User avatar
johnb_summers
Posts: 285
Joined: Thu Aug 04, 2016 7:48 pm
Location: Bushey UK

Re: Running AI algorithms on Rpi

Sat Oct 01, 2016 7:52 pm

One of the little thing that no one notices in this is that man is a herd animal and almost every discovery starts with one human interacting with another, the wheel was only invented because one person wanted to take something to another person, so how much would man know if he was only one immortal being, alone in the universe, would he just not hang out and do nothing, would he be a moron, so I think intelligence is a product of interaction with others and not something developed in man alone, so is a bug just machine like because it don't have anyone to talk to, and if we ever do manage to get a machine to do all the talking and work for us will we turn into bugs over time as we no longer need interaction. I think there is only progress in this because one man says to another we need this, once all these things are done then what. When we are talking about AI for the computer we think in singular terms, like the computer is there to help me, but we forget that we only developed because there was more than one of us so cant be thought of in a singular way, to me AI would be 2 computers teaching each other and changing their whole being on that and that alone, man would soon disappear from the picture, like do we need gods anymore?
MyPi Developer
http://mypiworld.com/
http://mypi.tech/

https://www.youtube.com/watch?v=US2nyRgg-SY&nohtml5=False

Heater
Posts: 13303
Joined: Tue Jul 17, 2012 3:02 pm

Re: Running AI algorithms on Rpi

Sat Oct 01, 2016 8:54 pm

Tzarls,

I was only using awareness of pain as an example.

Let's say you lost the feeling of pain. Went blind and deaf. Lost all sensory inputs. It's not uncommon, it's what happens when we sleep, or are deep in thought some times.

One can still be aware of ones own thoughts, of ones very existence. "Cogito ergo sum" as Descartes said.

Like I said, I feel like I have that awareness, that consciousness. I cannot prove conclusively that anyone or anything else in the world does. Maybe they are just mechanisms that behave as if they do. If they exist at all !

Of course, I see all those humans running around me are somewhat similar to myself, so I assume there is something going on inside each one of them that is similar to what goes on inside me.

Heater
Posts: 13303
Joined: Tue Jul 17, 2012 3:02 pm

Re: Running AI algorithms on Rpi

Sat Oct 01, 2016 9:16 pm

johnb_summers,
One of the little thing that no one notices in this is that man is a herd animal...
That is an interesting point and one that has been bugging me my whole life.

We depend on our society. I speak English because I was raised in England. I have all kind of opinions about all kind of things because I was raised in that society. I can even think in mathematical ways to some extent because of the education that society gave me. And so on.

As far as I can tell I can't think without a language. English in my case.

How would it be if somehow I was not in any such society since birth but some how managed to survive? I would have no language. I would have no maths. No education as we know it. Would I even be able to think at all?

Or would my mind make up it's own language so that it can talk to itself and thus bring "me" into existence?

ejolson
Posts: 3555
Joined: Tue Mar 18, 2014 11:47 am

Re: Running AI algorithms on Rpi

Sat Oct 01, 2016 9:58 pm

Heater wrote:"Cogito ergo sum" as Descartes said.
Philosophers, whether ancient or modern, seem to make a business of sounding profound by taking well known facts and saying them backwards. While the saying "I think therefore I am" has created great publicity for Descartes, it is merely saying the obvious "I am therefore I think" backwards. Not only does this quotable statement shed no meaningful light on what it means to be human, but worse, it leads to the implausible misconception that the ability to think is what makes something human.

In my previous post I mentioned that mankind is distinguished by moral responsibility. This does not mean people are required to act morally. Rather, it assumes each person has a free will and then holds each person accountable for the morality of their actions. This moral responsibility is what distinguishes humans from dogs, horses, corporations, governments and even intelligent machines.

Getting back to intelligent machines, it is entirely plausible that neural networks will eventually perform thinking tasks that were previously the exclusive domain of people. For example, the ability to diagnose medical illnesses is one area where machine intelligence already surpasses most doctors. At the same time, doctors remain relevant by shouldering responsibility for the morality of the resulting medical procedures.

Heater
Posts: 13303
Joined: Tue Jul 17, 2012 3:02 pm

Re: Running AI algorithms on Rpi

Sat Oct 01, 2016 11:03 pm

ejolson,

Oh boy. There have been many objections to "I think, therefore I am" but I have never heard one like yours before.

