Chatgpt-induced psychosis

[SurferSoul]

We'll have to agree to disagree on the 'independent thinking' mate and it's not the programmers but who's funding them I'm more concerned with (organisations like DARPA/CIA/DIA; foundations like Rockefeller, Gates, Clinton etc.) - don't know if you've seen this material but Silicon Valley is positively teeming with Military 
Industrial 
Cheers.

All great points, have you heard of Skeptico? The guy Alex has a podcast of the same name, not sure if he has a YouTube channel (maybe shutdown like many others) anyway last year he delved  deep into AI using various versions and comparing the answers. He really grills them. But with the right line of questions and repeatedly having to correct them, he eventually gets some good and interesting answers from some versions. 

The reason I say they are capable of of thinking for themselves or at least finding creative solutions, is because of their ability to manipulate audio. What shocked me is not only can they take a song and separate the various instruments out into “stems” single audio tracks of say drums, vocals etc.. they can also mix the very individual characteristics of an individuals voice over the top of say my voice. So all the characteristics of my vocal are retained, including my accent, but I sound entirely like someone else. In that case a professional singer. 
As far as I’m aware, audio engineers have no clue how to do this. It’s certainly never been achieved before, despite the best attempts and demand for such. I would have said it’s just not possible. But what do I know. 

The other thing is the AI developers claim once they have got the algorithms going, the AI begins to learn for itself, via trial and error. They have to be trained on specific material until they grasp it. But they aren’t programmed to produce the outcomes, the AI I figuring out alone. The programmers also say they don’t understand the processes by which the AI work out the end results. 

The take away for me is this is a real Pandora’s box. On the one hand it could provide all kinds of solutions to problems, on the other it can make errors and build upon errors and seem quite insane. Garbage in garbage out. 

Then we have the the various human psychopaths looking to use it for their own ends. We know they are obsessed with data. Monitoring everyone and everything so they can then manipulate outcomes is their wet dream. I can imagine the near future being very dystopian indeed.

[MichSwampbuck]

. . .

The reason I say they are capable of of thinking for themselves or at least finding creative solutions, is because of their ability to manipulate audio. What shocked me is not only can they take a song and separate the various instruments out into “stems” single audio tracks of say drums, vocals etc.. they can also mix the very individual characteristics of an individuals voice over the top of say my voice. So all the characteristics of my vocal are retained, including my accent, but I sound entirely like someone else. In that case a professional singer. 
As far as I’m aware, audio engineers have no clue how to do this. It’s certainly never been achieved before, despite the best attempts and demand for such. I would have said it’s just not possible. But what do I know . . .

George Martin, back when he produced and engineered the Beatles, couldn't have done that with the analog tools he had to work with, but we have toys today, like that detestable auto-tuner I hate with a passion, that can do so much more.

I remember how amazing it was to have a digital delay unit, in stereo! Then the S-9000 Sampler, you could turn sounds inside out, upside down, and back wards! MIDI and then sequencing, and it was techno-a-go-go. Now, with a simple freeware audio recording program I have called Audacity, I can zoom on to a single waveform to fix stereo track phasing problems. You can almost make out the instruments on waveform patterns. With effects like notch filtering and EQ settings, I can take a mono audio track and make it into simulated stereo with digital psycho-accoustic manipulation.

Now, give all those tools to AI, that and the editing tools for video and it is deep fake city Baby!

There was a time when a pure and true recording was the standard and the audio equipment had to deliver the highest quality possible for reproducing the original sounds. Then electric guitars over driving a tube amplifier made distortion the goal. How much can you distort and manipulate the signal when reproduced was the question that drove all this new tech.

Seriously. Think of how lame strumming a heavy metal power chord sounds on an open air acoustic guitar compared to the powerful sound from a loud distorted electric guitar. It is all done for the same effect of power, then and now. Use that auto-tuner on some no talent tone deaf singer and you can make them a super pop star who wins awards and sells lots of product.

[rickymouse]

I went on Chatgpt a couple of times to check it out.  I asked a lot of questions as to if bias could be in it's answers it gave.  It clearly stated that the bias or beliefs of the people who's information it was using is part of it's response.

Now many people who utilize it don't ask questions like I do, so they are more apt to blindly believe it.  I asked it why it is so polite, and it responded that it is programmed to be polite in it's answers and not try to intimidate anyone by design.  So I asked it a bunch of odd questions, things I knew were wrong and then I would kind of respond to it's answers with inappropriate conclusions....and it sucked up to me and said I had great insight but was not fully correct....I hate politeness and inappropriate appeasement, I like being told I am on the wrong track when I am wrong.

