Intellegence and Machines

Combine intellegence and machines and you get a forum about intellegent machines. Intellgent machines are machines with brains.

I think brains are machines. I think brains are interesting machines. I quit reading science fiction when I started building brains for machines.

Can we get back to the subject?

 

Log in or Sign up to hide all adverts.

i like brains too

is there something you'd like to discuss?

 

Log in or Sign up to hide all adverts.

Anything anyone is doing with brains, robots, animated machines, whatever.
Start out with questions:
What is a brain? Where do brains start? Are there brains below neutrons or does a neutron have a brain? How does one describe a brain?
Pick any one to discuss.

Harold

 

Log in or Sign up to hide all adverts.

Emergent behavior...

I've always held the view that neurons aren't really brains in themselves. I believe that neurons are simple machines that just react to the chemical signals coming in and send out chemical signals to other neurons. This seems like a simple design, much like our electronics. However, because there are SO many neurons working in this way they exibit behavior that is what we call intelligence. I believe it is much like the emergent behavior that we see in small creatures. Using simple rules that are in no way intelligent, creatures like fish and birds can seem to act as a single organism (schools and flocks). I believe that there is nothing inherent about the biological brain that prevents it from being emulated or copies in silicon parts. I believe a study of emergent behavior can also help in this research.

-AntonK

 

If I knew how the brain worked I'd be a millionaire. I don't think it's impossible to emulate human intelligence, but It's going to be really, really hard since we don't know how it works ourselves. I've often thought about various aspects of my own brain functions and how they could be duplicated in a digital environment.

How does the brain store information about objects. when I look at a pen of a different shape, size, or color than I've ever seen before, what does my brain do to identify it as a pen? How does my brain organize information? When I react to a statement what process does my brain go through in order to form a response?

 

Bottom-up or Top-down?

I think duplicating brain function in a digital environment comes down to whether you are of the camp that believes in top-down design or bottom-up design. Brains (I believe) don't really have data types for storage as we think of them. I think the bottom-up approach is better. It seems to me that the brain functions we have may just be emergent behaviors of a sufficiently large (yet simple) network of simple ruled neurons - as opposed to a large network of sophisticated computational algorithmic and data storage.

-AntonK

 

All good.
Emergent behavior is considered.
Object detection is only one of the problems.
Both top down and bottom up are needed.

Brains are of different sizes. By observation.
Then brains can be sorted by size.
Then there must be a least brain.

Brain is between sensors and motors. By observation.
Brains are made of neurons. Common knowledge.
Neurons have many inputs and a single output.
That means neurons make trees. Every synapse represents another neuron.

From this we can determine that brains must mediate between sensors and motors. Also, brain must be a list of sensors, neurons and motors. If we have a list of neurons and connections, we have the instructions for building.

Back to the least brain.
Let it be 0. Then we have sensors connected to motors.
You can make machines of those easily.

The least brain is one that mediates two sensors to one motor.
So instead of building a large brain, all we have to do is to build a simple two input, one output function.

What is the simplest , multi bit two input, one output function you can think of?

Harold


.

 

Well there are multiple 2 input 1 output functions. There are 16 actuually... AND, OR, NAND, NOR, XOR, etc.

Where does this take us?

-AntonK

 

brain and cognitive sciences

I work from the Top-down perspective. I'm not saying it's any better than bottum-up. I know people on both perspectives who tend to ignore the other side, which i find a very stupid thing to do. We need poeple to consider the problem from different angles if a lot of progress is going to be made.

kinda why i'm trying to learn more neurobio. I don't think the whole "brain" problem will be solved simply by thinking about the brain as a bunch of inputs. just cause i know a building is made of bricks don't mean i know how to build it. but it is good to know the "lower" brain stuff.

bottom line is there are lots of unknowns and problems. one approach isn't going to solve all of them.

 

My only problem wiht the top-down approach is that it seems to go against the way the brain was developed in the first place. It wasn't put together mimicking something that already existed. It started as Harold would put it as a minimum brain, simple neurons connecting inputs and muscles (btw...does anyone know what creature had the fewest number of what we today still recognize as neurons?)

It seems to me that we should try to develop both a brain and intelligence the same way nature did and that is from the bottom-up. However, there are tons of things we've seen that we can do better than nature (if not better than surely more efficient), so perhaps top-down does have possibilities.