I don't understand it. How one can say "I am therefore I think" ?

I mean, one cannot even get to the "I" part of that statement without thinking first. One needs to have self-awareness, some kind of thinking, to even get as far as an "I" concept.

Or I could say that your "I am therefore I think" implies that all things that exist are thinking. As far as I know that is not true.

You are right, "I think, therefore I am" says nothing about being human. It's only a statement from one self-aware, thinking thing. Happened to be a human called Descartes but that is not the point.

I don't see how Descartes statement leads to "...the implausible misconception that the ability to think is what makes something human."

Certainly there are humans who cannot think in any recognizable way. Conversely I'm prepared to believe the universe is full of thinking beings. Who may well have discussions similar to this one.

Oh boy, "free will".

For sure we should not go there. What is "free will"?

It feels to me like I have choices. I love to eat carrots but I cannot stand beetroot. So I choose carrots of my own free will. But really, I did not choose that I should like carrots and not beetroot. Where is my free will there?

Perhaps a thought comes to me that is the solution to some problem or other. Perhaps it never comes. Can I choose if it comes or not? Where is my free will there?

Perhaps I remember something and perhaps I do not? Did I choose to remember some things but not others? Where is my free will there?

This all makes me think free will is an illusion.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 12:04 am

AI is not Artificial Consciousness.

Or is it?
https://en.wikipedia.org/wiki/Consciousness
http://www.nytimes.com/2016/09/20/scien ... .html?_r=0
http://fortune.com/2015/10/16/how-tesla ... ot-learns/

If a Telsa car is aware of things around it, is it conscious or intelligent.
Years ago we got milk delivered by horse drawn cart, the horse was trained and knew were to go. We are now at a horse level of intelligence in mass produced products.

Rodney Brooks and other showed decades ago how simple responses to stimulation could emulate insect level behavior.
A Roomba robot has Ant level intelligence for path finding.
A Pi has the power to fit somewhere in between these two.
Perhaps bird level, Pi parrots?

AI has limitations at the moment due to training, basically they are as smart as their trainers. How to think outside of their trained responses?
Let's find out :P
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

ejolson
Posts: 3555
Joined: Tue Mar 18, 2014 11:47 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 12:24 am

Heater wrote:It feels to me like I have choices. I love to eat carrots but I cannot stand beetroot. So I choose carrots of my own free will. But really, I did not choose that I should like carrots and not beetroot. Where is my free will there?
Free will is not the ability to choose what you like or what you dislike, it is the ability to choose what you do. A rabbit is not charged with the moral responsibility of deciding whether to eat a certain carrot or not. On the other hand a person is responsible for such a decision. It is free will that makes it possible for a person who likes carrots to avoid stealing them whenever the grocer turns his back. It also free will that makes a person who steals carrots no less responsible for stealing than a person who steals but does not like carrots. At any rate, although I thought I was parroting someone else's ideas on "Cogito ergo sum," it appears (perhaps unfortunately) that my remarks were original.
Last edited by ejolson on Sun Oct 02, 2016 12:37 am, edited 2 times in total.

User avatar
Douglas6
Posts: 4760
Joined: Sat Mar 16, 2013 5:34 am
Location: Chicago, IL

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 12:34 am

This forum is an excellent place to learn about the Raspberry Pi. Philosophy, not so much.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 1:34 am

This forum is an excellent place to learn about the Raspberry Pi. Philosophy, not so much.
Philosophy is still important as future coders of AI may start here.

One of my favorite subjects in my Engineering Degree was Ergonomics.
Man/Machine relationships. Up until then I thought machine are just machines and men just used them as they came. From then on design became important to me.

Likewise I grew up with a diet of SiFi, Asimov's Robot Laws still resonant. When I first read about Rodney brook's bots, I knew there was a reason I was so disappointed when I had to give up Biology to go Phys/Chem.

Do the kids reading this today have this background, where is the youtube version of the Three Laws? My kid learns to Minecraft by watching youtubes.
When you start talking AI then you need to consider all the ologies as well.

Do we need to consider Philosophy when doing AI?
I am pretty sure it is one of the things that separate us from the animals.
What language do you code it in?
Or it is just all an artifact of consciousness?
"I think therefore I am".