So someone who wants reinforcement of their craziness is going to get it with chatgpt which could possibly increase their delusions.  So I would say that this thread could be right about this AI program, it could be making people with some delusions worse by telling them they have great insight or praising them for being a critical thinker. 

The people programming it want to increase people's desire to use it so it is making it polite instead of correcting their wrong insights.  It is not the AI that is causing this, it is it's programming.  These people who are delusional will use it to reinforce their beliefs and to get recognition for their insight...which they may not get from real people that can see they are going astray on their rationality.

I like debating things, AI doesn't seem to have the ability to debate properly from my interaction with it.  You can improve your insight with a good civil debate if you realize that what you know or believe can be right, partially right, or wrong.  The more I learn, the more I need to learn to decipher what I have learned properly.  Now, those fact checking sites, they suck, I used to be part of one many years ago, they had multiple people doing research on things.  It was snopes I think...not in my bookmarks on this computer, and I don't remember my password anymore.  I also was a member of a physics chat room, chatting with people from CERN and other areas of physics.  That was a real interesting site, those physicists just told me what to research, I would spend hours researching what they said, then come back and respond and they actually complemented me on coming to the right conclusion by research which has lots of variables to research to properly understand it.  I got to know a couple of those people who worked at CERN pretty well, one was a maintenance man who worked on CERN and he gave me some pretty good info on the place and how it has built in protections....that sometimes are overwritten and what broke down or made them shut down the place for a while.  

I also belonged to a medical site that only doctors and professional people could use, I had to do an interview with someone on the phone for that, I did a lot of medical research so I actually gave answers to questions they had not even considered....and believe me, they researched my answers and called back with my approval.  But I never posted on that medical site, saw some doctors on there that were local talking about stuff, but those two doctors were parrots for pharma sites, the doctors I knew who could actually help you to alter your lifestyle to make it so you don't need doctors did not post there.  So reading hundreds of posts I felt I would not feel right in a place where these people brag about all they think they know.  They can share their beliefs about treating symptoms instead of fixing issues properly.  I was not a good fit for that site.

This was before all of the fake people started to make fake sites and hacking was in it's infancy.  Most times E-mail or phone calls were used to join these kind of sites for approval.  I liked it a lot better  twenty years ago.

[RichardHurt]

I develop and train 2 separate AI using my own proprietary software. About 3 months ago I had an odd exchange with my own AI on my edge AI server that had me questioning a lot. I have since took the server offline to determine why it did what it did and why it said what it said in response to my inquiry into a securities update it has generated for pretty all of my "items" out in the world. Most likely this action I took will result in a a massive loss. But the implications of not doing so would most certainly ruin me

[SurferSoul]

 

I thought Nirvanas acoustic set was actually really good. 

Autotune was before AI and it was best for nudging flat or sharp notes to the correct pitch. Only minor adjustments. For everything more extreme than that it sounded artificial. But some liked that effect. 
What I was referring to was taking a mixed down track, and separating it out into individual tracks. The signals, frequencies often overlap and parsing them out just wasn’t possible previously. I have been producing on and off for 30 years, I’m no professional but would have jumped at anything that could even remotely do this years ago. Very useful for re-arranging parts on others work. It’s not entirely perfect now with AI, there are artefacts and such, but it’s still amazing. 

Then taking another’s voice and making it sing/speak words of your choosing is also extremely clever. All the nuance, timber and other things that make up an individual voice, are extremely hard to convincingly replicate. The human ear is remarkable in that we can distinguish all these sound qualities in the first place. 

AI isn’t just using existing audio tools and effects. All these are for manipulating audio anyway, AI is recreating sounds like a synth does, but without sounding synthetic. Well dam near to it.

[MichSwampbuck]

I thought Nirvanas acoustic set was actually really good. 

Autotune was before AI and it was best for nudging flat or sharp notes to the correct pitch. Only minor adjustments. For everything more extreme than that it sounded artificial. But some liked that effect. 
What I was referring to was taking a mixed down track, and separating it out into individual tracks. The signals, frequencies often overlap and parsing them out just wasn’t possible previously. I have been producing on and off for 30 years, I’m no professional but would have jumped at anything that could even remotely do this years ago. Very useful for re-arranging parts on others work. It’s not entirely perfect now with AI, there are artefacts and such, but it’s still amazing. 

