primaryobjects / ai-programmer Goto Github PK

This project sounds amazing! But due to my lack of knowledge on this, I have no idea what to do. Like what do I need to install or do? To get it working I would love feedback this is probably my first time using A.I I am more of a malware type of gal but I don't get

All love

Cherri

Hi, I had a look at your project. I find the idea interesting. I got a bit bored waiting for it to run so made some performance improvements, which you can see in my fork:
https://github.com/GilesBathgate/AI-Programmer

I removed some stuff you probably wanted so its not likely you want to merge it. But just thought I would let you know. I reckon its about twice as fast (a test that took 30s initially took 14s afterwards)

But have a try.... or not, up to you ;)

I just want to start off by saying this is fascinating, great work.

I have been developing software for almost 20 years but have only just started to scratch the surface on AI.

There is a ton of useful information in this project that has been very helpful to me, so thank you.

I did notice a couple of things regarding the scoring system and elitism.

I think scoring is by far one of the best methods to increase performance, especially if performance itself is also what is being scored.
In other words, if two genomes have a score of 700 but achieved that fitness in different ways, then a secondary score for performance or time to execute could drastically enhance scoring and reduce the time it takes to test, allowing for more generations to be created and tested.
This additional performance scoring could also be used to dynamically determine max number of applicable loops, instead of hard coding a max.

Regarding elitism, by definition it is supposed to ensure the best genes are retrained in case mutation has a negative impact, in essence it provides stability. Looking at the code however it seems the last two genomes of the current generation are being added to the next generation. I could be missing something and probably am, but shouldn't this be picking the two best scoring genomes of the current generation to be copied into the next generation?

Thanks again for the great code and articles!
Nick Hanshaw

wasn't really sure where to leave you a msg so I figured here was as well as anywhere else.
2 noob thoughts.

1. Im wondering if the GPU was used (if possible) would this run much faster due to a great number of small operations being used repeatedly.
2. If there was a second fitness rating that was only run after a new best primary fitness was found that had a max value of say 750 where 250 is awarded for a successful compile 250 awarded for no infinite loops and the remaining was rewarded for every char under 250 utilized. after reaching 500 or the first rating drops, the best match that did not harm the first rating is allowed to resume finding matches. at the end you can have an option for finding more optimized second ratings.

Not sure if that second option would harm the randomality as the extra compiler info might lend to it however it could still be used after a successful completion for finding optimized code.

sorry for the bad English its not my strong suit.

I want to know its basics, like to put AI in my model

What's the license for this code?

I think this project is really interesting and wanted to modify somethings(like adding cudafy) but I was wondering if you had given any thought to adding a license to the project? Thanks!

Hello,

I tried to reproduce your work in the Julia programming language for my master thesis.
I managed to get impressive results and very good convergence for all the most simple tasks that you covered (string, addition, countdown, logic, ...).
The only one that is resisting me at the moment is the program that simply prints "hello". This looks quite stupid considering that "Hi!" and "reddit" work great.

Here is my idea of why it fails :
In the beginning of the GA, I quickly see patterns appearing in the code that look like this : "100+[++++..]"
Indeed, because of the strange nature of the word hello, this program gives a fitness that is only 4 points away from the perfect score. The problem is that this structure is appearing so fast, and has such a good fitness, that it conquers the whole population very quickly, and the algorithm gets stuck in an evolutionary dead-end.

I don't know if you are still invested in this project, but if you have any insight that could help me I would very much appreciate it !

Thanks,

Jonas Gillain
Royal Military Academy of Belgium

what scenarios of it in reality? Whether It can be used to generate code for common business, like E-commerce website or webpage.