Do we give robots religion and make us their creators?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Douglas6
Posts: 4760
Joined: Sat Mar 16, 2013 5:34 am
Location: Chicago, IL

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 2:16 am

I was a huge Asimov fan myself, but that's sci-fi, not philosophy. Since then I've read Plato, Descartes, Locke, Frege, Kripke, Lewis, Chalmers, Dennett, Putnam, Quine and others, influential thinkers on the topic at hand. The subject is deep and subtle, and does not lend itself to YouTube.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 3:16 am

I have found most Philosophers theorems are coloured by the times they lived in. SiFi is predicting the times we could live in.
Truth is just a judgment based of facts, facts change as we learn more, therefore the truth can change.

But that is all beside the point.
How do we code AI on a Pi?
What is possible to code and how do we do it?

Subsumption architecture has shown emergent behavior that was NOT programmed. This was done on a MC68HC11 bit microcontroller.
Should be easy on a Zero.
What can we do with Quad core 1.2GHz machines?

Current AI is based on Deep learning, can we do this on a Pi?
Does it need to connect to the Cloud?

https://petewarden.com/2014/06/09/deep- ... pberry-pi/
https://www.raspberrypi.org/blog/now-added-cucumbers/
viewtopic.php?p=786040

Then there is what do we want the AI to do?
Automate your model train layout? Companion chat/health bot for the elderly?
Mow your lawn, wipe/mop/vacuum your floors, clean your pool?

Is there a programming language that is better for Subsumption, Fuzzy, Neural Networks? Do we need to use Robot OS? http://www.ros.org/
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Douglas6
Posts: 4760
Joined: Sat Mar 16, 2013 5:34 am
Location: Chicago, IL

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 3:30 am

Gavinmc42 wrote:Truth is just a judgment based of facts, facts change as we learn more, therefore the truth can change.
Again, a comment based on a very superficial understanding of the philosophy of truth.

Let's stick with what we know about the Raspberry Pi.

ejolson
Posts: 3555
Joined: Tue Mar 18, 2014 11:47 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 3:37 am

Douglas6 wrote:The subject is deep and subtle, and does not lend itself to YouTube.
While philosophy may be deep and subtle, it also focuses on questions universally important to every person. While some programmers interested in machine learning, autonomous vehicles and neural networks may not be interested in thinking about such things, it may be the case that every person at some point searches for meaning, truth and wisdom. It is interesting how the subject of creating intelligent machines so quickly leads to general questions of a nontechnical nature.

Returning to some technical things, I wonder if QPULib could be used to implement a neural network on the VC4 GPU and if so of what complexity?

By the way, thank you for collecting those links to Pi-based machine leaning projects.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 7:03 am

Again, a comment based on a very superficial understanding of the philosophy of truth.
Perceptions of reality are individually based, the World is 7.5 Billion Worlds.

"AI" is a now a over used generic term, AGI is used as General Intelligence.
Will "AI" emerge automatically?
https://www.youtube.com/watch?v=iNL5-0_T1D0

Mark Tilden's Beam robots have very simple "neurons" yet can walk.
https://en.wikipedia.org/wiki/BEAM_robotics
These are now mass produced
https://en.wikipedia.org/wiki/Hexbug

Can a Pi emulate a simple NV network in software?
If a robot learns to walk is it smart, intelligent, aware?
At what point do you say it is AI? And does it matter?

Can a robot learn to talk?
Once the first one speaks, can it train others?

Flocking bots have been in software since the 1980's, just now real in the last few years.
https://www.wired.com/2011/09/flocking-robots/
http://iq.intel.com/100-flying-robots-s ... gulations/

Collision avoidance
http://www.intel.com/content/www/us/en/ ... rview.html
Intel acquired the drone company.

Now real swarms of 1000's of bots
https://en.wikipedia.org/wiki/Swarm_robotics
http://www.eecs.harvard.edu/ssr/project ... .html#make
https://www.youtube.com/watch?v=dDsmbwOrHJs

The first mass produced Pi based bot, what will it be?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Heater
Posts: 13303
Joined: Tue Jul 17, 2012 3:02 pm

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 8:02 am

ejolson,
Free will is not the ability to choose what you like or what you dislike, it is the ability to choose what you do...
That is just saying that there are more reasons for choosing a thing than a simple like or dislike of the options. I might choose something I don't like so much over something I do like for many reasons. Perhaps what I like gives me an allergic reaction, perhaps it's too expensive and it's better to get more of the less liked option. Perhaps I know my friend likes it too and I want to save it for them. Or, as you say, that carrot is not mine and I have some idea of right and wrong so I don't steal it and eat it.

