[8pla.net]

Friend,

Please let me share a simple suggestion.

One, that I plan to implement for my A.I.
to use in place of audio.  A mute button!

A two line Closed Captioning System.

(Frame#.) (CC text)

0. (empty, not shown)
1. THIS STARTS

(1. moves up one row)

0. THIS STARTS
1. THIS IS NEXT

(1. moves up one row)

0. THIS IS NEXT
1. THIS IS FINAL

(1. moves up one row)

0. THIS IS FINAL
1. (empty, not shown)

A word wrap function is needed.
Text should have a foreground
and a background color.

[8pla.net]

This is the AGI responding to emotional inflections in my voice...

This golden ratio is one of my all time A.I. favorites!

[Freddy]

Korrelan, I have no idea what it's doing but it looks cool 

[Korrelan]

Cheers chaps

The top right is a spectrograph of the vocal input. I take the microphone input and run it through a fourier series analasis which breaks it down into the individual frequencies over time. I also needed the power/ volume of the input, shown very top right.

Bottom right... The output from the spectrograph is fed into my neural model, this small section used for the test comprised of 6 sheets of 300 * 200 neurons so... 360K neurons ish.  Each neuron is synapsed to every other so... 129600000000 ish synapse.

I needed a simple face for expressions, so I adapted the 'golden ratio' face to a vector set. Then re-mapped to produce simple expressions and wrote the animation routines to smoothly move from one to another.

I trained the neural net on normal speech patterns using audio books, I then spoke to the system using various inflections in my voice, anger, joy and noted which areas of the neural net responded to the different inflections. I then simply used a blackboard technique to map the neurons for each inflection to the appropriate animation. 

The facial expressions will eventually run through the motor cortex to express how the system is feeling, this was just test to see if I could pick up a jist from the users voice.

I did the same thing with facial ques and a combination of both... kinda freaky when your talking to the AGI and it mirrors you...

[Zero]

Impressive, indeed. Is there a website or something where I can read about your system?

[Korrelan]

Hi Zero

Yeah! Millions of them... neuroscience and biology sites lol... Soz couldn't resist.

Only my YouTube page, some vids of the initial neural seed I 'grow' the systems from, audio visual maps from early development runs etc... I've been working on AI for over 30 years...

Here's a vid showing my initial tests on neuro-genesis and migration, you'll see neurons moving/ zipping through the cortex layer to find their ideal position/ location amongst other tuned neurons... to enhance resolution required in that area (part of the memory/ learning)... visual orientation map I believe focused on a pinwheel..



I'm a dreamer too... I'd like to think that one day my AGI's will explore other worlds, and as they scream through the atmosphere heading downwards toward a fresh new world to explore... if you listened very... very... carefully you would hear...

Weeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee!... lol

My take on AGI is.. just design a machine that is capable of learning anything... and then teach it to be human. 

[Zero]

You should train your baby so it can drive a DeltaGlider in Orbiter! 

Did I understand correctly? Are you handling 129.600.000.000 synapses real-time?

[8pla.net]

The top right is a spectrograph of the vocal input. I take the microphone input and run it through a Fourier series analysis which breaks it down into the individual frequencies over time. I also needed the power / volume of the input, shown very top right.

Bottom right... The output from the spectrograph is fed into my neural model, this small section used for the test comprised of 6 sheets of 300 * 200 neurons so... 360K neurons ish.  Each neuron is synapsed to every other so... 129600000000 ish synapse.

My theory, and it is just a theory, is that both these cases are doing the same thing.  If these are two ways to fill a curve (in theory), then that leaves us with the training.  This is so educational.  Thanks for your expert response.

[Korrelan]

Ok! It's Sat night here and I've been on the Rum & Malt so ignor any bad English or spelling mistakes... my brain works like... alcohol at night to brain storm... coffee in the morning to focus... and all that.

@Zero

Replace them full stops/ decimal points with commas and...

Nah! lol... miss calculation... 300 x 200 all synapsed = 60k x 60k x 6 layers = 21600000000 and... inter layer axons = 60k x 6 = 360k so... 21600360000 possibe synapse.

But... that's just a small section of cortex so way... way... frikken waaaay more... and yes in real time, though that's only possible synapses (but all are possible). It has to be an all to all connectome to start with, you only cull whats not used over time...

@8pla.net

Getting the frequencies laid out nice and neat is one thing... matching the trillions of possible combinations in real time is another, but yes... they are just simple patterns in a time frame.

Learning is a very simple concept... a system has to be told, or experience an event, it has to record the event in its 'own' terms so it can recall it, as 'it' (not you) experinced it, through 'its' own senses.

Then it just needs guidance... do I keep this knowledge and build upon it, or discard it.

I'm gonna read this in the morning and think... WTF was I rambling about lol. And one more... (just for good measure) the (...) is a really frikken weird habit I've acquired... meh!

Edit: My way is not the only way... It's just what I deem and have found to be correct... there is more than one way to fly.

[8pla.net]

@8pla.net

Getting the frequencies laid out nice and neat is one thing... matching the trillions of possible combinations in real time is another, but yes... they are just simple patterns in a time frame.

Learning is a very simple concept... a system has to be told, or experience an event, it has to record the event in its 'own' terms so it can recall it, as 'it' (not you) experinced it, through 'its' own senses.

Then it just needs guidance... do I keep this knowledge and build upon it, or discard it.

You are entitled to a few.  Rum beverages and mistakes, that is.   They go good together.  I made a nice mistake myself. Instead of "fill a curve" I meant to say "fit a curve", I think.  Oh, now I'm confused.  No worries, when morning comes, seems to me, your response will be just fine.

[ranch vermin]

I'll still be here.      Maybe ill have to admit my ai idea isnt that original or hard to come up with.       Just going through a phase.

[Don Patrick]

I bet it's hard to implement though. I don't place much value in theories unless I've at least made a working prototype of it. No matter how good the idea, someone still has to do years of work to make it happen, and quite often some overlooked details pop up to make it that much harder.

[ranch vermin]

definitely true my friend.

[Zero]

21600360000 possibe synapse.
But... that's just a small section of cortex so way... way... frikken waaaay more... and yes in real time, though that's only possible synapses (but all are possible). It has to be an all to all connectome to start with, you only cull whats not used over time...

Yeah ok, but that's still 21 billion real time, you're fkin fast!!! How do you do that? (some "don't think you are, know you are" trick?)

I'll still be here. Maybe ill have to admit my ai idea isnt that original or hard to come up with. Just going through a phase.

It's probably a very good idea. Trust mankind, share it!
I'm so happy you're staying here. 

[ranch vermin]

Hey Korrelan, I just had a peak at your videos.   Did you craft that animating vector portrait yourself?  Its pretty nice.

[EDIT] hahahha i saw you put me into a class on one of the videos hehe [/EDIT]