What is AI, really?

[joluoto]

And you think humans learn to talk by doing something other than anticipating what words will yield the desired results based on how they have observed words being used?

Humans actually think about what they say, not just generate words based on probability.

 

[stevecanuck]

And you think humans learn to talk by doing something other than anticipating what words will yield the desired results based on how they have observed words being used?

You keep drawing a false equivalence between thinking with an actual brain and a LITERALLY mindless machine following a algorithm.

 

[AConcernedCitizen]

Humans actually think about what they say, not just generate words based on probability.

What distinction are you making there?

Humans fire a series of signals from one neuron to another to generate output behavior from input data. ANNs likewise send signals from neuron to neuron to generate output behavior from input data.

What metric are you using to draw a distinction between one algorithm and another?

 

[AConcernedCitizen]

You keep drawing a false equivalence between thinking with an actual brain and a LITERALLY mindless machine following a algorithm.

It isn’t literally mindless. That is the point.

Suppose you want a function that takes an integer as an output, and returns twice the value as an output.

That process is so simple that even a human can wrap her brain around it.

int DoubleValue(int Value)
{
return Value * 2;
}

But what if you want to take an mp3 of you speaking as an input and produce an mp3 that sounds like Donald Trump saying the same thing? What kind of mathematical operation are you going to perform on an array of bytes that will return an array of bytes that will sound like Donald Trump?

No human programmer knows the logic involved in that.

Instead of trying to figure out what process needs to happen to that array of bytes, the programmer creates a network of neurons. A mind. She determines the activation function from neuron to neuron, and sets rules for how the mind will learn and grow.

Then the created mind figures out what it needs to do to the input data to return the desired output.

 

[Gatsby]

STOP!!!!! Yer killin' me.

I didn't know it was even possible for that level of ignorance to exist.

@AConcernedCitizen is absolutely correct, and you are wrong.

 

[Gatsby]

Humans actually think about what they say, not just generate words based on probability.

What objective test would you suggest be used to determine if an allegedly-intelligent entity is "actually thinking about what they say" or just mindlessly generating words? What does it even mean to "actually think," instead of just performing complex algorithms with one's neurons which result in some sort of output?

 

[stevecanuck]

Humans fire a series of signals from one neuron to another to generate output behavior from input data. ANNs likewise send signals from neuron to neuron to generate output behavior from input data.

Humans fire a series of signals from one neuron to another to generate output behavior from input data. ANNs likewise send signals from neuron to neuron to generate human-created-algorithm-driven output behavior from input data.

Fixed it for you.

 

[AConcernedCitizen]

Humans fire a series of signals from one neuron to another to generate output behavior from input data. ANNs likewise send signals from neuron to neuron to generate human-created-algorithm-driven output behavior from input data.

Fixed it for you.

I never suggested that humans weren’t the ones creating the artificial intelligence.

 

[stevecanuck]

It isn’t literally mindless. That is the point.

Suppose you want a function that takes an integer as an output, and returns twice the value as an output.

That process is so simple that even a human can wrap her brain around it.

int DoubleValue(int Value)
{
return Value * 2;
}

The foundation of your argument seems to be that computing should have been called AI from the day it was invented.

But what if you want to take an mp3 of you speaking as an input and produce an mp3 that sounds like Donald Trump saying the same thing? What kind of mathematical operation are you going to perform on an array of bytes that will return an array of bytes that will sound like Donald Trump?

No human programmer knows the logic involved in that.

Signal matching has been around for decades. both inputs are separated into their component parts of frequency, phase, and amplitude. A digital operator is then created to match one to the other. Decades.

Instead of trying to figure out what process needs to happen to that array of bytes, the programmer creates a network of neurons. A mind. She determines the activation function from neuron to neuron, and sets rules for how the mind will learn and grow.

Then the created mind figures out what it needs to do to the input data to return the desired output.

ohdearohdearohdearohdear.

I guess you'll never understand the difference between actual thinking and simulated thinking. Don't see the point of continuing this.

 

[stevecanuck]

What objective test would you suggest be used to determine if an allegedly-intelligent entity is "actually thinking about what they say" or just mindlessly generating words? What does it even mean to "actually think," instead of just performing complex algorithms with one's neurons which result in some sort of output?

You answered your own question. Complex algorithms are the product of real thought by real people. Even algorithms that are advanced enough to create other algorithms come back to the original human-created set of instructions. Seriously, read the underlined bit that YOU wrote until you get it.

(btw, it's not just me saying this. My toaster oven explained it to me in so much detail that it got carried away and burned my toast. Intelligence my ass.)

 

[AConcernedCitizen]

The foundation of your argument seems to be that computing should have been called AI from the day it was invented.

It is not.

Signal matching has been around for decades. both inputs are separated into their component parts of frequency, phase, and amplitude. A digital operator is then created to match one to the other. Decades.

AI has also been around for decades. The perceptron was developed in 1958.

 

[stevecanuck]

It is not.



AI has also been around for decades. The perceptron was developed in 1958.

You dodged.

I answered your question directly, and you ignored it.

That was your last chance. Bye.

 

[AConcernedCitizen]

You dodged.

I answered your question directly, and you ignored it.

That was your last chance. Bye.

There wasn’t anything to dodge, since you didn’t make any arguments. You made a non sequitur statement about how long signal matching has been around.

 

[stevecanuck]

There wasn’t anything to dodge, since you didn’t make any arguments. You made a non sequitur statement about how long signal matching has been around.

Ohmygod.

I thought I was out, but you pulled me back in.

Here it is again, and please pay attention this time:

You asked: "But what if you want to take an mp3 of you speaking as an input and produce an mp3 that sounds like Donald Trump saying the same thing? What kind of mathematical operation are you going to perform on an array of bytes that will return an array of bytes that will sound like Donald Trump?

I explained: Signal matching has been around for decades. both inputs are separated into their component parts of frequency, phase, and amplitude. A digital operator is then created to match one to the other.

 

[AConcernedCitizen]

Ohmygod.

I thought I was out, but you pulled me back in.

Here it is again, and please pay attention this time:

You asked: "But what if you want to take an mp3 of you speaking as an input and produce an mp3 that sounds like Donald Trump saying the same thing? What kind of mathematical operation are you going to perform on an array of bytes that will return an array of bytes that will sound like Donald Trump?

I explained: Signal matching has been around for decades. both inputs are separated into their component parts of frequency, phase, and amplitude. A digital operator is then created to match one to the other.

Signal matching is not a mathematical operation that you can perform on an array of bytes to return another array of bytes that sounds like Donald Trump.

You can use traditional signal processing to make a vocoder effect, shift pitch, or warp formants, but recognizable celebrity voice emulation is beyond what traditional signal-processing and hand-coded logic can accomplish. It absolutely requires AI tech.

 

[stevecanuck]

Signal matching is not a mathematical operation that you can perform on an array of bytes to return another array of bytes that sounds like Donald Trump.

You can use traditional signal processing to make a vocoder effect, shift pitch, or warp formants, but recognizable celebrity voice emulation is beyond what traditional signal-processing and hand-coded logic can accomplish. It absolutely requires AI tech.

You have a nice day.

 

[AConcernedCitizen]

You have a nice day.

You as well.