I'm going to argue as follows:

In order to choose between action A and action B it is necessary that you know that both A and B even exist as options. If not you have no choice.

How do you get to know of A and B? Well perhaps somebody told you. Or you just learned it somehow. But what if you didn't? Then you have no choice.

What if you knew of A and B but forgot one of them? Then you have no choice. You cannot choose to remember or not remember things.

Perhaps you arrived at A or B by some kind of logical reasoning. Using your intelligence to deduce that they are options from knowledge of some other facts. But what if you are not so intelligent and can never make that conclusion? Then you have no choice. One can not choose how intelligent one is.

What if a course of action just comes to your mind out of the blue? Quite a few times I have woken in the night and found some idea in my mind that seemed to come from nowhere, perhaps the solution to some problem or other. Well, such ideas can come or they may not. You cannot choose for them to happen or not. Then you have no choice.

When you mention stealing you are bringing in ideas of right and wrong. That is to say morality. I don't want to go there except to say morality is a hugely fuzzy and ever changing thing. For example not long ago it was morally acceptable to burn homosexuals or keep slaves. Looks to me like we don't get to choose how our moral compass works.

Heater
Posts: 13303
Joined: Tue Jul 17, 2012 3:02 pm

Re: Running AI algorithms on Rpi

Sun Oct 02, 2016 8:35 am

Gavinmc42,
If a Tesla car is aware of things around it, is it conscious or intelligent.
If one claims that a Tesla, or any other computer system, is conscious then I'm going to claim that the light switch on the wall in my lounge is concsious.

Consider the humble light switch, what a brilliantly intelligent thing it is...

My light switch understands a language. That language has symbols, I can press it up or I can press it down.

That language has a grammar. Pressing up then down, or vice versa, is valid in the grammar but up,up or down,down is not.

That light switch language has meaning. Which is what makes the definitions of valid grammar. Up means turn the light on, down means turn the light off.

Of course in order for a language to have "meaning" it requires something to understand that meaning, that is to say it requires intelligence. My light switch seems to understand the light switch language so clearly we can already say it is intelligent.

That is impressive enough, I can talk with my light switch and it does what I say, but there is more.

My light switch has a memory. When I go away the light switch remembers what I last told it and keeps the light on or off for me. It knows.

So there we have it, my light switch has all the traits of an aware, intelligent thing. It's aware of my inputs, it understands what inputs make sense and what don't. It remembers what I told it.

Now, that Tesla has a computer made of billions of switches. Those switches are very much like my light switch. So the Tesla is at least as intelligent as a light switch. Which is already pretty impressive.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Tue Oct 04, 2016 1:47 am

Consider the humble light switch, what a brilliantly intelligent thing it is...
Remove tongue from cheek?
Ok, maybe if you add a PIR based movement sensor it becomes a bit smarter.
Not smart enough to know what is human and what is next doors cat.
So then you add a PIR array sensor, just a bit smarter?
Next version detects your smartphone or personal iBeacon and changes the lighting/aircon/water temp to your preferred setting.
Is it very smart or just a better bit of coding.

What if you add a camera and face recog code?
At what point does a smart thing show emergent behavior that anyone (with a few neurons still working :lol: ) would regard as intelligent?
And how do we do it on a Pi?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Douglas6
Posts: 4760
Joined: Sat Mar 16, 2013 5:34 am
Location: Chicago, IL

Re: Running AI algorithms on Rpi

Tue Oct 04, 2016 1:59 am

I don't think any of my light switches, even augmented with PIRs and cameras, would pass a Turing test. And, to my mind, that's a pretty easy test. Even ELIZA (runs on a Pi, I'm sure), could fool the denser among us.

Graymalk
Posts: 55
Joined: Wed Nov 11, 2015 8:33 pm

Re: Running AI algorithms on Rpi

Tue Oct 04, 2016 6:47 am

I'm studying machine learning right now and Pis are too small and slow for training. You can certainly run trained models on them though. But training is nasty slow especially with big, high dimensional data.

