Is it really AI if it's just doing text to speech? We've had that technology for years, but now suddenly it's "AI"? I think people are throwing around that term just to jump on a bandwagon far more than any actual AI being present. There's more to AI than just writing a program. Some sort of machine learning is supposed to be taking place.
Machine learning has been used to teach it to do text to speech.
We've had different text to speech technologies for decades, but this one has been taught with machine learning, so it's an AI.
Describe this "machine learning", please. Because in my mind, all machines can ever do is what they are programmed to do. There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
Is it really AI if it's just doing text to speech? We've had that technology for years, but now suddenly it's "AI"? I think people are throwing around that term just to jump on a bandwagon far more than any actual AI being present. There's more to AI than just writing a program. Some sort of machine learning is supposed to be taking place.
Machine learning has been used to teach it to do text to speech.
We've had different text to speech technologies for decades, but this one has been taught with machine learning, so it's an AI.
Describe this "machine learning", please. Because in my mind, all machines can ever do is what they are programmed to do. There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
I had intended to share the story of how machine learning had been used to teach a computer to beat Pong in some number of hours, at least several hundred as I recall.
So I find this story from last year where human neurons in a dish learned to beat Pong in 5 minutes.
Facinating yet scary at the same time as you just know where this leads....like, where does someone acquire human brain neurons, especially in quantity.
Just trying to live long enough to play a new, released MMORPG, playing New Worlds atm
Fools find no pleasure in understanding but delight in airing their own opinions. Pvbs 18:2, NIV
Don't just play games, inhabit virtual worlds™
"This is the most intelligent, well qualified and articulate response to a post I have ever seen on these forums. It's a shame most people here won't have the attention span to read past the second line." - Anon
Is it really AI if it's just doing text to speech? We've had that technology for years, but now suddenly it's "AI"? I think people are throwing around that term just to jump on a bandwagon far more than any actual AI being present. There's more to AI than just writing a program. Some sort of machine learning is supposed to be taking place.
Machine learning has been used to teach it to do text to speech.
We've had different text to speech technologies for decades, but this one has been taught with machine learning, so it's an AI.
Describe this "machine learning", please. Because in my mind, all machines can ever do is what they are programmed to do. There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
A computer program is given: 1. Large set of examples 2. Ability to modify its own algorithm 3. Ability to determine which of its algorithm modifications produce results that best match the given example
Then the program creates some modifications of its own algorithm (2), tests them to determine which one is best (3), picks the best one and uses it to create further modifications (2), tests those further modifications to determine which one is best (3), after test again picks the best modification and modifies it further (2), tests it again (3), and so on.
As long as the examples given to the computer have some common rule(s) that the program can learn, with enough repetitions the computer usually manages to learn an algorithm that is fairly good at following those rules.
Machine learning can so far only be used to problems that are limited in scope, but as long as the problem and its solution are something that can be run by our programs, it's true learning.
Please note that this is very simplified version of how machine learning works.
so is AI making the voices or are they prerecorded?
I just asked Microsoft Copilot and it told me this:
According to the description of the VoiceOver addon by MrThinger1,
it uses AI speech software to generate voices for Classic WoW quests,
rather than recorded voices. The addon uses a neural network model
called ChatGPT, which is trained on a large corpus of text and speech
data, to produce realistic and natural sounding voices for NPCs. The
addon also allows users to customize the voice settings, such as pitch,
speed, and accent, for different races and genders. The addon is
compatible with both the original and the Burning Crusade versions of
Classic WoW.
Err ... the problem is not speech, the problem is emotional speech.
And that isnt solved by classic reading programs, who speak of course in a completely emotionless way, since they have no way of knowing the emotion and no way to express the emotion.
How much AI can do this, well, that remains to be seen. Frankly I expect a flood of just godawful cases. People enormously overestimate what AI can do, and underestimate how extremely expensive in both development (well, training) and execution AI is.
Is it really AI if it's just doing text to speech? We've had that technology for years, but now suddenly it's "AI"? I think people are throwing around that term just to jump on a bandwagon far more than any actual AI being present. There's more to AI than just writing a program. Some sort of machine learning is supposed to be taking place.
Machine learning has been used to teach it to do text to speech.
We've had different text to speech technologies for decades, but this one has been taught with machine learning, so it's an AI.
Describe this "machine learning", please. Because in my mind, all machines can ever do is what they are programmed to do. There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
I had intended to share the story of how machine learning had been used to teach a computer to beat Pong in some number of hours, at least several hundred as I recall.
So I find this story from last year where human neurons in a dish learned to beat Pong in 5 minutes.
Facinating yet scary at the same time as you just know where this leads....like, where does someone acquire human brain neurons, especially in quantity.
That's very interesting. I was thinking from the start of that post that there needs to be some means for the "AI" to recognize that something is "good", and wants more of that. Then they talked about using real neurons, and that opened that door. The bit about feeding the neurons with reactions... "If the neurons “hit” the ball—that is, showing the corresponding type of electrical activity—the team then zapped them at that location with the same frequency each time. It’s a bit like establishing a “habit” for the neurons. If they missed the ball, then they were zapped with electrical noise that disrupted the neural network"... is interesting. Harmony replaces dopamine? I assume they already knew of this effect because of making that choice.
