Kazami42
Budget Morpheus, soldier #1,128,300 reporting in.
Intellectual development/because/refinement and stress testing/comfort/entertainment from going mad of boredom/"escapism."
The summarization that ignores the human component is all the word wall below.
Ignoring an endless list of mental pursuits and case by case aspects:
On average it is akin to stress testing and recalibration. Every form of technology no matter the sophistication or advancement will always need this. "goofing around" or other things often viewed as enemy activity by some, like training on video games themselves, are the best means of stress testing all forms of media altering, mimicry, and creation software that exists. AI is also entirely capable of being DIYed and democratized, which means they may lack all the.... Interesting, bits of data, datasets, and resources and thus their development, refinement, and testing stages need to use whatever is on hand. Even if in an idealist utopia, people would still be found doing this. The same as how people would still drink, smoke, or whatever else. Just imagine the technological equivalent.
In this case, we have multiple things we have to test at once:
The creation of an algorithm usually for sorting.
The creation of a software to recognize vocal aspects like inflection, tone, pitch, shifts, and if sophisticated (but not advanced) enough then to shift them.
The creation of yet another system networked or a series of unrobust algorithms to act like a camera to replay, analyze, and other shenanigans audio.
This may or may not involve using audio software itself to invent up new voices, or correct mistakes and failures to fill the gaps.
It is to then also be synched up to a script usually digitally written, to be recited by this "intelligence."
And that is the normal process if one doesn't care about going all the way with it, or merging it with existing models, algorithms, etc. And like most other systems, hybrid designs are better but so anti-robust that new data smashes them to smithereens without remorse or mercy and needs to be retrained from scratch. Then, we can test it out doing random crap. Like Obama playing league, or seeing how well a chatbot works if more extensive, or seeing if it can work as well in something else then correcting it if not. And then you pretty much gut most of it and remake a new one in a couple of years to do the same thing again but faster.
Chatbots can do this better though, but often work on speech biases needing to periodically be lowered but not too much than its level or incomprehensible, consist of so many different systems all compounding to make something new that it may as well be a blackbox, or some rare ones use machine reinforcement learning, the equivalent of a "genetic" or "evolutionary" algorithm set, and random stuff unrelated to being used purely for communicative or assistant purposes. (Because others including DIYers aren't just making it up for corporate reasons.) In other cases, the system is already done.
A super model like a GPT being merged with various voice systems and then used for varying degrees of voice emulation for singing songs, playing games, and reading sci-fi papers is to display publicly the work. At times it is akin to being used like a technological resume for a job.
The rest of the time it is literally because someone got bored, took an existing thing as is, and used it for entertainment and decided to share the end results publicly. But yet, anything that isn't purpose or cripplingly role specialized people for some reason despise, despite generalization tending to be the superior means forwards. And this would indeed, include leisurely activities.
The summarization that ignores the human component is all the word wall below.
Ignoring an endless list of mental pursuits and case by case aspects:
On average it is akin to stress testing and recalibration. Every form of technology no matter the sophistication or advancement will always need this. "goofing around" or other things often viewed as enemy activity by some, like training on video games themselves, are the best means of stress testing all forms of media altering, mimicry, and creation software that exists. AI is also entirely capable of being DIYed and democratized, which means they may lack all the.... Interesting, bits of data, datasets, and resources and thus their development, refinement, and testing stages need to use whatever is on hand. Even if in an idealist utopia, people would still be found doing this. The same as how people would still drink, smoke, or whatever else. Just imagine the technological equivalent.
In this case, we have multiple things we have to test at once:
The creation of an algorithm usually for sorting.
The creation of a software to recognize vocal aspects like inflection, tone, pitch, shifts, and if sophisticated (but not advanced) enough then to shift them.
The creation of yet another system networked or a series of unrobust algorithms to act like a camera to replay, analyze, and other shenanigans audio.
This may or may not involve using audio software itself to invent up new voices, or correct mistakes and failures to fill the gaps.
It is to then also be synched up to a script usually digitally written, to be recited by this "intelligence."
And that is the normal process if one doesn't care about going all the way with it, or merging it with existing models, algorithms, etc. And like most other systems, hybrid designs are better but so anti-robust that new data smashes them to smithereens without remorse or mercy and needs to be retrained from scratch. Then, we can test it out doing random crap. Like Obama playing league, or seeing how well a chatbot works if more extensive, or seeing if it can work as well in something else then correcting it if not. And then you pretty much gut most of it and remake a new one in a couple of years to do the same thing again but faster.
Chatbots can do this better though, but often work on speech biases needing to periodically be lowered but not too much than its level or incomprehensible, consist of so many different systems all compounding to make something new that it may as well be a blackbox, or some rare ones use machine reinforcement learning, the equivalent of a "genetic" or "evolutionary" algorithm set, and random stuff unrelated to being used purely for communicative or assistant purposes. (Because others including DIYers aren't just making it up for corporate reasons.) In other cases, the system is already done.
A super model like a GPT being merged with various voice systems and then used for varying degrees of voice emulation for singing songs, playing games, and reading sci-fi papers is to display publicly the work. At times it is akin to being used like a technological resume for a job.
The rest of the time it is literally because someone got bored, took an existing thing as is, and used it for entertainment and decided to share the end results publicly. But yet, anything that isn't purpose or cripplingly role specialized people for some reason despise, despite generalization tending to be the superior means forwards. And this would indeed, include leisurely activities.