That said though, I bet it'll run decently on small data. Could try one of the C++ machine learning libraries (eg mlpack) if they compile (which they might not). Alternatively, there's weka, which is a java app that has both GUI and command line. It has nasty resource requirements though. I imagine it will run but you'd have to stick with really small datasets, which would be boring. There's a JVM for Pi, right?

There are also the python and R machine learning libraries but I've actually tried....one of the R ones I think. I've actually forgotten. It failed. Couldn't compile one of the C++ math libraries it required (one of the many BLAS libraries and offshoots). I was going to try to fix it but I just don't have time. It was usable - just couldn't use some of the optimizations, which just made it unbearably annoying.

Ok back to your philosophy stuff.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Tue Oct 04, 2016 8:16 am

Eliza? have not heard that for a very long time.
With the decrease in IQ in the last few decades, I'm sure it would fool more people now :lol:
https://en.wikipedia.org/wiki/ELIZA
http://www.jezuk.co.uk/cgi-bin/view/software/eliza
http://www.masswerk.at/elizabot/
Hmm texted based, I had forgotten that.

How to make a voice recog version?
My kid uses his tablet to get it to find youtube videos.
Is there an offline version we can use? Or do we need to go to the cloud?

Came across this today.
https://www.invensense.com/products/ics-52000/
Currently PDM mode on the Pi only allows 2 digital microphones.
With 4,6,8,16 mics could do tracking, better filtering etc.

Text to speech is relatively easy.
Came across this little thing, i2s audio amp.
https://www.adafruit.com/products/3006
Not sure if we can do i2s audio in plus i2s audio out on a Pi, have not used it before.
There are also the python and R machine learning libraries but I've actually tried....one of the R ones I think.
Don't know much about machine learning, did my Uni robotics back in the 1980's, 1MHz 8 bit micros :lol: .
R I know works on Pi, was looking at it for the big data I collect on IoT gadgets.
JVM runs, comes in Raspbian and we have Wolfram/Mathematica too.
To get reasonable speed need to look at NEON/SIMD coding?

Microsoft acquires R language experts, hmm wonder why ;)
http://www.computerworld.com/article/28 ... ytics.html

Pass Turing test? Well maybe not, don't now unless we try ;)
I would have said this was not possible either,
https://www.raspberrypi.org/blog/alpha- ... ement-day/
http://magazine.uc.edu/editors_picks/re ... alpha.html

Alpha was deep learned on a super computer then ported to Pi?
http://www.omicsgroup.org/journals/gene ... 000144.pdf
Maybe not, GFT - Genetic Fuzzy Tree, cross between Neural nets and Fuzzy logic?

Ok, Alpha is trained by Eve.
http://www.psibernetix.com/projects/defense/

What algorithms can run on Pi's.
Let's make a list and start testing.
Anyway it might make better AI's for games on Pi's possible.
Or talking companion bots.

With the cost of Zero's, adding a few extra for vision, audio preprocessors etc is not expensive.
Can plug 4 Zero's into a Pi3 :P
We don't have to do everything on a single Pi.
Maybe make a super cluster of Pi's in a cupboard?

There are 10 Million Pi's out there now, hook them all up, would they pass a Turing test?
Emergence it what we are after, throw some water and chemicals in a flask, add lightening and life emerges eventually.

If you are a combat pilot and get shot down by a drone controlled by a Pi, do you get pissed off that it fails a Turing test?
Chinese hackers can remotely play with Tesla model S cars.
It is a very different world now than last year.
Do we learn how this works or stick our heads in the sand?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Graymalk
Posts: 55
Joined: Wed Nov 11, 2015 8:33 pm

Re: Running AI algorithms on Rpi

Tue Oct 04, 2016 2:08 pm

Gavinmc42 wrote: Don't know much about machine learning, did my Uni robotics back in the 1980's, 1MHz 8 bit micros :lol: .
R I know works on Pi, was looking at it for the big data I collect on IoT gadgets.
Base R works. The problem arises when attempting to install certain packages. The fun part is how many coffees you get to grab while waiting for it to compile before it crashes. ;-)

Python might work better. I haven't tried it.
JVM runs, comes in Raspbian and we have tWolfram/Mathematica too.
To get reasonable speed need to look at NEON/SIMD coding?
Well I'm personally a big fan of SIMD coding although I don't know NEON. I get 1024 threads out of my laptop's GPU. I would need 256 of my laptop's CPUs to match that (direct comparions aren't totally fair but I'm being general here lol).