The alternative to using living neuron cells is to make artificial ones. That requires a lot of code, and goes back to what I said: "There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it." Even going back to this example of using living neuron cells is subject to this, because those cells are making decisions and being rewarded based on coding that tells these neural networks what's what.
So again, they are not really learning, they are not really making decisions, or making choices. They are doing what they are programmed to do.
Is it really AI if it's just doing text to speech? We've had that technology for years, but now suddenly it's "AI"? I think people are throwing around that term just to jump on a bandwagon far more than any actual AI being present. There's more to AI than just writing a program. Some sort of machine learning is supposed to be taking place.
Machine learning has been used to teach it to do text to speech.
We've had different text to speech technologies for decades, but this one has been taught with machine learning, so it's an AI.
Describe this "machine learning", please. Because in my mind, all machines can ever do is what they are programmed to do. There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
A computer program is given: 1. Large set of examples 2. Ability to modify its own algorithm 3. Ability to determine which of its algorithm modifications produce results that best match the given example
Then the program creates some modifications of its own algorithm (2), tests them to determine which one is best (3), picks the best one and uses it to create further modifications (2), tests those further modifications to determine which one is best (3), after test again picks the best modification and modifies it further (2), tests it again (3), and so on.
As long as the examples given to the computer have some common rule(s) that the program can learn, with enough repetitions the computer usually manages to learn an algorithm that is fairly good at following those rules.
Machine learning can so far only be used to problems that are limited in scope, but as long as the problem and its solution are something that can be run by our programs, it's true learning.
Please note that this is very simplified version of how machine learning works.
That's interesting, but I'm not convinced. In Kyleran's example above using Pong, all that computer is doing is putting two or more things it's already got in its memory to come up with a solution. Is that really learning?
I have to admit that I'm having trouble coming up with an example to use here. I thought of two things that ancient man learned. Starting fire, and smelting iron out of ore and then turning iron into steel. But I realized that in each of these examples, ancient man probably witnessed something from experiences and then combined that with another thing witnessed to come up with these things to try, and eventually succeed. No different than the Pong example, really. Just much more detailed.
I don't know yet. There's a lot to think through. (i.e. I'm not quite willing to concede your point yet.)
Comments
Philosophy of MMO Game Design
Because in my mind, all machines can ever do is what they are programmed to do.
There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it.
Once upon a time....
So I find this story from last year where human neurons in a dish learned to beat Pong in 5 minutes.
Facinating yet scary at the same time as you just know where this leads....like, where does someone acquire human brain neurons, especially in quantity.
Brrr....
https://singularityhub.com/2022/10/18/neurons-in-a-dish-learned-to-play-pong-in-virtual-reality/
"True friends stab you in the front." | Oscar Wilde
"I need to finish" - Christian Wolff: The Accountant
Just trying to live long enough to play a new, released MMORPG, playing New Worlds atm
Fools find no pleasure in understanding but delight in airing their own opinions. Pvbs 18:2, NIV
Don't just play games, inhabit virtual worlds™
"This is the most intelligent, well qualified and articulate response to a post I have ever seen on these forums. It's a shame most people here won't have the attention span to read past the second line." - Anon
1. Large set of examples
2. Ability to modify its own algorithm
3. Ability to determine which of its algorithm modifications produce results that best match the given example
Then the program creates some modifications of its own algorithm (2), tests them to determine which one is best (3), picks the best one and uses it to create further modifications (2), tests those further modifications to determine which one is best (3), after test again picks the best modification and modifies it further (2), tests it again (3), and so on.
As long as the examples given to the computer have some common rule(s) that the program can learn, with enough repetitions the computer usually manages to learn an algorithm that is fairly good at following those rules.
Machine learning can so far only be used to problems that are limited in scope, but as long as the problem and its solution are something that can be run by our programs, it's true learning.
Please note that this is very simplified version of how machine learning works.
Then they talked about using real neurons, and that opened that door. The bit about feeding the neurons with reactions...
"If the neurons “hit” the ball—that is, showing the corresponding type of electrical activity—the team then zapped them at that location with the same frequency each time. It’s a bit like establishing a “habit” for the neurons. If they missed the ball, then they were zapped with electrical noise that disrupted the neural network"...
is interesting.
Harmony replaces dopamine? I assume they already knew of this effect because of making that choice.
The alternative to using living neuron cells is to make artificial ones. That requires a lot of code, and goes back to what I said:
"There is no such thing as a machine learning without said programming, and it will always be defined by that programming. There's no way around it."
Even going back to this example of using living neuron cells is subject to this, because those cells are making decisions and being rewarded based on coding that tells these neural networks what's what.
So again, they are not really learning, they are not really making decisions, or making choices. They are doing what they are programmed to do.
Once upon a time....
In Kyleran's example above using Pong, all that computer is doing is putting two or more things it's already got in its memory to come up with a solution.
Is that really learning?
I have to admit that I'm having trouble coming up with an example to use here.
I thought of two things that ancient man learned.
Starting fire, and smelting iron out of ore and then turning iron into steel.
But I realized that in each of these examples, ancient man probably witnessed something from experiences and then combined that with another thing witnessed to come up with these things to try, and eventually succeed. No different than the Pong example, really. Just much more detailed.
I don't know yet. There's a lot to think through. (i.e. I'm not quite willing to concede your point yet.)
Once upon a time....
"I am an AI and you need me!"