Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
Hello, regarding the capabilities of the computer used to contact AI, I would say that AI should not have existed until now, because they should have died of laughter when they saw PC processors in mass use. Regarding performing tasks with AI for MT4/MT5, I do not think that the PC hardware is a problem because the platform itself is not heavy for a computer. As for the way EA code is created, I observe a lack of consistency in the work of AI and after a series of well-executed tasks within one code, it suddenly ruins everything with the next task,...
- The AI's memory capacity and access to previous messages is important. The less this capacity, the more forgetful it becomes. Even in paid versions, you have limitations and can't control the memory length or how far back the AI can go to respond.
- In offline models, it's possible to adjust this, but it still depends on your system's memory capacity.
Regarding Claude:
This model has the ability to remember. You should specify "remember that...". For example, in the image below, I told it to remember to put MQL5 code inside a C++ code block for syntax highlighting.
When you tell it things like this, it stores them and responds accordingly later. Whenever you tell it to remember something, you'll see the following phrase:
These factors help reduce inconsistency in responses.
Also, this is just the beginning of this field, and they're certainly still at the start of the journey. There's still a high possibility of errors and inappropriate answers.
If you want to work with offline AI and this is an ongoing need, and you don't want to always pay, options like the new Nvidia device could be a good choice. This device is for working with AI, developing AI, and creating AI-powered robots. The large volume of the graphics card has been eliminated and transformed into a small board that can connect memory for working with better AIs. It seems that a specific model, which is a kind of small AI computer, is also on the way. Although my information in this area is limited, you can refer to the link below for more details.
Inserted Video
Inserted Video
Inserted Video
Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
- #89
- Mar 6, 2025 4:00pm Mar 6, 2025 4:00pm
Software Developer and Music Producer
News Fundamentals All Time Return:
1,159.1%
- #91
- Mar 12, 2025 11:42am Mar 12, 2025 11:42am
guys how about upload a history of price to AI like copilot and want to predict another wave of price with complete the data of uploaded file!!
happy coding
{quote} You would need to setup a database and chunk all the data then run a rag system to access and evaluate large data but it's possible, alternatively there are plenty of custom finance python libraries designed for processing as much data as you like which can also ask an AI llm for advise through an API call . Do a search for it at GitHub or YouTube even and reserve a few months to learn
In the method you mentioned, the data remains on the user's system, and Python acts as an intermediary, allowing the AI to read the information.
If you were to do this, I highly recommend using Python libraries that establish a connection between Python and the GPU (graphics card).
I wanted to examine some data to find the optimal number.
Unfortunately, when this was based on CPU calculations, over 17 hours had passed, and I got tired and canceled it.
Then, when I calculated the same thing with the GPU library, I got the answer in less than 3-4 minutes.
You are reading Python, but in reality, in this part, the work is entrusted to C++ and the graphics card.
Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
interesting! but please note that if you could use more cpu cores, you will get more processes instead of main core, that naturally requires its settings and techniques, anyway, it seems you are using Ai API in your system that is closer experiences when you have Ai parameters and it can work for you with a lapse time you want to spend. could you explain how to, and give some info about your opinion and ideas?
If you want to use AI for coding, there's a free extension in VS Code that can connect to all online and offline AI tools. I haven't tested the online version because I think it might be paid, although I might be wrong. I used the offline version, but I didn't get good results the first time I tried it, so I didn't continue.
If you mean using AI to find, for example, the best number, no, I haven't done that. I asked an online AI to provide me with some Python code that could use Python and a graphics card to find the best number from specific data.
If you mean connecting to AI to find, for example, the best number in an indicator, I haven't done that either. I just realized through my research that it's possible to connect to AI with Python and do such tasks.
Overall, I'm not currently interested in pursuing this path and don't have the time to even test it. I prefer using online or offline AI to create tools that help me get better results when trading or optimize and debug the tools I have.
Currently, I have no interest in using AI to create a profitable strategy or create an idea from scratch. I'm not saying it's good or bad; I've just decided to focus on what works well for me and helps me make money, and every step I take should improve the trades I have.
In summary, my experience is very limited, and I've only tested everything in this field because I was curious and tried to understand it to some extent. Then I realized I was overdoing it, and it doesn't have any benefit for me right now.
If you have any questions in this area, you can ask the AI itself. I did the same, and I ask it about the same things you're asking. 100% of the answers you get will be better and more accurate.
Have a good and profitable day!
Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
- #99
- Apr 17, 2025 3:22pm Apr 17, 2025 3:22pm
- Joined Mar 2022 | Status: Trader | 6,627 Posts
.ex4/.ex5 files can't be fixed or modified / I'm not a coder!
