Lecture
The AI in the project is created by quantitatively empirical development on FoxPro - so it can be called.
Stage 1 - deductive: Semantic Text Analyzer
In general, now at the first stage it is done so that several thousand texts of several hundred or thousand lines each are uploaded to the system (in different areas, for example, in history, or in physics, or in programming, etc.), and so that the system can then reasonably answer questions on this text. In the future, so that questions could be absolutely any within the framework of deductive (unconditional) conclusion. This is of course a difficult task, but it is solved here little by little. I think the implementation period is about 1 year (if I’ll borrow at all with this whole affair).
2nd stage - inductive: Text generator and problem solver, associative and non-verbal thinking
At the second stage, when there are already several million texts in the system and it will “weigh” gigabytes of, say, 200-300, it will be necessary to make an inductive (probabilistic) conclusion in the system, and so that it itself can already generate, to give some lyrics, maybe even poems or novels. And also so that she could solve problems. For example, if a textbook on physics is uploaded into the system, then that it should also solve the problems in physics that the user will offer to it. In general, almost as a person so that she could do everything. The implementation period can also be 1 year.
3rd stage - consciousness: A lively and rational person, a soul
Well, at the third stage it is a launch in the system of consciousness, the formation of a personality, and other questions about this.
***
(PS: I transferred my various “extreme” statements to the “about the author” section.)
29/04/12
Why on FoxPro? I tried to write AI on a bunch of MS SQL Server + MS Access. That is, MS Access used to organize the interface with the person, and MS SQL Server for data storage and calculations. MS SQL Server has such advantages as complete transaction processing, reliability, large amount of guaranteed data to be processed, etc. But I was tormented by speed, somehow slowly. Then I just took and rewrote everything on FoxPro, one to one. And it turned out that everything works just 30 times faster. What used to take from me, let's say half an hour, on FoxPro began to be processed in 1 minute. And for AI it is very important, much more important than all sorts of transactions and the concept of the server. After all, the user is still one - it's me, and so it will be a very long time. But it is necessary after all so that at least something really works. Therefore, I chose all the same on FoxPro.
I also tried to write AI in C ++ & MFC earlier. But, however, I was tortured with garbage collection in the RAM, it still accumulated in me, and secondly - the speed of such a system is determined primarily by working with the database, and only then by working with the OP. And with the database I had a connection through ODBC, of course, which is rather slow. And FoxPro directly processes the DBF-ki, therefore and faster. Yes, and writing in C ++ is still quite laborious, although I remember getting used to it somehow, the concept of the Universal List Container (UIC) helped me in particular, I invented it myself. For AI in my version, the thing turned out to be very useful. But still decided to switch to FoxPro. It would be possible to write in C ++ the entire AI using the fact that modern OP computers are already very large - 2 GB, and then 4 GB is already the norm. And it is already possible, in theory, to put the entire database of AI AI in OP. Previously, the hard drives were allowed 1.6 GB, and that seemed a lot. But I think that in this case there will be problems with speed, if you use the usual scan in the list in the OP when searching for data. Certainly indices, the so-called B-trees, will be required. I tried them once to write and debug myself, and frankly I was tortured. This, I think, is a separate and very complex topic. There is no time and effort. Therefore, I think instead to get it later, or write myself, if there is no such, a certain Windows driver, so that it simply creates a virtual disk of 3 GB in the OP-memory, and copy and run Eva on it completely. So the indexes that are in FoxPro, will be perfectly worked out, and the performance will be high. Well, here is such a shorter situation. Amen.
10.25.2012 07:09
In the course of the last few days there was a strong enough thought that Eve could write as much in Assembly language. The reason is to be completely independent. And indeed, if you adapt, you can write pretty fast there. At the same time, the whole AI was in general in RAM, the benefit is that now it has already grown up to 4 gigabytes. This is a lot, it would be enough for the whole system, if we consider that the first AI was launched in the late 70s on a computer that had 4 MB of operational and 30 MB of external (on magnetic tapes and drums) memory. But writing on the Assembler itself is still impossible. It is necessary to make a macro assembler, of course, and even more likely a type C superstructure. The same variables, because you won’t write cell addresses on a piece of paper, it’s better that the computer itself did it, and so on. Thus, it is necessary to go all the way that others have already gone. And because of the lack of funds, it is quite expensive to go in another round, and it’s also possible that I’m almost as close as ever with this AI. If I download the final paradigm system of 90 thousand words that I already have, and finish the subsystem of issuing simple answers to the texts, then this will already be something. I'm feeling it. And if you write in Assembler, then adding a bootloader, you could get your operating system at all, which would immediately start up when the computer was booted, bypassing Windows. It would be very interesting, but unfortunately I don’t have the strength to do such a large amount of work. So we'll see. Amen.
