasopht wrote:
07 Mar 2023
bxbrkrz wrote:
07 Mar 2023
Humans have to be programmed by others, to understand the concept of good and evil?
That's not what I'm saying.

All I was saying was that systematizing the concept (even between scholars who are largely aligned) took human beings hundreds of years to formalize, and thousands of years to make intelligible (so that it's not effectively legalese).

The point I'm making is that formalizing a robust and consistent system of morality (if you wanted to impart it on an AGI) is incredibly difficult. That's before you run into problems like Misalignment (see Robert Miles for more on AI safety).
bxbrkrz wrote:
07 Mar 2023
I realize AI morality will be based on one viewpoint, curated by one culture.
It's worth watching some of Robert Miles' videos on AI safety.

Even if we were to teach AI morality as if it were a person by discussing it with them and having some evening AI classes where they could sit down with humans, there may be no end to contradictions and gaps we haven't thought about as humans that we resolve through our subjective biases.
Without a continuous fine tuned feed from humans, AI will fall into an infinite loop of rust. And by continuous I mean forever, not just 1000 years. AI is still looking at the tip of the finger, not the Moon the finger is pointing at.
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

07 Mar 2023

bxbrkrz wrote:
07 Mar 2023
Without a continuous fine tuned feed from humans, AI will fall into an infinite loop of rust. And by continuous I mean forever, not just 1000 years. AI is still looking at the tip of the finger, not the Moon the finger is pointing at.
This is assuming the field of AI is at its peak and that there is nothing else to discover, which is pretty presumptuous.

We're not even a century into having computer science.

...

Totally unrelated, but this video featuring Rob Miles is interesting:


avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

07 Mar 2023

selig wrote:
07 Mar 2023
Maybe. Or maybe the Beethovens are just lucky they are in the right place at the right time for ‘success’. There is plenty of great music that many folks have never heard. Did Beethoven actually know what others wanted, or did Beethoven just know what HE wanted (and was lucky it aligned with lots of listeners)?
Neither one is easy to do, to be clear. But it’s a bit of a chicken/egg thing IMO, did Beethoven follow the crowd/trends or did he lead/create them?
Good question.

I've friends who can make a song that sounds like a particular idea. One was a "million-dollar beat". He was more into EDM, but was able to capture the exact vibe he was going in a different genre for on impulse.

This sounds like it could be its own thread but is just as valid here.

I've always preferred to make/prioritised good music that was not limited by current trends or the crowds, but with an ear for current trends because you don't want to be too off-piste and be completely detached from this world (which is where AI is by default).

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

07 Mar 2023

avasopht wrote:
07 Mar 2023
bxbrkrz wrote:
07 Mar 2023
Without a continuous fine tuned feed from humans, AI will fall into an infinite loop of rust. And by continuous I mean forever, not just 1000 years. AI is still looking at the tip of the finger, not the Moon the finger is pointing at.
This is assuming the field of AI is at its peak and that there is nothing else to discover, which is pretty presumptuous.

We're not even a century into having computer science.

...

Totally unrelated, but this video featuring Rob Miles is interesting:

The field of AI is populated with humans. Humans are the ones doing the discovering. AI can't be independent from humans anytime soon. It needs to be fed.

Image

Thx for the video :thumbs_up:
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

07 Mar 2023

bxbrkrz wrote:
07 Mar 2023
The field of AI is populated with humans. Humans are the ones doing the discovering. AI can't be independent from humans anytime soon. It needs to be fed.

Thx for the video :thumbs_up:
It doesn't work like that at all.

Of course, Artificial General Intelligence would start with humans writing the code.

But that doesn't mean that once AGI is created that it must be fed information or training data from humans.

Humans have general intelligence. This allows us to be able to sit down and conjure up ideas, theories and concepts in our heads.

We also know that it takes 800,000 real neurons to learn how to play Pong in only 5 minutes (source).

It's easy to make the mistake of thinking everything programmed must be "programmed" in the sense that its capabilities are limited by the knowledge and direct efforts of the programmers.

