Hi gang!
Project ARIEL is unfortunately going to have to go in a different direction for the time being. At the moment, I'm looking into implementing the project as an online chatterbot using the haptek interface. Once I find somewhere to host the bot and get it coded (which should be much easier than the prior goals for ARIEL), I'll post a link here.
There are several reasons for this:
1) I'm using the chatterbot to learn more about the current AI field.
2) I'm using the chatterbot to learn the haptek interface.
And last but not least:
3) After ALOT of reading, studying, and contemplating, I've come to the conclusion that what I want to accomplish will require more than software. I've been reading up on this almost non-stop for several days (one tiny benefit of disability is free time). The software that I would need to build would take more processing power than most desktop machines today can handle. Unless I suddenly win the lottery, there's just no way I can afford to build the necessary hardware. At this point, I'm thinking that the only way we're going to get past the chatterbot stage is with distributed hardware and some serious programming. I know enough about distributed hardware to get myself into trouble, and enough programming to do some simple software.
Here's the reasoning behind my thoughts:
The human brain and consciousness are more than simple software. The brain is a distributed network of hardware, with interfacing software. Each individual piece of hardware is running a simple process, which it shares with other pieces of hardware in it's cluster. Each cluster then collaborates with other clusters to form the "big picture" of understanding and reasoning. The sheer number of processes running at one time will require a much larger processing capacity than what a current desktop computer can handle - and this would just be for true understanding of human language. Add in other processes, such as image recognition and cognition and the processes grow exponentially.
Don't misunderstand me - I still think that a true AI is possible using today's technology. For the time being, however, it's beyond my means, and most other people's means. To truly get something like this off the ground would require large sums of money and years of collaborative work among programmers, neuro-scientists, and psychologists before we can get something approaching true Artificial Sentience.
The ARIEL Project is not dead!
I still plan to keep moving forward with the project, but at a much more realistic pace. First, I need to start with fully grasping the current AI practice of mimicking human conversation and pattern-matching. From there, I'll hopefully be moving into the areas of building small blocks of processes, one at a time. Just as certain areas of the brain carry out single processes and then feed this information to other areas, each of these smaller blocks will communicate with other blocks of processes, until the whole begins to take shape. Natural language processing will be the first goal of this second stage of development.
I believe at this point, that a single, multi-processor computer can serve as a single hardware block for each process of handling natural language. I'm not talking about super-computers here. I'm talking about relatively inexpensive systems that can serve as a single block or node in the overall network of hardware to process simple functions rapidly and efficiently, then pass the processed information forward to other hardware and software for higher-level processing. A single, dual-core processor may be able to handle a single process. On that side, I'll just have to wait and see how things develop.
Here's an example of what I'm talking about.
Let's say that the user enters the statement, "Dogs are mammals." This input is parsed via a single hardware/software node to determine parts of speech. The subject "dogs" is sent to a lower-level hardware/software node, where it is processed against a database. If the information pertaining to dogs is not in the database, then the word is added and flagged for future processing. The verb "are" is sent to another hardware/software node, where a similiar process takes place. The noun "mammals" is sent to still another hardware/software node where it too is processed. These results are then sent to a higher-level hardware/software node that flags "dogs" as belonging to the class "mammals". The final process is that a final hardware/software node is told that the information is processed, and it in turn notifies the user that this has occurred. The distributed system has just learned the fact that dogs are mammals.
Of course, this information can be added beforehand, so that the entire system already knows this fact before a user becomes involved.
Now, I'm sure some of you are going to say that this is an overly-complicated view of how the process might work. You may be correct, but I'm looking at it from the point of view that human language is extremely complicated. We still don't quite understand how we ourselves understand it. A complicated system requires another complicated system to comprehend it.
My opinion is that by using this distributed network of hardware and software packages, it will allow the entire network to begin to learn and comprehend. Artificial sentience won't take place on a single machine, but on a network of machines and the network will become the sentient entity.
As I said, the ARIEL Project is not dead. Far from it. It is only beyond my financial means and programming capabilities at the moment.
I come from a hardware background (many years in the field), although I am a bit behind on current hardware tech, since I was forced out of the field due to my illness. By taking a step back and re-analyzing the problem, I can take the time to reaquaint myself with the latest hardware and sharpen my programming skills at the same time. While doing this, I can also build a firmer paradigm of how I believe artificial sentience can be achieved.
In the long run, the ARIEL project will benefit from this shift in thinking.
As always, I'll keep everyone posted on how things are going.
Thanks,
Darkwolf
