Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
- Joined Jul 2022 | Status: Trader | 477 Posts
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
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
- Joined Jul 2022 | Status: Trader | 477 Posts
Theory of Everything is Liquidity...https://www.youtube.com/@FxArt.Trader
- Joined Jul 2022 | Status: Trader | 477 Posts
{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