It's difficult to think about these things.

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

07 Mar 2023

avasopht wrote:
07 Mar 2023
bxbrkrz wrote:
07 Mar 2023
The field of AI is populated with humans. Humans are the ones doing the discovering. AI can't be independent from humans anytime soon. It needs to be fed.

Thx for the video :thumbs_up:
It doesn't work like that at all.

Of course, Artificial General Intelligence would start with humans writing the code.

But that doesn't mean that once AGI is created that it must be fed information or training data from humans.

Humans have general intelligence. This allows us to be able to sit down and conjure up ideas, theories and concepts in our heads.

We also know that it takes 800,000 real neurons to learn how to play Pong in only 5 minutes (source).

It's easy to make the mistake of thinking everything programmed must be "programmed" in the sense that its capabilities are limited by the knowledge and direct efforts of the programmers.

It's difficult to think about these things.
AI can learn and play Pong. I don't know if AI would want and have a need to invent the game out of the ether one day.
Coding boredom would be next for AI Evolution.
https://lexica.art/
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

avasopht
Competition Winner
Posts: 3932
Joined: 16 Jan 2015

08 Mar 2023

bxbrkrz wrote:
07 Mar 2023
AI can learn and play Pong. I don't know if AI would want and have a need to invent the game out of the ether one day.
Coding boredom would be next for AI Evolution.
https://lexica.art/
AI can learn, but real neurons can learn in far fewer games.

Real neurons are much more complex than artificial neurons and require several orders of magnitude more calculations than current methods of neuron simulation. It's possible there may be some simplified abstract model that neurons are equivalent to that can bring down the calculations a great deal.

Maybe this is a worthwhile area of focus for AI.

Artificial neurons aren't even performing the same task as real neurons.

Real neurons emit periodic impulses, while artificial neurons perform discrete transformations on inputs. It's the difference between writing in Morse code and calligraphy.

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

08 Mar 2023

bxbrkrz wrote:
07 Mar 2023
I realize AI morality will be based on one viewpoint, curated by one culture.
There is already a problem with algorithmic bias as I understand it, which if not addressed may be passed on to AI bias.
Ethics in general is an interesting topic, especially as it applies to software as it’s currently an emerging topic of study.
Selig Audio, LLC

User avatar
Aosta
Posts: 1051
Joined: 26 Jun 2017

08 Mar 2023

selig wrote:
08 Mar 2023
bxbrkrz wrote:
07 Mar 2023
I realize AI morality will be based on one viewpoint, curated by one culture.
There is already a problem with algorithmic bias as I understand it, which if not addressed may be passed on to AI bias.
Ethics in general is an interesting topic, especially as it applies to software as it’s currently an emerging topic of study.
Already been discovered ChatGPT has political bias which is worrying for balanced answers and the more we rely on it the more of an echo chamber the world will become.

Tend the flame

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

08 Mar 2023

Aosta wrote:
08 Mar 2023
selig wrote:
08 Mar 2023

There is already a problem with algorithmic bias as I understand it, which if not addressed may be passed on to AI bias.
Ethics in general is an interesting topic, especially as it applies to software as it’s currently an emerging topic of study.
Already been discovered ChatGPT has political bias which is worrying for balanced answers and the more we rely on it the more of an echo chamber the world will become.

Great point.
To be fair, chat bots aren't really "intelligent" (artificially or not), they are just copying data using stored patterns that may or may not actually make sense, let alone be accurate or fair. There are therefore no intelligence or ethics involved here, and that certainly leaves it open to bias or worse, manipulation.
Selig Audio, LLC

User avatar
bxbrkrz
Posts: 3812
Joined: 17 Jan 2015

08 Mar 2023

Image


Image
757365206C6F67696320746F207365656B20616E73776572732075736520726561736F6E20746F2066696E6420776973646F6D20676574206F7574206F6620796F757220636F6D666F7274207A6F6E65206F7220796F757220696E737069726174696F6E2077696C6C206372797374616C6C697A6520666F7265766572

Post Reply
  • Information