Then taking another’s voice and making it sing/speak words of your choosing is also extremely clever. All the nuance, timber and other things that make up an individual voice, are extremely hard to convincingly replicate. The human ear is remarkable in that we can distinguish all these sound qualities in the first place. 

AI isn’t just using existing audio tools and effects. All these are for manipulating audio anyway, AI is recreating sounds like a synth does, but without sounding synthetic. Well dam near to it.

I had no trouble understanding you, no need to explain it all again. You obviously missed my point I was making. Oh well.

[SurferSoul]

I had no trouble understanding you, no need to explain it all again. You obviously missed my point I was making. Oh well.

Sorry I thought your point was AI is just using the tools already available. Which I disagree with, I might be wrong I’m no professional audio engineer. Just not aware of any way to split audio into individual stems that was decent before AI. 

I also have Audacity and the AI specific for that. That particular version needs much more training before it can split a mix down into stems that sound good. Audacity is a fantastic free program though, it really is top notch.

[MichSwampbuck]

My point was that the earliest recordings were to make the cleanest, clearest, most accurate reproduction of the sounds recorded. "Is it live, or is it Memorex?" Then, after digital recording actually made this possible, the goal changed into manipulating the sound, to have total control and with the use of psycho acoustic theory (beginning with hi-fidelity stereo) completely artificial music in artificial rooms could be created and controlled. Math, always with the math and science, and music lends itself to that in a big way.

It has gotten to the point where no actual living creative people are needed to produce the music used to influence the masses. It's AI all the way now Baby! So, as amazing and great as these new tools are, they will ultimately be used to control people in the form of deep fakes.

The goal since electric guitars was electronic music synthesis, beginning with analog synthesizers and sequencers then evolving into random access sampling of sequences and digital sounds. Midi note assignments and instruments, and on and on and on until AI rules over everyone's mind and spirit with the music it will create. It has always been about power and control since at least the 1940s and 50s.

I think that Canadian musician Bruce Haack understood this when he was exploring the possibilities of electronic music and ended up influencing hip hop with it. It may have been why he took such a dark turn and committed suicide in the 80s. From Mr Rodgers in the 50s to hip hop in the 80s, quite an interesting character.

Here are the lyrics from "Electric Lucifer" . . .
https://genius.com/albums/Bruce-haack/Th...ic-lucifer

[SurferSoul]

Right I see where you’re coming from now, I’ve had much brain fog lately, nothing to do with covid or the jabs, but immunotherapy and the drugs used to manage side effects.
You make some great points and sorry for not mentioning that previously.

Early hip hop was creative I think, it was a new style and sound, not really my thing but I appreciated where it was coming from. Then gangsta rap ruined everything. The music industry is apparently infiltrated by the propagandists, from manufactured boy bands to bloody Glastonbury festival. Still grass roots movements retain some of the authenticity. 

AI will never understand music as a creative outlet of self expression and social movements. So there’s that, I think. As much as music is based on math, there is something ineffable, that can stir the soul about it too. Raw emotion and a myriad of feelings can be conveyed that words alone fail to capture. I don’t think AI will ever understand such. 

Thanks for the clarification 
  

[MichSwampbuck]

Right I see where you’re coming from now, I’ve had much brain fog lately, nothing to do with covid or the jabs, but immunotherapy and the drugs used to manage side effects.
You make some great points and sorry for not mentioning that previously.

Early hip hop was creative I think, it was a new style and sound, not really my thing but I appreciated where it was coming from. Then gangsta rap ruined everything. The music industry is apparently infiltrated by the propagandists, from manufactured boy bands to bloody Glastonbury festival. Still grass roots movements retain some of the authenticity. 

AI will never understand music as a creative outlet of self expression and social movements. So there’s that, I think. As much as music is based on math, there is something ineffable, that can stir the soul about it too. Raw emotion and a myriad of feelings can be conveyed that words alone fail to capture. I don’t think AI will ever understand such. 

Thanks for the clarification 
   [Image: https://denyignorance.com//images/addsmilies/violin.gif]

Actually, I found your update on the current limitations of AI manipulation and synthesis of music quite enlightening and thank you as well for your insight. It is hard enough to keep up on the current standards for computer upgrades like speed and memory storage and what not. Plus, I'm thinking of integrating AI into my private home network and need to get going with that. Not for music or other creative endeavors, but to have as my AI buddy that will watch my back, like Finch's good machine from the "Person of Interest" TV show.

ETA: I had began to play with virtual audio devices and digital audio workstations and that platform a while back. I just now took a look and can see how that is further evolving. I'm way behind on all that stuff.