Getting started in Artificial Intelligence?

Question

Hi, I am an undergraduate computer science/software engineering major who has an interests in AI (especially anything concerning artificial general intelligence). I am particularly infatuated with research and theories concerning creating an AI with capabilities similar to those of a human (which I realize is a LONG ways off in the future). But I do like AI in general from what I have read.

I already do a lot of reading and researching on AI, but it is usually just me aimlessly floundering for things concerning the topic. I would like to know how to REALLY get started learning about AI as a discipline. What are the fundamentals? Foundational theories and anything else I need to get started (apart from what I will inevitably learn in my undergraduate studies, I want to learn about the things that I might be taught in grad school if I can or anything else of interest)?

Thanks for your time.

Answer 1

Engineers refer to this topic as "embedded systems". These consist of computer chips which function as special purpose computers which are then used to control mechanical devices. The devices can be as simple as microwave ovens or as sophisticated as industrial robots, or the cockpit controls on a commercial airliner.

These chips contain what is known as an EPROM. This is where the chip stores programs downloaded to it. Things such as microwave ovens have EPROMS whose instructions are literally burned permenantly onto the chip with UV light. Programmable chips are hooked to computers and the software is then downloaded to them. The BASIC Stamp and the ATMEL chip is programmed this way, and both can be purchased as complete kits ready to be hooked to a PC.

Embedded system chips recieve inputs and generate outputs through "ports" which are addressed through programming statements contained inside the EPROM. The inputs can be digital signals from VCR buttons, or perhaps various sensors which detect radar waves, microwaves or infrared light. The output signals frequently drive what are known as SERVO motors. These electric motors have rotors which turn precisely in relation to a series of digital signals. This allows them to be controlled absolutely precisely.

The programming itself is usually a form of assembly language. All computer chips contain instruction sets consisting of a series of numerical codes. These are integers between 0 and 255. The computer reads long strings of these integers sequentially to determine what it is to do. Computers understand it perfectly, but humans do not. Assembler is an attempt to make it more like english. The code is a series of three letter "minoics" which correspond to the instruction set of the particular chip one is programming. The instruction set is also very simple, meaning complicated behaviour is going to require quite a lot of programming in assembly language. The C++ language can also be used to make programming faster. It looks more like english and is basically a library of pre-written assembler programs which handle things like printing text strings.

Robotics and Artificial intelligence are still largly undeveloped fields in engineering. I tend to think it is because computer chips are still not dense enough to contain the number of components necessary for even simple insect reflexes to be faithfully reproduced. Neurons are smaller than the transistors on silicon chips and they are far more sophisticated than individual transistors. Neurons are much more like entire PC's when it comes to computing a series of many inputs and then deciding whether to transmit a signal. One other problem is that nervous systems are self programming and computers are not. A bonified self-programming algorithm would be a great step foreward in robotics.

Answer 2

Yes, AI is not tied to any specific language. The most fundamental concept in AI is searching. Most problems in AI involve searching for solutions. Familiarize yourself with trees, Depth-First-Search and Breadth-First-Search. Next, familiarize yourself with graphs. This will lead into graph searching algorithms (A* and Dijkstra's) and heuristics. At this point you'll be able to write path-finding algorithms for robots and video games. This leads into Game Theory, and min-max trees. Master these, and you'll be able to write an unbeatable chess/tetris/tic-tac-toe/whatever opponent. Finally, you'll learn how Neural and Bayesian Networks simulate brain activity. These will teach you how to make computers learn from examples, and from their mistakes. At this point you're basically re-creating the human brain. =P

Answer 3

AI, is still very new.
just learn to program well.
Then learn to program with code that changes variables by inputed sensor data.
then learn to use self-modifying code.