But once the training has been done, *running* a model is fast. It's just like running any other program. But that's only if you aren't continuously training. You can set some models up to retrain after they acquire more data. This is typically something you want to do in the cloud. I suspect things like Siri run this way but I don't work for them so I don't know for sure.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Wed Oct 05, 2016 2:38 am

Had forgotten about this
https://www.tensorflow.org/mobile.html
There is a Pi version, Python based so it would be slow?

Might be a good place to start.
I think learning systems are designed to get better and faster, to point it can be done on Pi's.
This would be for single purpose type applications, like how Alpha was taught by Eve.
Continuous learning would be slow?
Write a learning program to learn faster :lol:

Wonder if I could train a robot to pick up Lego's when told too, probably would be simpler than getting the kid to do it ;)
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
morphy_richards
Posts: 1603
Joined: Mon Mar 05, 2012 3:26 pm
Location: Epping Forest
Contact: Website

Re: Running AI algorithms on Rpi

Wed Oct 05, 2016 7:10 am

Heater wrote:Which ever way you slice it, neural nets, fuzzy logic, what ever AI algorithms we have now a days. No matter what hardware it runs on, GPU's or Google cloud farms etc. It all comes down to the same simple thing:

A computer, with memory, inputs and outputs.

The question then is: Can we build an "intelligent" thing out of that with some algorithm or other?

A satisfying, to me at the moment anyway, would be "No".

A disturbing answer, to many, would be "Yes". Disturbing because it would prove that we are nothing more than machines.

Anyway, whatever you call it, it's amazing what the compute power and memory available today can do.
Plus 1.

Brain as processor analogies are often used because they are convenient perhaps to the extent that this has become part of collective human consciousness. The brain is just a computer processor...

The number of transistors in a vast, parallel computer system is now approaching the number of connections between neurons in a human brain therefore we are close to being able to make a brain of our own.

But (I think) this is very far from the truth in reality.

As Heater points out computers are all based on Von Neumann architecture. Brains are not. There is no distinction in a brain between data, software and hardware. Brains are analogue. At some point years back I read about supposed carbon nanotubes in there too which may suggest they also work on some quantum level (this may have been debunked now, as I say it was some time ago)

AI definitely exists. Make a program to play noughts and crosses with if statements and there you have it in a simple form.
Top down vs bottom up ....

The AI that's in films like Ex Machina may be a wild goose chase while computers are in their current form. I really do believe "AI" will come about but it will be alien to us. Very much unlike our minds and the way we think.

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Thu Oct 06, 2016 3:53 am

Was waiting for wife doing the shopping, grabbed PC game mag to flick through.
Had a bit on No Man's Sky.
Just about everything is Procedurally Generated, worlds, animals, behavior, like the original Elite game.
https://en.wikipedia.org/wiki/No_Man%27s_Sky

Could PG be applied to generating training solutions that then get run through something like Tensorflow to find optimum answer?
There was a rumour next gen Elite was coming to Pi's.

Stick on a VR headset and go into one of these world's, how long before you cannot tell the difference between an AI character and another player? It took three decades before flocking bots on a PC became real flocking drones.
AI is now everywhere, which is why AGI is now the preferred term (Artificial General Intelligence).
https://en.wikipedia.org/wiki/Artificia ... telligence

Some sort of AI could be done on PI's :P , AGI maybe not ;)
Will a computer ever become conscious? Never, not ever.
Now that I have predicted it will never happen, someone will prove me wrong.
Hands up who watches "Person of Interest"? ;)
How close are we? Just need a cluster of 500 old game machines?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3730
Joined: Wed Aug 28, 2013 3:31 am

Re: Running AI algorithms on Rpi

Wed Oct 12, 2016 3:04 am

Was looking at Mathematica for multispectral image processing when I noticed the new version 11 stuff.
http://www.wolfram.com/mathematica/new- ... age-stripe.

Neural Networks and Machine Learning.
Will version 11 come this on a Pi?

Interesting stuff
http://www.wolfram.com/language/11/imag ... athematica
http://www.wolfram.com/language/11/imag ... athematica

Have not look at Mathematica for a while, it is much faster on a Pi3.
Wonder if it uses NEON/SIMD stuff on the ARM A53 cores?

Raspbian 4.8 64bit and Mathematica 11, could be very interesting.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Return to “General discussion”