18/08/13
It is necessary to make the system accumulate knowledge from the texts. And made on the basis of them the logical conclusions that will be required by man.
27/08 / 13-24 / 12/13
All the same, more and more again I come back to the 7-level model of AI. Indeed, the PSOE could bring real benefits now to me, for all the programs that I now have to develop and maintain. This 7-tier model looks like this:
Each upstream level is based on the capabilities of previous levels. Convergent approach, of course.
That means. And the PSOE is essentially the way to bring the FoxPro relational database management system simply to the level of hierarchical database management systems that were once classified as military. Here is such a garbage, damn nafig. Wherever you throw, everywhere wedge, e-mine. Tree sticks. :) Amen.
Creating artificial intelligence (AI) - is the main task of mankind.
All other tasks will be solved by the AI itself.
Maybe AI can solve the problems of poverty, disease, scientific problems, etc.
Maybe he can even give us immortality.
AI is the intellect created by the human intellect.
He will be able to think, feel, do science, work in various fields, etc.
He will possess consciousness, self-awareness, memory, etc.
Now there are many approaches to creating AI.
My opinion: of all these approaches, only one is promising.
It seems to me that in order to create an AI, you need:
STEP 1. Thoroughly study the workings of the human brain, i.e. study how the human brain works.
This must be done at any cost.
It is believed that the human brain works by 5%.
It seems to me that, accordingly, the person has divined the brain by 5%.
We have to figure out how the human brain works 100%.
So that nothing remains unexplored.
Moreover, it is necessary to take into account both anatomy and physiology.
The biggest mystery here is how consciousness, self-consciousness and thinking arise in the brain.
STEP 2. Create an AI in the image and likeness of the human brain.
There is such a science - bionics.
According to Russian Wikipedia, this is “applied science of application in technical devices
and systems of principles of organization, properties, functions and structures of living nature. "
So, we will create AI according to the principles of bionics.
That is, the first AI will be a device designed as
to copy the principles of the human brain.
But in the future we must not just make a copy of the human brain.
We need to improve it.
Perhaps the AI of the later generations will be much larger than the human brain.
And then "quantity will turn into quality" (Hegel) - (recall how quantity
turned into quality with an increase in the monkey's brain: it turned into a man)
Or maybe AI will no longer have a human brain,
but we will come up with some improvements.
When a genuine AI is created, the further development of AI, the emergence
new generations and new versions of AI - all this will be dealt with already created AI.
The same AI will solve ALL the problems of mankind: treatment and vaccination against all diseases,
overcoming poverty in the third world countries, achieving immortality, scientific resurrection
of all the dead (Fedorov), the AI will solve all unresolved scientific problems,
tell us about the meaning of life, tell you how the world works,
will tell all that the human mind can know about God,
will find a source of infinite energy, will create a perpetual motion engine (not just a river and a turbine),
help us fly to the stars and meet aliens,
will AI create new versions of AI (which may already be arranged according to other principles,
not according to the principles of the human brain),
AI will be engaged in creativity in all imaginable and unimaginable areas, etc.
These two steps above will lead mankind to the creation of AI.
All other ways are unpromising.
Evolution has worked for billions of years to create a human brain.
Billions of species of living organisms appeared and disappeared, gradually becoming more complicated in order
so that the human brain appears in the universe.
And we want to create intelligence from scratch.
This is not true.
The Universe has already worked on the creation of intelligence, and it does not matter
did God (or something like that) participate in the evolutionary appearance of the mind?
The main thing is that the universe created the mind - the human brain.
And on the basis of the device of the human brain, we will create AI.
In addition, since the creation of AI is the main task of mankind,
I believe that only science has the right to exist
which helps humanity bring the creation of AI closer.
The only problem: how to force AI
Comments
To leave a comment
Experiments and Results
Terms: Experiments and Results