-AntonK

 

37! (in a row?)

i guess it depends on if you're trying to design a brain from scratch, or understand how ours works. even if you are trying to design one from scratch, knowing how the current model works will help, and top down is equally important for that.

oh, and to answer you question, the lowerst i've heard is about 37 or so. some sort of flatworm. they know exactly what each one does. very cool stuff.

Please Register or Log in to view the hidden image!

 

How does a neuron "input" work? Is it binary, or is it something else? Can it only distinguish between "no signal" and "signal"? Or can it distinguish between various levels of "signal", as in different voltage or different density or ammount of whatever chemical signals neurons use?

I'm asking because I have no idea how neurons work, I'll need to read something on the subject.

 

I did hear about that flatworm. I believe its a form of nematode (but im wrong a lot). I'm sure there must have been some creature before it though that had, say... only 36. I'm curious as to how the neuron developed and what its function was before it became what it is now.

-AntonK

PS - I dont have time to find it right now but I believe a group actually build that worm brain into a small robot, they replaced the stimulus with light and the darn thing actually worked!

 

interesting question.
consider me looking for answers.

in other words, all ask my neurobio freinds

 

C. elegans is the worm you are talking about. It has over 300 neurons. If you count the functions, you can determine the neuron tree fanout. It is 2. That is, the average neuron is connected to 2 other neurons in this worm. And, yes, you can make the hardware for this. In fact you can make the hardware for any brain. We have the tools. It is know how that is lacking.

Neurons grew from general cells. It first connects by sending out fibers that find a synapse. If there is no synapse for it it dies.

The synapse connection is messy, to say the least. A neuron has both positive and negative inputs. The information passed from neuron to neuron must be an integer. Nothing else will work. So there is no Boolean logic here. It is all arithmetic.

To build a neuron, put a computer between every two synapses and a computer between every two computers to the final computer which is the axon. Then you have a model neuron.
That is the order of complexity here. Does not matter. We have $5.00 computers now.

Every thing must be done top to bottom to top.

 

What have you built so far Harold? May we see any of it?

-AntonK

 

Sure.
I have the algebra done that matches to hardware nodes I have made. These are connected with cables that plug in, so you can build and take apart at will.
I have about 20 different nodes now, from CNode which is a comparison, subtractor node to done which emits a number when the derivative of what it is watching is zero.
Nodes combine into agents a la Marvin Minsky, who introduced me to AI in 1959. Gave up AI in 1994 when I found the least brain.
My web site has a bunch of papers that you can have for asking.
I also have kits to play with.
Biggest project is Big Al which is a human sized animachine that can sit, stand, walk, etc. Big Al has pain, touch and hearing. Vision is primative so far. Most of Big Al is first level stuff which is reflex. It knows how to sit, but not how to find a chair.
The design scheme is bottom up. You start with any motor, define its actions. Then add nodes to go and not go until you run out of ideas and you can stop with a constant.
These structures build trees, so adding additional actions just adds to the group.
Next level is to identify objects and their properties which make comcepts. Eventually, the name property is added. I think nouns come first, then verbs, then adjectives. I think there is an arithmetic of language but it is hard to dig out.
A is some B is near to A = some + B. So is, was, has, had are easy. Generally a motor is Gerund = verb - not verb.
Then fill out verb = x, not verb = y and so on until you define everthing.
I am working on a better book for all this.

What is left to discover is top level of the brain and language.
I assume that I know nothing until I can test it on the bench. So there is not much out there to help anymore.

What this algebra does is to allow one to make experimental brains. If your algebra works, the hardware will work. In fact any structure you make will do something, even if you make it random.

That is how brains develop in animals. Each neuron finds another from its location. The parts that don't work are pruned.

This is going on too long for the forum. Ask me for papers.
When you get the algebra, we can swap designs.

Harold

 

Harold,

May i ask what your first language is? I am intrigued by your ideas but I feel sometimes there is a language barrier.

-AntonK

 

AntonK,
Don't know what you want. My first language is hill billy English.
Other language comes from mathematics and the computer.
I do tend to ramble sometimes on the forum where things come out as they come.

My basic assumption is that the universe and everything in it is a mathematical machine and that anything can be encoded as a number. If that could be proved wrong, I would join Pythagoras.

What is your language of choice?
If you have a question, be specific or quote it.
I want to spread this, since time is flying. If it is a matter of language, I can adapt.

Bye for today

Harold

 

hlreed,

what sort of things have your HAL brains actually done?

Do you get alot of business on that site, and what sort of things do people ask for?