Let's Make Robots!

Artificial Intelligence Questionaire

Hi Guys,

I'm new to robots etc. I am currently doing a school project on artificial intelligence and wanted to do some research. I thought why not go talk to the experts so here I am. I have already done some background research and already know the general definition but part of the project requires me to gain information from a range of sources. This is a kind of questionaire/interview to gain new sources from experts.

Could you please explain in your own way "What is Artificial Intelligence?" and "How can you prove when a robot has AI?". Please make the answers as detailed as possible.

My chosen project is to design and create a robot (i.e line follower) and discuss whether or not it posseses AI. 

Also could you suggest a robot ( line follower or maze solver) which would be closest to AI in order for me to raise a good discussion.

Thanks for the help!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

I do not need one definite answer, i need different views

 

What is AI?


   AI or Artificial Intelligence with respect to robots means that the robot will learn from previous behavior with some manner of positive or negative reinforcement. To contrast this with many of the robots found on LMR, if you look at the programming that allows them to function, you will likely notice pattern recognition in the program, or, level based conditional statements, such as, dangerlevel = 80 for example.

 

How can you prove when a robot has AI?


   Based on my personal description of AI above, AI should be visible based on improved iterations of a task without outside interference. In the case of your line follower, if you program it to read its sensors and react to specific patterns, it does not contain AI. If, however, your robot is programmed to reinforce beneficial behavior when following a line, I would then say it has AI. I guess, based on my previous statements, you could simply look at the program and see how it is organized to know if it is programmed to be simply autonomous, or, programmed to have AI.

 

there was a test made up by some scientist a long time ago which stated that an AI should be
considered if it could have a conversation with someone via text and maintain the illusion to the
conversational partner of being human.

as it turns out even very crappy AI's can fool people on irc and such places.

anyways i might define AI as something that can respond to dynamic and changing situations
without getting stuck or becoming obviously reactive.
'a machine that appears to make decisions'

that is what I think about AI, I am french, my english can be not perfect, but my differente culture can see the AI diferently :

I think we can say a robot has an AI when the robot acting a non-programmed action against an element of his environment.

All other posibilities of action, re-action, or environmental is only the result of programming.

As for any philosophical questions, wording is the key.

Artificial : do we mean fake (good enough to be cheated) or synthetic (like ours but not from carbon life form). My choice is synthetic. I will not call AI a chess program.

Intelligence : what is intelligence? The meaning goes from calculating fast (as in IQ) to understanding your neighbour (as in CIA).

I may call intelligent a group of plants that communicate chemically and become poisonous when one of them is being eaten.

But once you understand the mechanic behind these observation, you cannot really call that intelligence.

The word I would really think about is consciousness. Consciousness is that shadowy, unwordable (what is the word Lovecraft used?) thing that links all the inputs to the outputs. Science (not freudian psycho crooks) shows that unconscious processes play a big part in the final response or even in the desire.

I am convinced we will create consciousnes (and unconsciousness) on something else than carbon. It will know thirst and hunger, though it will call it battery level. It may needs words like love and hate even towards other AI, but its definitions will be very different from ours. In fact, we may be closer (emotionnally) to any carbon lifeform in the universe than to our AI.

It is a very interesting question anyway. But it touches so much on the pride and fear of human against machine, with some religious belief to help everybody keep reasonable (Did you notice the sarcasm? I'd appreciate you did!), it is hard to expect a thorough, scientific answer.

 

 

 

Besides all AI definitions out there here s mine:

I guess there are different levels of AI, simple, medium and advanced.

The simple stage is learning by experience. Here is an example, the maze solving robot from Patrick http://letsmakerobots.com/node/25735

Ok, it's pre-programed but each brain has to be programed in a way. It find it's way through the maze and remembers it. The second attempt will use the learned path to avoid dead ends and go directly to the exit. For my understanding, this is a simple AI.

Medium and advanced AI is just more advanced, faster and more  versatile. It can learn in different ways, make decisions which of the learned date to use to get the task done in the most efficient way. A chess program is not AI, it's just a pre-proigramed pattern. If the chess program adapt to the opponent and learn how he is playing to response then it might be close to an AI. For example, if the human player is using mainly the horse that the AI learn that and develop a strategy to take out the horse, block it or protect his figures from the horse. I am sure something like that is already included in the chess program but yeah...just my 2 cent

thanks for the wonderful answers! 

AI = Memory + Behavior(Memory) + Emergence(Behavior(Memory)) + Magic

Memory can be an abstract data-structure that gets filled with data on events from the environment.
Behavior are pre-programmed algorythms or a set of rules that are executed as an action or reaction to the environment.
Emergence is the result of the combination of the behaviors.
Magic is the unknown that makes the difference between a rock or metal to a flower or human.

So one could say that AI is a function of memory:    AI(Memory)

 

Testcase. A test case for AI could be: If the robot makes progress on what it does by exclusively using the own memory then this robot has AI.

Example. You could use a floor cleaner as an example. If this cleaner can lean where to clean and how to get there, is it then an AI cleaner?

Some of the best work in AI comes from google. Whan clumsy tpyist like me serch... It still understands what I mean. lol.

I think AI is a moving goal. What we once would have considered incredulous is now taken for granted. I agree that behaviours emerging that were not programmed is an acceptable measure for AI but also like to consider any form of automated/algorithmic  thinking as AI in a simpler sense.

Sophisticated AI is after all only a combination of algorithms which are each automated thinking. A calculator is an example of automated thinking and while it is not by itself a favourable example of AI, mathematical calculations are a necesary part of all algorithms.

On the other hand as mathematics originate from the human mind then perhaps making a machine calculate is a form of Artifical intelligence.

Just look at how manufacturers of some products use the word "intelligent" in their promotion of their product. Most of the time this is purely a formula computed to adjust some value in the system.

Whenever I think about this question I find it difficult to pinpoint exactly the right answer. We all have our favourite robots from movies, books and cartoons with personality. And so I guess the question of consciousness comes to mind. Is it possible to create a conscious robot and is this the ultimate when it comes to AI?

but i would say that ... im a biologist.

Take DNA for example - its essentially a four part code (ATGC) [rather than digital 01] leading to 3 part codes (Amino Acids) [letters] which when combined in the right order (proteins) [commands] can create interacting structures (cells) [programs] - these small structures can then be combined to create larger structures which can interact (organs) [large programs].

so why mention DNA? where do you think the human brain comes from?

Human inteligence comes from a two fold evolution - first genetic then life experience - we call it nature and nurture.

The reason why DNA works is because it has evolved to integrate mistakes into its development - from the grass root Amino acid coding to the muscle development and brain tissue organisation. and it doesnt stop there the brain is constantly reorganising itself based around a central pattern.

we like to think of ourselves as non pattern following entities but frankly we couldnt be further from the truth - were just so damn complicated it can be difficult to see the patterns without focusing on the tiniest detail.

even our celebrated conciousness is simply a REALLY big program doing exactly what its supposed to.

the turing test is the perfect example of our arrogance in this field - if it can fool us it MUST be inteligent.

so - how to define artificial inteligence?

its the insertion of a random element within an organised structure which allows evolution of the structure not based upon its origional design.

 

i think... ;)

 

dom