[8pla.net]

My animated, talking 3D avatar aired on CNN in primetime to like fifty million homes in the USA with cable TV, not to mention tens of millions more overseas. So, isn't it to be expected that many of us here in our research, have the latest versions of our A.I. that we don't release publicly?  In other words, don't we keep our best A.I. for ourselves?

[LOCKSUIT]

Stop talking salad into your mouth!

I'm actually on image 11 of the animation, and it's lookin good and ALL processes are shown, so you know when things start, end, and that there's "nothing" left that may show up in the actual "build" once built, therefore proving to the viewer I do have AI.

Art, give me your email,  maybe by PM. I can send you the 11 images to flip-through.

[DemonRaven]

I personally get turned off by someone who is aggressive and thinks they know all the answers. The truly wise person knows that  no one knows all the answers and also knows that aggression only begets aggression.  So I am sorry but the second i saw you attacking others for not buying into your grand ideas i shut down and was not interested in hearing any more.

Wisdom comes with age and experience and is not necessarily the same thing as being intelligent.

[Korrelan]

Hi BF33… I’m curious…

What kind of hardware would be required?
What methods do you envision for reading the senses into the system?
What sensors/ modality/ resolution would be required?
Would it be a parallel or serial system, connectionist/ heuristic/ etc...?
Would it run real time or simulated real time?
What language(s) do you envision creating this AGI with?
What’s the data structure & method for storing/ computing information?
What language/ structure/ method for generating the intelligence (kind of crucial)?
What pattern matching methods/ algorithms used?
What learning techniques will/ does it use?
Do you have working examples or a prototype to prove your ideas/ theories?

I could go on but you get the idea…

These are the types of questions we individually/ as a community strive to find the solutions too.

If I was claiming to have the total solution to AGI, and I was asking/ requiring people to follow/ join me… I would make sure I could easily answer these kinds of questions.

Flapping your arms won’t make you fly.

[squarebear]

Jeez! Are you guys still feeding this troll? Multi national companies like Google and Facebook can't create AGI, with resources of hundreds of people and budgets of millions, yet this clown comes along reckons he has created something groundbreaking while offering no proof and everyone is pandering to him....

[Art]

Why can you not just show us this AI of yours and be done with all the bantering?
Put your AI on a website and let the rest of us See and interact with it.

Squarebear has been online for many years entertaining thousands of visitors (if not more...I have no real numbers but it's a lot, I'm sure).

You will get some helpful feedback from visitors as well so quit with the pictures and promises and post your AI for us to experience.

This is just a friendly nudge and not an attack against you or your credibility, ok?

Good luck!

[ivan.moony]

P.S.
Take a few books, read on and you'll see what you lack in your theory. If you have enough knowledge, no one can stop you. A paper always beats the rock. And scissors cheat if you ask me

[LOCKSUIT]

---

[Art]

Your intro title says that:

You HAVE an AI.

You wish to show an animation of it working.

These are two distinct premises.

1st, Show us this AI, either working in a testing environment like an exchange of text statements / questions & answers or some other repeatable methods.

2nd, Animations can and are made by everyone from 8 to 80 everyday so that really doesn't lend itself toward proving very much.

3rd. It you DO have a working AI of which you are proud, you shouldn't hesitate to show it at least to us who some might regard as your peers. After witnessing your AI perform, you might wish to exchange some constrictive criticism and some positive feedback...that is...IF you do have an AI like it states in your title.

Consider this and make it available or simply bow out and save a lot of time and space.

Best to you!

[LOCKSUIT]

---

[8pla.net]

Art said, "Squarebear has been online for many years entertaining thousands of visitors (if not more...I have no real numbers but it's a lot, I'm sure)."

And, Squarebear's website aired on CNN during the Bruce and Sue Wilcox interview by Morgan Spurlock.  Not to mention, Squarebear's been on the BBC and other news outlets.

During my interview on Tech TV, Leo Laporte plugged my site.  Wow! It went offline with so many hits so quickly. The monthly bandwidth quota, which was generous to begin with, was exceeded in no time.  Whoah, but it was a lot of fun though, being on TV!

Reference:
http://www.cnn.com/TRANSCRIPTS/1501/22/inm.01.html

Hey BeautifulFlowers33, draw us a flowchart by hand.

[LOCKSUIT]

---

[ivan.moony]

Every time my robot generates actions that get it more food/ect (5 senses) better, it saves actions just done. This makes us!

Ok, now we got the first inforcement learning: to get some food (I assume electricity for batteries in our case). But then a robot wouldn't ever get away from electric recharger. I think we need more reinforcement feedback than this.

BUT HEY ALL, THIS MIGHT BE A VERY GOOD IDEA!!!

The only reinforcement that would be built in could be the one of a fear from being unplugged and running out of batteries.

Let's analyze: a criminal robot would risk to be turned off and chopped out for an old iron. The more good the robot does, the more probability is to be turned on forever. But this would arise a nasty bug: a robot could do something unacceptable, just to keep itself turned on. The robot should accept to be turned off if humans want to do it, but this is currently not the case with our only fear-from-death reinforcement treatment.

Does anyone have any idea how to avoid this nasty bug?

[DemonRaven]

They already started a reward system for machine learning:

http://www.sciencedaily.com/releases/2015/10/151001142152.htm

More-flexible machine learning
Giving machine-learning systems 'partial credit' during training improves image classification
Date:
October 1, 2015
Source:
Massachusetts Institute of Technology
Summary:
Researchers have presented a new way of doing machine learning that enables semantically related concepts to reinforce each other. So, for instance, an object-recognition algorithm would learn to weigh the co-occurrence of the classifications 'dog' and 'Chihuahua' more heavily than it would the co-occurrence of 'dog' and 'cat.'

[LOCKSUIT]

---